How We Win

Why Now — Intro Summary

The case for acting now is not a manufactured deadline — it is a base that is already turning. Two enterprise renewals in the last ~2 years (a ~$200K-relationship medical-device client and an automotive client) were retained only after answering, at the renewal turn, "are you going to be investing in the modern workflow we're being called on?" — and ~24 months of sales-qualification calls have ended in "hard pass," with prospects reporting "nothing changing in your platform" (per eric-meeting-2026-08-07 §2 — private transcript, provisional). The company weathered a storming-to-normalizing transition and then stalled there — retention held, but the step past "normalize" into performance never landed, and rehire capacity is now predicated on growth that the current product is not producing (same source). Acting now is the answer to a clock the clients and the market are already running, not one this plan imposes.

The credibility of that claim rests on what it does not assert. There is no two-month mandate behind this urgency — that "deadline" was a workspace framing, never the CEO's words, and it has been retired; the confirmed horizon is ~9 months ("winning the next battle," CEO 2026-08-07; reconfirmed 2026-08-10) (per eric-meeting-2026-08-07 §8; ceo-framing-cluster calibration). And there is no invented runway figure behind this urgency — the financial pressure is real, but the hard number stays a named flag below, not a fabricated one. The pressure is real precisely because it is sourced to client voices and dated renewals, and honest because the gaps in the timing case are named, not smoothed over.

That timing — a base eroding at each renewal turn, a prospect pipeline that has gone quiet, and a confirmed ~nine-month window to answer — is the pressure behind the ask the prior section makes; this section establishes the runway, urgency, and competitive context that make "now" the only honest answer, before handing to the market context that makes it urgent.

The Renewal Clock — Clients Are Already Running It

The single strongest timing signal is client-voiced, dollar-denominated, and time-bounded. Two enterprise renewals in the last ~2 years — a ~$200K-relationship medical-device client and an automotive client — were retained only after being able to answer, at the renewal turn, "are you going to be investing in the modern workflow we're being called on?" (per eric-meeting-2026-08-07 §2 — private transcript, provisional). That ~$200K medical-device renewal is now corroborated by name: Stryker Corporation closed-won at exactly $200,000 (01/13/2026) — the largest direct deal in the FY26 book — and the "automotive" renewal likely maps to Ford Motor Company's $90K open renewal (per fy26-financial-baseline §3 — internal projection, provisional). In both original cases the clients kept their logins but ran PowerPoint and Excel alongside the platform, a visible hedge against a tool they had not yet decided to trust fully. This is the closest thing the evidence offers to a real deadline: each renewal cycle is an implicit clock that turns on the modernization question, and each turn we do not answer erodes the base a little further.

The risk side of that clock is now concrete, not just qualitative. The same FY26 book records $128K of attrition across seven accounts — Enterprise Transformers ($28K), Charter Next Gen ($25K), EON UK ($23K), Fluence ($18K), Patriot-Hc ($14K), Vinarchy ($13K), dental bauer ($7K) — every one flagged High-risk before the loss (per fy26-financial-baseline §4 — internal projection, provisional). Closed-renewal retention sits at 82.7%: the marquee renewals land, but the base erodes at the margins. Each renewal turn the platform cannot answer the modernization question for is a candidate for that attrition column, not merely a delayed decision.

That renewal pressure is corroborated, not isolated. The same source reports ~24 months of sales-qualification calls ending in "hard pass," with no new modules shipped and executives reporting "nothing changing in your platform" — the prospect pipeline has gone quiet, skeptical-quiet turned to indifferent-quiet. And at the company level the storming-to-normalizing transition (cost-cutting and habit-fixing, then client-and-employee retention) has stalled at "normalize": the step into performance never landed, and the capacity to rehire is now predicated on the growth only a renewed product is expected to produce (same source).

What is proven here is the pattern — renewals turning on modernization, prospects hard-passing, growth stalled — and the cadence is now substantiated too: the specific renewal dates and dollar exposure for the next twelve months are filled in 08_renewal-cycle-data-gap from the same FY26 baseline (flagged provisional, not authoritative). The one figure still absent is the hard cash runway, named separately in 07_runway-figure-gap.

Finances — A Factor, Stated Honestly

The timing signal here is financial, and its direction matters more than its exact magnitude. The Owner's intent (Thom — Owner & Sole Investor) is endurance and growth, not a short-term harvest: the operating register is "innovate or die," and building something competitive is framed as the way to be more valuable in any future transaction (per eric-meeting-2026-08-07 §3 — private transcript, provisional). A GSI partner's ventures lead has framed the strategic option as "recap, rebuild, win" — recapitalize, rebuild the product, go compete — recognizing the retained client-logo base as the differentiated asset (same source).

The implied clock inside that intent is a funding-drain horizon: the outcomes the CEO names are net-new revenue, growth, and "reduced investor funding drain in 2027" (per eric-meeting-2026-08-07 → M13 module-impact row — private transcript, provisional). That is a qualitative runway signal — Owner-funded endurance has a discernible pressure point around 2027 — not a quantified one. The hard runway figure (months of cash, burn rate, the dollar exposure) is not in the evidence base and is named as a flag in 07_runway-figure-gap (G-M02-RUNWAY).

The honest synthesis is that the financial pressure is real enough and directed enough to make "now" the right read, while the precise number that would let the plan quantify the runway remains a leadership question — one this section surfaces rather than fills in.

Competitive Timing — Structural Drift, Not An AI Race

The competitive timing pressure is real, but it is not the "everyone is adding AI" pressure it might first appear to be. Across the seven-dimension, confidence-tagged matrix, every serious incumbent has converged on the same diligence-stage AI feature set, which makes that AI a commodity rather than a differentiator — the "AI reflex" is table-stakes everyone already has, not a window that is closing (per competitor-feature-parity-matrix). Framing "why now" as "everyone is adding AI" would be both weak (Devensoft would be late to a commodity) and mis-attributed — the "AI reflex" was a workspace framing, not something the CEO asserted (internal calibration).

The honest competitive-timing pressure is structural drift: rivals are modernizing the package around the commodity AI while Devensoft holds. DealRoom — the most dangerous structural rival — has moved to per-deal/unlimited-user pricing that directly attacks Devensoft's seat-license friction, runs a live MCP connector, and ships Looker BI with editable PowerPoint export against Devensoft's image-only reporting; Midaxo carries an IDC Leader badge Devensoft lacks (same source). Each of those is a modernization axis where the gap widens with every quarter Devensoft does not move, and none of them is an AI feature that could be copied late. The one place the matrix shows greenfield rather than convergence is the integration/PMI seam — and that depth belongs to M07's AI thesis, cited here as the opening, not re-argued.

So the competitive "why now" is not "AI is here" — it is "the package is drifting away from us on pricing model, BI, and connectors, and the one uncrowded seam is the one we would have to choose to invest in." The clock is the drift, not the AI launch.

The Horizon — Nine Months, Not Two

The confirmed planning horizon is ~9 months (roughly three quarters), framed by the CEO as enough to "win the next battle" without needing to plan past it, and reconfirmed two days later by the enterprise architect (CEO 2026-08-07; reconfirmed 2026-08-10) (per eric-meeting-2026-08-07 §8 — private transcript, provisional). Delivery inside that horizon is incremental and iterative, not big-bang.

A calibration footnote, because it matters for credibility: an earlier "2-month mandate" appeared in the workspace's framing and even reached the leadership intake questionnaire, but it was a fabricated placeholder, never an external deadline — it traces to a workspace input, not to the CEO's words, and the CEO's later 9-month statement supersedes it (per ceo-framing-cluster calibration; retired 2026-08-10). The 2-month figure is referenced here only to retire it; it is not a constraint on this plan and must not be cited as one.

The one honest caveat on even the 9-month horizon is execution discipline, not the date itself: the actual timeline depends on leadership holding the ~90/10 attention firewall when client urgencies arise, or the reactive-client-work pattern that overran the c.2020 modernization mission recurs and consumes the rebuild window (per eric-meeting-2026-08-07 §6). That execution-timeline risk is owned by M10/M12; M02's claim is only that nine months is the confirmed, lowest-solid timeline input — not that nine months of disciplined execution is guaranteed.

The Timing Mandate — Resolved, Not Re-DecidedDecide

The "timing / mandate" decision this section once carried is already settled elsewhere, so M02 resolves it rather than re-litigates it. The horizon — ~9 months, incremental delivery — is the CEO's confirmed input and has been amended into BTDR-0004 (G4: the capacity-vs-horizon framing, horizon-supersession amended 2026-08-10). The invest-vs-exit question is settled in the opposite direction the word "mandate" might imply: the Owner's intent is endurance and growth, exit explicitly not on the table — a decision that belongs to M12's risk/decision surface, where the exit branch is refuted (per M12 intro). And the shape of what gets done inside the window — the 90/10 attention firewall, the fundamentals-first sequencing — is BTDR-0004's hedge and the roadmaps of M06/M10, not a timing matter.

What M02 therefore contributes on the mandate is a synthesis, not a new decision: the timing mandate is "act inside the ~9-month window the CEO confirmed, against a base that is already turning at the renewal cycle, with Owner endurance (not exit) as the intent." Nothing in that sentence is newly decided by this section; each clause traces to a locked BTDR or a CEO-origin fact. This follows the M05/M06 precedent — sections that establish and synthesize rather than decide — and means M02 produces no new BTDR of its own.

The decisions that do remain open at the timing surface — the hard runway figure, the exact renewal-cycle exposure — are flags, named in the next two components, not silent assumptions.

Flag — The Hard Runway Figure Is Still Absent (Revenue Baseline Narrows It)

The qualitative financial signals — Owner-funded endurance, an implied 2027 funding-drain pressure point, an "innovate or die" register — are all stated in 03_finances-funding-drain and sourced. What was wholly absent is now partially filled: an internal FY26 projection books ~$1.9M of revenue (incl. the Aviator reseller line), with $700.8K already closed-won in Q1 and $128K of attrition, against a closed-renewal retention of 82.7% (per fy26-financial-baseline §1 — internal projection, provisional). That gives the plan a revenue baseline to reason against — it is no longer sizing the base from zero.

What remains genuinely absent — and why this stays a flag — is the number that actually defines runway: months of cash, monthly burn, and the dollar exposure at risk on the cost side. The FY26 workbook is a revenue/renewal projection, not a P&L or balance sheet; it carries no cash position and no burn rate. This is gap G-M02-RUNWAY (per gap-analysis → M02): the financial-pressure signals are now quantitative on the revenue side and still qualitative on the cash side.

This is named as a flag rather than filled in because manufacturing a runway number would be the kind of stance drift the validation protocol guards against — an uncited or invented cash figure in an investor-grade strategy fails provenance (P4) and honesty (principle 4) at once. Closing the gap is a leadership input: the actual runway figure lives with the Owner and the finance function, not in any source the workspace has ingested. Until it lands, the timing case is honest about being quantitative on the client-voiced renewal clock and the revenue baseline, and qualitative on cash runway.

Flag — Renewal-Cycle Dates And Dollar Exposure: Substantially Resolved (Provisional)

The renewal-contingency clock is named qualitatively in 02_renewal-contingency-clock — two enterprise renewals in the last ~2 years turned on the modernization question. The cadence gap that flag once named (no exact dates, no dollar exposure beyond the ~$200K relationship) is now substantially closed: an internal FY26 renewal/upsell/attrition projection puts real dates and negotiated contract values against the next twelve months (per fy26-financial-baseline §3 — internal projection, provisional). The open/projected renewal book — Accenture V&A $350K (09/01/2026, 23% of the open pipe), Ford $90K, Graybar $19K, Fulton Financial Y2 $13K, IBMG $4K — plus the $334K of FY27-exception deals pulled into the FY26 window (J&J $93K, Marmon $92K, Kenvue $78K, Hexagon $36K, NovAtel $22K, Stanley Black & Decker $14K) now gives the timing case a calendar, not just a pattern.

This stays a flag rather than a clean resolution because the baseline is a point-in-time sales projection, not an audited financial statement and not recognized ARR — it blends Q1 closed-won actuals with Q2–Q4 open/projected pipeline, and two of the open accounts (Graybar, IBMG) are already flagged High-risk. Naming the resolution level keeps the timing case honest: the pattern (renewals turning on modernization) is proven and load-bearing; the cadence is now directionally quantified from an internal projection that leadership should treat as better-than-nothing, not authoritative. Converting it to recognized ARR needs the finance-system data deferred to the M13 baseline census.

This is the second of two named flags in M02 — together with G-M02-RUNWAY (07_runway-figure-gap), they bound exactly what the timing case does and does not claim. The renewal-cycle flag is now substantially resolved; the runway flag narrows but stays open on the cash/burn side.

Market & Competition — Intro Summary

The M&A-lifecycle software market is worth winning, but not because it is empty — it is worth winning because Devensoft already holds ground the crowd hasn't reached. The nearest sized proxy for the category, the virtual-data-room (VDR) market, runs roughly $3.3B today toward $5.6–6.0B by 2029–2031 (per enterprise-market-opportunity — published estimate, not Devensoft-verified), with enterprise deals concentrating about 71% of that spend. Every serious incumbent — Midaxo, DealRoom, Datasite, Intralinks — has converged on the same diligence-stage AI feature set, which makes that AI a commodity rather than a differentiator; Devensoft's real opening sits one stage later, at the PMI/synergy seam where no competitor has shipped AI at all (per competitor-feature-parity-matrix).

This read is grounded, not asserted: the sizing traces to published market-research figures (Mordor, MarketsandMarkets), and the competitive picture rests on a seven-dimension, confidence-tagged matrix built from direct vendor-site captures and per-vendor profiles, not secondhand analyst summaries. Where a figure is a published estimate rather than something Devensoft has verified, that is flagged inline rather than presented as fact — the credibility of this section comes from naming that boundary, not hiding it.

That urgency — a market every rival is racing to reposition around AI, right as Devensoft's own growth has stalled — is exactly the context the prior section establishes; this section turns that urgency into a concrete map of where the crowd is and where Devensoft's opening actually is.

Market Sizing

There is no clean published total-addressable-market figure for "M&A lifecycle-management software" as its own category — the only sized proxy available is the virtual-data-room (VDR) sub-segment Devensoft's category sits inside. That proxy puts the VDR market at roughly $3.3B in 2025, growing to an estimated $5.6–6.0B by 2029–2031, a 10–18% CAGR range (per enterprise-market-opportunity, citing Mordor Intelligence 2026 and MarketsandMarkets 2024 — published estimates, not Devensoft-verified).

Two details make this figure more than a headline number. First, enterprise-tier deals account for roughly 71% of VDR revenue, meaning the dollars concentrate almost exactly where Devensoft already competes rather than in a segment it would need to newly enter. Second, vendor concentration is only moderate — the top five vendors hold an estimated 55–60% of the market — so this is not a two-horse race Devensoft is arriving late to; there is real room to take share. The regulatory backdrop reinforces the timing: EU DORA (effective January 2025) and rising data-sovereignty requirements are pushing buyers toward auditable, compliance-grade M&A tooling, a tailwind that plays directly to Devensoft's audit-log strength (per enterprise-market-research-external).

The Competitive Set

Four vendors define the crowd Devensoft competes against, each occupying a distinct position. Midaxo is the closest like-for-like rival — a full-lifecycle platform carrying an IDC MarketScape "Leader" badge (2025) that Devensoft does not have, with the broadest native AI in the set, including cross-project synergy-flagging (per competitor-profile--midaxo). DealRoom is the structural pricing attacker: a per-deal, unlimited-user model that directly targets the seat-license and heavy-onboarding friction Devensoft and Midaxo both carry, paired with the first live MCP connectors to Claude, ChatGPT, and Copilot (per competitor-profile--dealroom). Datasite and Intralinks sit in an elite, VDR-heritage tier serving investment banks, mega-PE, and law firms — a different buyer than Devensoft's integration-management-office champion — but both define the ceiling for AI maturity and security certification the whole category is converging toward (per competitor-profile--datasite, competitor-profile--intralinks).

Across all four, key AI capabilities have converged: document summarization ships in every competitor's product, clause/risk extraction ships in some form at all four, and redaction ships at the VDR-heritage tier — Datasite and Intralinks — though neither Midaxo nor DealRoom has shipped it; MCP connectors are already live at DealRoom and Datasite. That convergence is the decisive dynamic — it confirms the diligence-stage AI everyone is racing toward is table-stakes, not a differentiator, and Devensoft currently ships none of it (per competitor-feature-parity-matrix). Underneath that, the research also names Devensoft's actual #1 competitor as the status quo — Excel, PowerPoint, and SharePoint — since users revert to it under friction, not to a rival platform (per enterprise-market-opportunity).

Where Devensoft's Opening Actually Is

Devensoft is not behind on differentiation opportunity — it is behind on shipped table-stakes. The feature-parity matrix shows every competitor concentrating its AI investment at the diligence stage; none has shipped AI at the post-merger-integration or synergy-tracking stages, which is exactly Devensoft's historical center of gravity (per competitor-feature-parity-matrix). That gap is the opening: PMI/synergy-stage AI, reachable through the existing Deven Connect API surface, is genuinely greenfield rather than a race to match a shipped rival feature.

A second opening sits in trust and compliance rather than features. Devensoft's audit log is a demonstrated, DORA-aligned moat today — no named competitor ships a comparable regulator-ready audit trail. The tenant-isolation hierarchy (ADR-0040) points at a second, related moat, but it is still target architecture rather than a shipped control, so it should be read as provisional, not yet real (per competitor-feature-parity-matrix, which flags both as largely unmonetized) — procurement collateral the sales motion isn't yet using as a selling point. A third, adjacent opening is pricing: DealRoom's per-deal, unlimited-user model is actively attacking the seat-license and onboarding friction Devensoft shares with Midaxo, and a self-serve, per-deal product tier would blunt that attack on the enterprise side while doubling as a wedge into the lower-middle-market diversify play (per enterprise-market-opportunity §8–9). None of these three openings — integration-seam AI, moat monetization, re-packaged pricing — requires competing head-on where the market has already commoditized.

Open Decision — Where To DifferentiateDecide

Part of this question is already settled. BTDR-0004 (Accepted, 2026-08-06) locks the AI lane of differentiation: a layered posture where reporting-AI is a conditional table-stakes floor, Deven Connect + BYOM satisfies the "we have AI" ask, and integration/PMI/synergy-seam AI is the deliberate win-tier bet — the one place in the matrix no competitor has shipped AI. That BTDR also locks a model-governance rule (BYOM-mandated at the enterprise tier; a hosted model permitted only inside a standalone lower-middle-market product).

What remains genuinely open is broader than the AI lane, and it is a leadership call, not an architect's. Three differentiation levers are still undecided: whether Devensoft leads commercially with the audit-log/isolation-hierarchy moat as procurement collateral, how aggressively to re-price or re-package against DealRoom's per-deal attack, and whether a self-serve product ships as an enterprise PLG wedge, a distinct lower-middle-market Diversify product, or both. Each is an appetite-driven commercial choice that depends on how leadership weighs evidence-certainty against growth-alpha and how much execution risk the ~nine-month planning horizon can absorb. This is therefore an open decision-ask for Eric and Thom: the evidence ranks fixing fundamentals and reporting as the highest-ROI, lowest-risk move (per enterprise-market-opportunity §8), while moat-monetization and re-pricing carry higher upside but compete for the same constrained capacity inside that ~nine-month window — so leadership needs to choose the sequencing and resourcing priority among these three levers before M08/M10/M11 can commit a roadmap.

Fog — No Standalone TAM

There is no clean, published total-addressable-market figure for Devensoft's actual category — full-lifecycle M&A software spanning pipeline through post-close integration. The $3.3B–$6.0B figure cited elsewhere in this section is a proxy ceiling drawn from the narrower VDR sub-segment, not a sized estimate of Devensoft's own serviceable market, which is a broader-scope, narrower-buyer subset of that proxy (per enterprise-market-opportunity).

This is named here as an unresolved gap, not estimated ourselves. A commissioned, analyst-grade sizing (IDC or Forrester-style) remains an available follow-up but is treated as optional rather than blocking, consistent with the speed-first research posture adopted for this workstream.

Fog — Residual Gaps In The Competitive Map

The competitive matrix underlying this section is deliberately speed-first, and several cells remain thin rather than confirmed. Midaxo's and DealRoom's API depth is unverified — neither advertises a priced, commercial OpenAPI product the way Deven Connect is packaged, so any comparison there is inference, not confirmation. Pricing for Midaxo and Datasite is contact-sales-only; DealRoom is the sole vendor in the set with a public pricing model, which limits how far the pricing comparison in this section can be pushed.

Devensoft's own security and AI-governance posture is itself still provisional in this matrix — internal architecture decisions describing tenant isolation and BYOM are targets, not yet shipped, and certifications are asserted "by inheritance" rather than independently verified. Whether competitor MCP connectors support read-only access or full write-back is also unconfirmed across vendors, and Intralinks' public review footprint (G2/Capterra) has not yet been captured, unlike the other three vendors. None of these gaps are resolved here; they are named as open items for a future research pass (per competitor-feature-parity-matrix — "Biggest Gaps" queue).

Customers & Segments

Devensoft's enterprise base is bought by a well-evidenced four-persona buying center led by an Integration Management Office (IMO) champion, resisted at the gate by IT/security, and defended by fixing the fundamentals those buyers already ask for — not by chasing AI (per modus-enterprise-insights; per demo-dwight-customer-voice). Growth beyond that base has two distinct paths: net-new enterprise wins on the same buying center, and a Lower-Middle-Market (LMM) diversification play that remains commercially unvalidated. This section is grounded in four independent, cross-corroborating customer-voice streams — 640–1,200 HubSpot support tickets, a 108-issue YouTrack census, 15 Modus enterprise interviews, and a recorded customer demo — that agree on the same buying center, the same top pains, and the same absence of AI demand (per hubspot-customer-voice; per youtrack-demand). Building on the market opening established in §03, the customers who actually transact in that market are enterprise buying centers organized around an IMO leader — the segment this section names, defends, and weighs against the unproven LMM alternative.

The enterprise buying center

Enterprise deals are decided by four recurring personas, not a single buyer: the Strategic Executive (Economic Buyer — VP Corp Dev, Director Corp M&A) who wants exec-ready reporting and "time to synergy"; the Finance & Ops Lead (Champion — Senior Finance Director) who owns adoption; the Integration Manager (User-champion — Director PMI, Global PM) who needs a stable, low-cognitive-load tool; and the M&A Strategy Consultant (Influencer — Accenture) who shapes client-facing self-service expectations (per modus-enterprise-insights, drawn from 15 enterprise interviews spanning J&J, Accenture, DuPont, Hexagon, and others). The recorded customer demo corroborates this shape from the champion side: the IMO leader is "the person who wants the tool," and the economic-buyer role has drifted from the deal's managing director toward the IMO over roughly nine to ten years (per demo-dwight-customer-voice). This four-persona map sharpens the vault's flatter "IMO leader" framing into a buying center with distinct objections — reporting granularity for the Economic Buyer, adoption cost for the Champion, cognitive load for the User-champion — each of which is a named lever for the fundamentals-first defense argued in §05/§08.

The resistance: IT and security as gatekeepers

The buying center's most resistant faction is not a competitor evaluation — it is the customer's own IT/security team. In the recorded customer demo, the advocate names this directly: "some of the most resistant... is the IT team, surprisingly" (per demo-dwight-customer-voice, .vtt ~54:24–54:57), a resistance the sales/demo motion must disarm upfront rather than treat as a late objection. Modus's independent enterprise research corroborates the same gatekeeping dynamic from the delivery side: IT/security concerns surface through the platform's technical-maturity debt (Implementation scored 1/5) rather than as a named persona with its own asks (per modus-enterprise-insights). Because this resistance is procedural — access, permissioning, and reliability — rather than feature-driven, it is directly addressed by the same access/reliability fundamentals work this plan already prioritizes (§05/§08), not by new capability investment.

The Lower-Middle-Market (LMM) segment thesis — provisional

Flagged provisional: this entire thesis is Devensoft-originated ideation research (2026-04/05), synthesized via NotebookLM — market-unvalidated (no fake-door or customer signal yet), so treat every figure below as directional, not confirmed (per lmm-thesis). The thesis targets a $5M–$50M sweet spot within a broader $5M–$500M Lower-Middle-Market band, driven by a macro wave of retiring founders in fragmented industries, and pitches two buy-side-only personas — the solo Search Fund Operator and the Portfolio/Programmatic Acquirer — who are underserved by a "False Choice" between point-solution tools (the Lean Trap) and legacy enterprise platforms like Datasite, Midaxo, and Devensoft itself (the Enterprise Trap, per the research's own framing) (per lmm-thesis). The research explicitly places Devensoft's own enterprise platform inside the Enterprise Trap it proposes to bridge past, implying any Diversify product is a distinct product line rather than an extension of the enterprise platform. The thesis is explicitly conditional on four kill-conditions (Enterprise Trap immutability, displacement-gap volume, infrastructure as force-multiplier, low-CAC community reach) that have not yet been tested against real demand.

Where the demand actually concentrates

Two independent internal-demand streams agree: client pull concentrates on reporting and on access/reliability fundamentals, not on new capability categories and not on AI. A 108-issue, census-clean 12-month YouTrack sweep finds Reporting & Analytics the #1 cluster (18 issues, 16.7%) and Access & Reliability #2 (16 issues, 14.8%, including the only demand-side Show-stopper); rolling in bug/stability and performance, reporting-plus-fundamentals accounts for 51 of 108 issues (47.2%) (per youtrack-demand). HubSpot's lower-bound 640-ticket extraction reaches the same conclusion from a separate channel: reporting is the single largest capability-demand cluster (~86 tickets), and basic access/authentication friction is the loudest recurring pain (per hubspot-customer-voice). Both streams converge on zero-to-negligible direct AI demand — YouTrack shows 0/108 (0.0%) AI/ML mentions, and HubSpot surfaces exactly one customer-authored AI ask (an API-driven chatbot) against dozens of reporting and access tickets — so demand-driven prioritization argues for fundamentals first, with AI earning its seat on non-demand grounds rather than as the headline (per youtrack-demand; per hubspot-customer-voice).

The decision: defend enterprise vs attack mid-market — resolvedDecide

This is not an open fork the plan is asking leadership to break: it is already decided. BTDR-0004 (locked 2026-08-06) resolves the enterprise-defend-vs-LMM-attack question as "Hedge, sequenced" — support both objectives, but sequence and scope them honestly rather than pick one exclusively (per BTDR-0004). Concretely, the bet is: a layered AI posture (table-stakes reporting-AI, BYOM at the enterprise tier, differentiated integration/PMI-AI as the real win), a risk-tiered model-governance rule that keeps hosted-LLM exposure confined to the standalone Diversify product, and an incremental ~nine-month delivery horizon that lands access/reliability fundamentals and positioning first — with the Diversify product itself (and the reporting-AI floor) more likely to land later in that horizon given current two-developer capacity (per BTDR-0004). The evidence in this section (strong, corroborated enterprise demand vs. thin, unvalidated LMM demand) is exactly what makes "hedge, sequenced" — rather than an all-in bet on either side — the defensible read; downstream modules (M06/M07/M10/M11/M12) inherit this resolution rather than re-litigate it.

Fog: mid-market sizing is unknownFlag

The LMM thesis carries a headline $5M–$500M valuation band and a $5M–$50M "sweet spot," but neither figure is a sized, addressable market — both are NotebookLM-synthesized framing from ideation research, not a commissioned TAM/SAM/SOM study (per lmm-thesis). No fake-door validation, landing-page conversion data, or paying-customer signal exists yet to confirm real demand at any scale within that band. This is named here as unresolved fog, not closed: the plan does not yet know how large the addressable LMM opportunity actually is, and the fake-door validation plan referenced in the LMM thesis remains the open gate before any sizing claim can be trusted.

Fog: persona depth is uneven across segmentsFlag

The enterprise buying center rests on real depth — 15 interviewed stakeholders across named companies, cross-corroborated by two independent ticket streams and a recorded customer demo (per modus-enterprise-insights). The LMM side has no equivalent: just two NotebookLM-synthesized personas (the Searcher and the Portfolio Manager), built from ideation research rather than direct interviews or transaction data (per lmm-thesis). This asymmetry is named here as an open gap, not resolved: any LMM go-to-market or product decision made before deepening these personas — through real searcher/PE-operator interviews or the fake-door signal — rests on thinner evidence than the enterprise side of this section.

Current State — Honest Baseline

EMA today is a real, sticky enterprise platform — five capability pillars covering the full deal lifecycle, extreme configurability, and a customer base that treats it as operating infrastructure rather than a nice-to-have. It is also, by its customers' own account and by an independently commissioned assessment, running well below its potential: the loudest and most frequent complaints are about logging in and staying logged in, not about missing AI, and an external 2023 review scored the product-strategy discipline behind the platform at 1.5 out of 5.

Naming both of these plainly — what works and what is genuinely broken — is the credibility test this section has to pass before any future-state story is trustworthy; a plan that skips the gaps to get to the pitch faster loses the room. This section answers what those customers experience today, honestly, picking up directly from the segments and resistance patterns just described.

The Five Pillars + the API Surface

EMA's current capability surface is five pillars: Pipeline Management (internally "Targets" — the first-built module), Due Diligence, Post-Merger Integration ("Programs"), Synergy Tracking, and Legal Workflow, with a Target→Program handoff as the seam where a won deal converts into an integration effort. Sitting alongside the pillars is Deven Connect, an OpenAPI commercial layer — $0 for existing/early-access clients through their current contract term, $75k/year MSRP for new or renewing clients (⚠ theoretical / market-rejected / $0 realized — clients have not accepted this price; BTDR-0007 Am.1) — sold as a capability, not a managed service (per vault Home business-context / deven-connect / history notes — flagged Ingestion-Phase/provisional; Tim Eisenhardt is listed there as current CTO but departed ~July 2026).

The same codebase runs as multi-tenant SaaS or single-tenant per client via feature toggles rather than code branching — the mechanism that keeps a single CORE version live for every client and that later sections lean on for downmarket feasibility.

The #1 Pain Is Getting In and Staying In

The single loudest and most repeated customer complaint is not a missing feature — it is basic access and reliability: blank initial passwords, "login denied / inactive" errors, MFA resets, and sessions that silently time out and force a re-login (per hubspot-customer-voice — at least 34 of 640 tickets in the 24-month window, a lower-bound count). YouTrack's independent, census-clean 12-month sweep corroborates this at a similar scale: Access & Reliability is the #2 demand cluster at 16 of 108 issues (14.8%), spanning login/403/access-denied, SSO scoping, and session-timeout tuning, and it contains one of two dataset-wide Show-stopper-priority issues (LGCYEMA-423 "Accenture V&A 403 error"; the other, LGCYEMA-447 "Wipro Import Load Screen", sits in the DD / Data-Import cluster — both client-demand-side).

Two independent evidence streams — client support tickets and internal product-requirement demand — agree on this, and neither surfaces "the competition has AI and we don't" as a driver; the fix the platform needs first is that it works reliably, not that it gets smarter.

The #1 Capability Demand Is Reporting

Reporting is the single largest capability-demand cluster in both evidence streams: HubSpot shows at least 86 of 640 tickets (a lower-bound count) asking for reports that reconcile, export cleanly, and expose the right fields, and YouTrack's census-clean sweep independently ranks Reporting & Analytics #1 at 18 of 108 issues (16.7%), including a Power BI migration effort and two Critical-priority report bugs (LGCYEMA-411, -399). Customers describe timed-out report builders that lose unsaved work and dashboards that disagree with each other on basic completion percentages — this is not cosmetic friction, it blocks governance and executive trust in the numbers.

A third, independent source — the 2023 Modus enterprise interviews — names the same gap from the outside: no executive-ready views and granularity too coarse for the C-suite, forcing customers to export raw data into other tools. Three separately gathered signals landing on the same answer is the strongest evidence this section has.

An Externally Validated 1.5-out-of-5

In April 2023, Modus ran 15 structured interviews (13 external enterprise stakeholders plus 2 internal) and a heuristic UX evaluation, and scored Devensoft's product-strategy maturity at 1.5 out of 5 overall: Validation 1/5 (no formal validation — decisions driven by customer requests and anecdote), Prioritization 2/5, Alignment/Decision-Making 1.5/5, Communications 1.5/5, and Implementation 1/5, characterized bluntly as "sales-driven, not based on actual user data." The engagement itself was scoped, then not carried forward — the inferred reason was cost (~$1.1M for the recommended platform modernization) plus the bluntness of the "sales-led, not product-led" critique, not disagreement with the underlying finding.

This score is a 2023 baseline, not a current measurement, and should be read as directional rather than exact — but it is corroborated by the current internal signals (the same reporting and access/reliability gaps still top the 2025–2026 support and demand streams), so the direction has not obviously changed.

Customers Use a Third of What They Pay For

The same 2023 Modus interviews found that enterprise users utilize only an estimated 30–40% of the platform's capabilities, and that it takes 20-plus training sessions to reach proficiency — a cognitive-load and onboarding cost that makes new-seat sales unusually expensive to support. That onboarding cost now has a dollar shape: a one-time $5k–$20k implementation/"training" fee scaled by org size and complexity (per commercial-pricing-baseline §1 - Diego guidance, provisional) — which partially recovers the delivery cost of those 20-plus sessions but, given Modus's findings, may not fully cover it. Seat-license structure compounds this: teams routinely need to loop in people who don't hold a full license just to assign or track work, forcing workarounds outside the platform — a pattern corroborated directly in current HubSpot tickets asking to assign tasks to "non-full users" without giving them full access.

Together, low utilization and licensing friction mean the platform is paying a real cost in customer-acquisition and retention economics for capability that customers already own but can't easily reach.

What We Know About the Debt — and What We Don'tFlag

The technical-debt register names specific, quantified critical debt: TernaryLibrary, a custom MVC framework on .NET Framework 4.8 that is still foundational and largely untouched, sitting in a blast-radius chain through Tracing, Database, and configuration-editing components; Targets/Programs coupling, scored HIGH (24/69 blast radius) in the current WebMono codebase; and SharedLibs chokepoints — ConnectionManager (CRITICAL, 29/275) and IEmaConfiguration (CRITICAL, 21/282) — that sit underneath nearly everything else. The platform is also still running as a single monolithic embedded Angular SPA with one shared build and router, which blocks per-surface versioning or independent deployment. Against this, real modernization has already shipped and should be credited: all clients are on the latest CORE version, the single-tenant-via-feature-toggle architecture is done, and an IoC/DI migration (Unity → Microsoft.DI) has partially released to production.

What remains genuinely open and unquantified: the blast-radius scores come from a single code-scouting pass (April 2026) and are sensitive to re-indexing, not a stable or independently re-verified metric; there is no consolidated cost or timeline estimate for retiring this debt (unlike the priced Modus UX recommendations); the in-progress IoC/DI phase is already running at roughly 136% of its original estimate, a live signal that further estimates on this debt carry real uncertainty; and the engineering capacity available to work it down — two in-house engineers plus five offshore contractors, with the CTO seat vacant since ~July 2026 — is itself unresolved. This section names the debt honestly; it does not claim a remediation plan or cost exists yet (the register itself is status: draft — treat the whole picture above as directionally accurate, not verified ground truth).

The Future State — Before to After

The future state this plan proposes is the completion of a value proposition customers have already validated: transforming M&A execution from a reactive, spreadsheet-driven process into a structured, data-driven operating model (per demo_led-by-customer-dwight-summary). The strategy that delivers this is dual-track: Devensoft v24 — the modernized, API-first, AI-enabled enterprise platform — anchors defense and growth on the existing installed base, while the Diversify first step is held as a leadership vehicle choice (the LMM Vehicle Choice — DevenLite and DevenDiligence the Tier-1-live candidates; the plan commits to none) that opens the down-market ladder when chosen, the remaining candidates gated on demand signal (per BTDR-0010 Am.2). Inside the balanced hedge, the anchor track leads on capacity direction (fundamentals-first sequencing per BTDR-0004); Diversify is conditional, not parallel. Across both tracks, the positioning holds as one master frame: the platform is deal-strengthening critical infrastructure to the deal, not another tool — language customer-tested in the demo that reframes the product from optional software to load-bearing deal infrastructure (per demo summary Positioning Guidance; BTDR-0007 Decision 7 product ladder).

This framing earns its credibility the hard way: it is grounded in customer voice, not aspiration. The Before→After transformation is not a hypothesis — it is what a 35-year M&A veteran who led the platform's deployment at Devensoft's largest client narrated unprompted, closing with the assertion that teams who operate with full visibility do not return to manual methods (per demo summary Closing Message). The demand evidence is equally grounding: 86 HubSpot support tickets and 18 of 108 YouTrack issues concentrate on reporting, while zero of 108 ask for AI directly — telling this section what the future state must fix first (fundamentals, reporting, structured data) and what it should not over-promise (per hubspot-customer-voice; youtrack-demand-sweep). The competitive matrix independently confirms the opening: every named rival has converged on diligence-AI commodity, and none ships integration/PMI seam-AI — exactly where Devensoft's center of gravity sits (per competitor-feature-parity-matrix D2 read).

This is the future state that closes the gaps the prior section named — the access and reliability failures that stall value before it begins, the reporting granularity deficit that fails the economic buyer, the configuration-heavy enterprise friction that locks the product in a segment the market is moving past, and the 1.5-out-of-5 maturity baseline underneath all of it.

Before to After — The Value Proposition

The customer's transformation IS the story; the product architecture that delivers it is the mechanism subplot.

The Before

Today's M&A execution is, in the customer's own experience, a reactive, spreadsheet-driven process. Work happens across emails, shared files, and disconnected tools. An analyst spends days pulling spreadsheets together, reconciling versions, and rebuilding status reports every week — effort that produces a backward-looking snapshot, not a live operating picture. Data Requests built from a 10,000-line-item Accenture deal become email chains held hostage by the slowest responder. Reporting cycles run days, not seconds. The information is disconnected, and the time lost is spent coordinating across tools rather than executing the deal (per demo summary Core Value Proposition; hubspot-customer-voice §2 Before-pain clusters).

This Before is not a caricature — it is the lived pain the HubSpot ticket stream corroborates independently. Customers describe exports that require significant manual rework to be usable in Microsoft Project, activity-map completion percentages that disagree with operational reports, sessions that time out without warning and lose unsaved work, and login failures that stall every downstream workflow before value begins (per hubspot-customer-voice §2A–B, tickets 45566369291, 39019687214, 41596525650).

The After

The future state transforms this into a structured, data-driven operating model: a single source of truth where all deal activity lives — tasks, risks, dependencies, financials, and documents fully connected and updated in real time. Reporting cycles compress from days to seconds. Manual data aggregation and reconciliation are eliminated. Decisions accelerate because the information is live, not assembled. The analyst stops rebuilding reports and starts acting on them (per demo summary Time and Efficiency Impact; High Impact Quotes).

The demo's distilled framing captures the threshold moment crisply: "Once teams experience operating with full visibility and connected data, they do not return to manual methods" (per demo summary Closing Message). That is the After — not a feature list, but an operating-model shift the customer owns and defends.

Why the customer axis is the headline

The transformation above is the story: reactive spreadsheet-driven execution becomes a structured, data-driven operating model. The product architecture that makes it possible — the strangler synthesis (greenfield PLG frontend served by the modernized backend via the DevenConnect API seam, F1-gated, per BTDR-0006) and the product ladder (Devenfy → DevenLite → Devensoft v24, per BTDR-0007 Decision 7) — is the mechanism subplot: how the story is delivered, not the story itself. This separation is deliberate. It keeps the future state anchored in the outcome the customer buys — operational transformation — rather than the technology that enables it, and it keeps this section distinct from M05 (which owns the product Before) and M10 (which owns the build path).

Positioning — Critical Infrastructure, Not Another Tool

One master frame, expressed as a ladder

The positioning frame that holds across the entire future state is a single sentence, customer-tested in the demo: the platform is a deal strengthener — critical infrastructure to the deal, not "another tool" (per demo summary Positioning Guidance; Eric's framing in the demo Q&A). The distinction is load-bearing. "Another tool" lands the product in a category the buyer already has too many of — a line-item to be rationalized, a seat to be cut, a budget to be deferred. "Critical infrastructure to the deal" lands it in a category the buyer cannot afford to lose: the system the deal runs on, the audit trail the regulator asks for, the single source of truth the IMO leader defends. The demo's positioning guidance distilled this into a reusable rule: anchor in real pain points, focus on execution not features, and reinforce that this is critical infrastructure (per demo summary Positioning Guidance).

This one frame holds across every segment tier — but the delivery mechanism shifts as the buyer moves up the product ladder. At the bottom, a solo searcher or small LMM team encounters the frame through self-serve PLG (Devenfy candidate, DevenLite Targets+Tasking) — the same deal-strengthening value, delivered without the 20-session enterprise onboarding that the LMM rejects. At the top, an enterprise IMO leader encounters it through Devensoft v24 — enterprise-configured, deeply customizable, with the audit-log moat and the isolation hierarchy that an enterprise procurement process demands. The product ladder is what makes one positioning frame work across segments: the value proposition is constant; the packaging adapts (per BTDR-0007 Decision 7 — product ladder + graduation path; competitor-feature-parity-matrix D6/D7 — DealRoom's per-deal model attacks the enterprise seat-license friction the ladder resolves).

Why this frame is defensible

The frame is defensible because it names something the competitor matrix independently confirms is scarce. Devensoft owns the back half of the M&A lifecycle — Integration/PMI, Synergy/Value-Realization, and Legal/Regulatory — where its center of gravity sits (Data Requests, TSA lifecycle, Activity Map, synergy/KPI tracking) and where no competitor ships AI (per competitor-feature-parity-matrix D1 read, D2 read). Its regulator-ready audit log is a unique enterprise moat that no other evidence stream surfaces in a rival (per competitor-feature-parity-matrix D3 read). "Critical infrastructure to the deal" is not marketing language layered on top of a commodity product — it is the accurate description of what the platform already does at stages 7–9 that competitors do not match, and what the future state modernizes and extends rather than replaces.

Scope — The Winning Move and the Boundary

The winning move (destination)

The differentiation anchor of the future state is integration/PMI seam-AI — the one greenfield no competitor ships. The competitor feature-parity matrix establishes this directly: every named rival (Midaxo, DealRoom, Datasite, Intralinks) has converged on diligence-AI — document summaries, clause extraction, AI redaction, Q&A over data rooms — making that surface a commodity, not a differentiator. But no competitor ships AI at the integration, PMI, or synergy-execution stage — exactly where Devensoft's center of gravity sits (per competitor-feature-parity-matrix D2 read: "commodity-in-diligence, greenfield-in-integration"). The strategic opening is to own that seam — entering via the DevenConnect OpenAPI layer and a DevenMCP server as the integration on-ramp, rather than a bolt-on AI feature — so that the platform where deals are executed becomes the platform where AI assists that execution (per BTDR-0004 Decision 1 Win-tier; competitor-feature-parity-matrix D5 read: "Deven Connect + an MCP server = the natural AI on-ramp").

Greenfield qualifier: the Midaxo deep-profile surfaces a cross-project/synergy-AI capability (predictive delay/risk-alerting) that narrows the opening without closing it — it is adjacent alerting, not dedicated PMI-execution AI at Devensoft's depth. The greenfield claim is qualified, not retracted (per BTDR-0004 G1 amendment 2026-08-11).

This winning move sits inside a composite dual-track scope: defend and grow the enterprise base (fundamentals + reporting + modernization through Devensoft v24) while opening the down-market ladder with the Tier-1 LMM vehicle held as the LMM Vehicle Choice — a deferred leadership decision (DevenLite and DevenDiligence the Tier-1-live candidates; no vehicle committed), the remaining Diversify candidates behind demand-signal gates (per BTDR-0010 Am.2). AI is one element of the future state, located at the integration/PMI seam — not the headline. The headline is the customer's Before→After transformation. AI earns its seat at the seam because that is where Devensoft's competitive moat is deepest and where no rival has staked a claim; the depth of why-AI-there-and-not-elsewhere is M07's argument to make. This section locates it; M07 justifies it (per the M06 decision review).

The nine-month marker

The full Before→After vision — strangler synthesis landed, product ladder connected, seam-AI differentiated — is a strategic destination, not a nine-month deliverable. What is real by month nine is the down payment: fundamentals resolved (the access/reliability/auth work that the F1–F6 clusters name — the #1 pain in both demand streams), reporting modernized (the #1 capability demand at 18/108 YouTrack issues and ~86 HubSpot tickets), and demand-signal validation for the Diversify ladder (fake-door or instrument, pending the leadership decision on method). Seam-AI is roadmap within the destination, not a month-nine commitment (per BTDR-0004 Decision 1 Win-tier = roadmap phase; youtrack-demand-sweep: reporting + fundamentals = 51/108 = 47.2% of all demand).

Out of scope

What is deliberately out of scope for M06 — and referenced, not duplicated — is the build-path detail: the phased frontend-extraction roadmap, the F1 critical-path intersection, the capacity-vs-ambition sequencing, and the "Beyond the Horizon" items (DevenMCP full runtime core, the Devenfy full standalone, monolith decomposition). Those belong to M10 (Enablers + Roadmap). M06 owns the destination — what the future state is and why it wins. M10 owns the path — what unlocks what, in what sequence, against what capacity (per the M06 decision review; BTDR-0006 Consequences — F1 critical-path named for M10/M12).

Strategic Posture — Hedge, SequencedDecide

Three locked decisions constrain the posture; this section resolves how they combine into the organizing principle for the future state.

The dual-track architecture holds — balanced, not weighted

The future state is organized as a dual-track strategy: Devensoft v24 — the modernized, API-first, AI-enabled enterprise platform — is the demand-backed anchor track that defends the installed base and grows net-new enterprise; the Diversify products (the Tier-1 vehicle pending the LMM Vehicle Choice — DevenLite and DevenDiligence the live candidates, no vehicle committed; Devenfy a candidate) are the optionality leg that reaches down-market. No single product is the protagonist; the strategy — the story of how defense and diversification compose — is the protagonist (per the M06 decision review; encoding per BTDR-0010 Am.2). The anchor is served by the strangler synthesis: a greenfield PLG frontend delivered through a modernized-extracted backend via the DevenConnect API seam, with F1 (auth/session/scopes) as the gating prerequisite (per BTDR-0006). The optionality leg is connected through the product ladder: Devenfy → DevenLite → Devensoft v24, with DevenDiligence cross-cutting, turning one-and-done churn into lifecycle LTV (per BTDR-0007 Decision 7). Both fronts are supported inside a single time-boxed, sequenced hedge — a genuine two-front effort, not one leg carrying the other.

Posture: Hedge, sequenced — fundamentals-first

The resolved posture is Hedge, sequenced (per BTDR-0004 — the committed decision of record; both Eric and Thom chose Hedge, time-boxed with re-evaluation). Both tracks run; neither is abandoned. The sequencing rule that orders them is fundamentals-first: BTDR-0004 G4 establishes that capacity is smaller than ambition at a 2-dev team plus a contractor-vendor transition, so the plan takes the fundamentals direction first — access/reliability/auth, reporting modernization, API-first extraction — and the Diversify leg advances as a deferred Tier-1 choice with gated remainder: the Tier-1 slot is the LMM Vehicle Choice (BTDR-0010 Am.2) — a leadership decision with DevenLite and DevenDiligence the Tier-1-live candidates (no vehicle committed, none recommended; a DevenLite choice delivering late-window behind the floor), while the remaining candidates advance behind demand-signal validation (instrument method pending leadership decision). The DealSetupProcessing coupling is carried openly as the coupling-dominated sizing caveat on the DevenLite branch; the F1 serial gate still sequences DevenMCP. The 90/10 attention firewall guards focus against relapse into reactive client work (per BTDR-0004 G4 — capacity < ambition, fundamentals-first, the 90/10 firewall).

The FY26 financial baseline is why the fundamentals sequence first — not a rule that Retain wins every capacity contest. Closed-renewal retention ran 82.7%; closed-won landed at $700.8K against a $1.92M projection; and the active base is roughly 22 companies, concentrated — Stryker's $200K alone is ~2.3× the median realized ACV (per fy26-financial-baseline, provisional — internal projection, not recognized ARR). A base this concentrated means a single lost renewal materially changes the cash position — so BTDR-0004 sequences the fundamentals first within the hedge; the hedge is not abandoned, but its Retain leg carries the sharper near-term risk.

Focus vs. capacity-direction: the two questions

This resolves cleanly into two axes, and a leadership audience will press both. The 90/10 attention firewall operates on the focus axis: keeping the team's energy on transformation rather than relapsing into reactive client work (the 2020 failure mode). The fundamentals-first sequencing operates on the capacity-direction axis: when capacity is smaller than ambition, it sets which direction goes first. Together they answer the two questions a leadership audience will press first: are you protecting the rebuild from client-urgency drift? (yes — the firewall) and given finite capacity, where does the direction start? (fundamentals first). Both stand under severity-triggered, monthly drift-checked discipline across the ~9-month horizon (per BTDR-0004 G4 horizon supersession — the confirmed planning horizon is ~9 months, incremental and iterative).

What this means for the future state

The future state, organized under the sequenced hedge, is an enterprise platform whose fundamentals are fixed (access, reliability, auth — the #1 pain in both demand streams), whose reporting is modernized (the #1 capability demand), whose backend is API-first (DevenConnect as the seam both tracks share), and whose AI earns its seat at the integration/PMI seam — the greenfield no competitor owns. The Diversify ladder extends the same value proposition down-market when signal justifies the build, not before. This is the destination the rest of the plan builds toward.

Exit Optionality — Flagged, Decision DeferredFlag

Flag — named, not resolved. The exit bet is a leadership decision that belongs to M14 (Recommendations & Biggest Bets) and M01 (The Ask), gated on uncaptured owner stance.

A factual consequence of the future state — not a recommendation, but worth naming — is that it is more sellable than the current state. Today's Devensoft v23 is a configuration-heavy monolith with a seat-license + 20-session-onboarding commercial model, near-invisible digital presence (G2 4.6/5 across only 12 reviews; not listed on Capterra; empty TrustRadius), and a realized DevenConnect revenue of $0 (per competitor-feature-parity-matrix D7 read; BTDR-0007 Amendment 1). The future state — modernized through F1–F6, API-first via DevenConnect, multi-product across the ladder, with the PMI/seam-AI differentiation anchor no competitor ships — is a materially different asset to a potential acquirer or partner. The strangler synthesis reuses rather than discards the backend (per BTDR-0006), so the modernization is accretive to enterprise value, not a sunk-cost rebuild.

This section flags that optionality exists as a consequence of the plan, not as a driver of it. Whether to treat exit as a first-class strategic objective — and on what timeline, and at what valuation threshold — is a decision that depends on Owner intent that this plan has not captured (Thom's stance on endurance vs exit is uncaptured; per BTDR-0004 Context: Owner intent = endurance, not exit, at the current evidence level). The exit bet — the recommendation to pursue or foreclose a sale — goes to M14 and M01, where the plan's closing decision and the call to leadership live. M06's contribution is the one-liner: the future state preserves exit optionality that the current state does not.

AI Thesis

AI earns its seat in exactly two places in this plan: as a defensive "yes, we have AI" posture at the Deven Connect API layer, and as a genuine differentiator at the integration/PMI seam — Devensoft's center of gravity — where no rival platform has arrived yet. It does not earn its seat as the plan's headline, and it does not earn its seat inside due diligence, where four competitor platforms (Midaxo, DealRoom, Datasite, Intralinks) already ship an identical commodity AI feature set. What follows resolves the future-state capability set carried forward from §06 (M06 Strategy/Future State) — where AI earns its seat in that future state — by anchoring it to evidence rather than to the industry-wide AI reflex every M&A software vendor is currently chasing.

This is a demand-evidenced ranking, not an AI-averse or an AI-first position. Zero of the 108 client-raised issues in the trailing 12-month YouTrack census requested any AI capability — the absence is itself the finding, and it is corroborated independently by the HubSpot support stream (per youtrack-demand). At the same time, every principal competitor has converged on the same diligence-AI menu, which makes AI-in-diligence a cost of entry rather than a source of advantage (per competitor-feature-parity-matrix). The two Decide: calls this section carries — AI's priority rank against fundamentals, and AI build-vs-buy — are not open: both are already locked by BTDR-0004, and this section names their specific terms rather than re-litigating them.

The Direct-Demand Floor Is Zero

In the census-clean 12-month sweep of the LGCYEMA YouTrack project (108 issues, 2025-08-07 to 2026-07-30, no sampling), not one summary mentions AI, ML, copilot, agents, or prediction — 0 of 108, 0.0% (per youtrack-demand). This is not a gap in a lower-bound estimate; the YouTrack count is a full census, and the independent HubSpot support-ticket stream corroborates the same absence. What clients are asking for instead is concrete and concentrated: Reporting & Analytics is the single largest cluster at 18 issues (16.7%), Access & Reliability is second at 16 (14.8%, including the only demand-side Show-stopper), and reporting plus fundamentals together account for 51 of 108 issues (47.2%) — essentially half of all demand. Any AI-first pivot would be unmoored from every captured signal of what Devensoft's own customers are actually asking for.

Every Competitor Has Already Converged On the Same AI Menu

Where clients aren't asking, the market has already moved: Midaxo, DealRoom, Datasite, and Intralinks all ship substantially the same diligence-AI capability set — document summarization/Q&A, clause and risk extraction, and (for DealRoom and Datasite) live MCP connectors into Claude, ChatGPT, and Copilot (per competitor-feature-parity-matrix). Devensoft ships none of it today, which is a real gap — but closing that gap would only buy parity, not advantage, because the feature set is now commodity across the category (per enterprise-market-opportunity §2 — "identical AI feature set... = table-stakes across Midaxo/DealRoom/Datasite/Intralinks"). The lifecycle stage furthest from that diligence-AI commodity race is PMI/integration/synergy — Devensoft's existing center of gravity — where no rival has shipped dedicated integration/synergy-execution AI at Devensoft's depth. The nearest adjacent capability is Midaxo's cross-project delay/risk alerting layered on its synergy tracker, which Midaxo itself lists as a strength (per competitor-profile--midaxo — "native mature AI (broadest incl. cross-project/synergy+predictive)"). That near-miss narrows, but does not close, the greenfield opening this thesis is built around — it is not the crowded diligence lane every vendor is racing to match.

The Credible On-Ramp Is the API, Not a Feature

Client demand does point at AI's prerequisite, even where it doesn't point at AI itself: the API/Integration cluster is the sixth-largest demand theme in the YouTrack census (10 issues, 9.3% — Core Web API build-out, secure credential storage, RAID-via-API), which the analysis reads as "AI's on-ramp, not AI itself" (per youtrack-demand). This is exactly the surface the existing Deven Connect commercial API layer already occupies, and it is the same surface the predecessor AI-integration work already designed against — routing client-owned model calls through the platform rather than hosting AI as a standalone feature (per ai-integration-deck). Opening and hardening that API surface is therefore the credible, demand-grounded path into any AI capability, not a new AI feature bolted onto the existing surfaces.

Decided: AI Ranks Behind Fundamentals, Layered Rather Than BinaryDecide

Where AI ranks against fixing fundamentals is not an open question here — it is locked by BTDR-0004 (G1, 2026-08-06) as a layered AI posture rather than a single AI-or-fundamentals choice. Three tiers: a Floor tier — table-stakes "yes we have AI," delivered as reporting-AI (natural-language-to-insight/report), targeted for the early phase of the ~9-month horizon but explicitly conditional on a time-boxed feasibility PoC; a Satisfy tier — answering the "integrations + AI" ask prospects raise, delivered as Deven Connect + BYOM (bring-your-own-model), also early-phase; and a Win tier — real differentiation via integration/PMI/seam-AI (DevenMCP), deliberately scoped to the roadmap horizon, not the initial delivery window. Ranked against fundamentals, fundamentals sequence first: the Floor and Satisfy tiers exist to answer the market's AI expectation without displacing the reporting-and-access-and-reliability work the 51/108 (47.2%) demand cluster actually calls for (per youtrack-demand). BTDR-0004's capacity check (G4) sharpens this further — with two capable engineers plus a contractor transition, the realistic early-horizon ship is access/reliability fundamentals plus positioning; BYOM-build and reporting-AI are likely to land later in the horizon, meaning the near-term "we have AI" story rests on positioning and Fake-Door signal, not shipped features. That gap between the Floor/Satisfy ambition and current capacity is itself surfaced as an execution risk in M10/M12, not hidden here.

Decided: A Risk-Tiered Model-Governance Rule, Not a Uniform Build-vs-Buy AnswerDecide

Build-vs-buy is resolved as a segment-conditioned rule, not a single answer, per BTDR-0004 (G1, 2026-08-06): at the enterprise tier, a hosted/general LLM offering is disallowed — BYOM is mandatory, and the customer's own model serves all platform AI uniformly, keeping Devensoft out of the AI-provider/sub-processor compliance path. This reverses the original full-governance design, which would have made Devensoft the AI sub-processor and required provider DPAs, data-residency, and no-train obligations Devensoft would otherwise have to own directly (per ai-security-compliance). A hosted general-model offering exists only inside the standalone LMM micro-product, where legal/MNPI stakes are lower and Devensoft-as-sub-processor is an acceptable trade — meaning "build" (a hosted model layer) is scoped narrowly to that one segment, while "buy/route" (BYOM atop the existing Deven Connect commercial API) is the rule everywhere else. BTDR-0004 names the resulting trade-off directly: enterprise BYOM is a different value proposition than rivals' fully-managed AI — Midaxo, Datasite, DealRoom, and Intralinks all ship native, hosted diligence-AI (per competitor-feature-parity-matrix) — and that positioning cost is accepted deliberately, in exchange for staying out of the AI compliance path at the tier where the legal stakes are highest.

Flagged: AI ROI Is UnknownFlag

The only worked-out AI cost model in the corpus is illustrative, not validated: a $709K/year total-cost-of-ownership estimate (indirect costs 42%, token spend 38%, throughput/egress 10%, model gateway/ops 7%, vector store 3%), built on assumed adoption figures — roughly 40 tenants, 20 active AI users per tenant, 50 calls per user per day (per ai-integration-deck). No revenue, retention, or win-rate return has been modeled against that cost, and the deck itself flags the model as needing validation before any pricing locks. Whether the Floor/Satisfy AI tiers pay for themselves — in retained enterprise accounts, in won deals, or in reduced churn — is genuinely unresolved. This section names that gap; it does not resolve it, and no cost-benefit claim in this plan should be read as an ROI figure until that validation work is done.

Flagged: The Vendor Landscape Is Still Only Partly MappedFlag

The competitor picture behind this thesis is solid on the questions that matter most — what AI features ship, and where — but several adjacent gaps remain open: Midaxo's and DealRoom's API/MCP depth is not publicly documented and is likely internal-only; Midaxo's and Datasite's pricing is contact-sales-only, so DealRoom's per-deal model is the only fully public pricing reference; and whether competitor MCP connectors expose read-only or read/write access to deal data is unverified (per competitor-feature-parity-matrix). Most of these gaps don't change the core read — commodity diligence-AI, greenfield PMI-AI — but one is more load-bearing: the per-vendor Midaxo profile (dated 2026-08-07, one day after BTDR-0004's G1 lock) surfaces a cross-project delay/risk-alerting capability layered on Midaxo's synergy tracker, which Midaxo lists among its own AI strengths (per competitor-profile--midaxo). That sits closer to the PMI/synergy seam than the matrix's original "no rival ships AI here" summary allows for, and the evidence postdates BTDR-0004's G1 lock — it narrows, without closing, the greenfield claim the Win tier rests on. This has been reconciled against BTDR-0004 (G1 greenfield-qualifier amendment, 2026-08-11): the Win-tier greenfield is qualified to "at Devensoft's execution depth; Midaxo cross-project/synergy-AI is a near-miss," not retracted. This section names the remaining landscape gaps as open; closing them is follow-up research, not a blocker to the AI thesis itself.

Core Capabilities

The future state this plan argues for (§07) needs a specific capability set to deliver it, and that set sorts into four tiers — not three. The first tier, Core IP, splits into lines that are settled by evidence or shipment today and lines that are candidates needing leadership's investment decision. The settled lines are: end-to-end lifecycle integration (the platform's five pillars — Targets, Due Diligence, PMI, Synergy, Legal — running in one shared data model across all nine M&A lifecycle stages, with no rival matching the back-half depth), Reporting & Analytics (the #1 capability demand across every independent evidence stream), and the audit log (a regulator-ready, shipped, uncontested enterprise moat). The candidate lines — Integration/PMI-seam AI, agent-led onboarding (DevenMCP), and Security & Compliance maturity — earn their place by competitive differentiation or vision, not demand convergence, and they need a fund/throttle/defer decision from Eric and Thom.

The second tier, Foundations, is the broken floor: access and reliability basics (the #1 pain across every evidence stream) and the UX/configuration-debt remediation that drives the platform's 20-plus-session onboarding burden. The third tier, Enablers, is the strategic investment layer — modularization, frontend/backend decoupling, and the Deven Connect API surface — that unlocks the future product family without being customer-facing IP in its own right. The fourth tier, Outsource / Non-build, distinguishes three structurally different postures: Buy (SSRS engine replacement, auth-provider migration), Don't build (commodity enterprise diligence-AI), and Customer-supplies (BYOM at the enterprise tier).

This inventory earns its credibility the same way the rest of this plan does: every line traces to a named source, not a feature wishlist (guiding principle 1 — evidence-led, not feature-led). The settled Core IP lines trace to demand convergence (per youtrack-demand; per hubspot-customer-voice), competitive profiling that names them uncontested (per competitor-feature-parity-matrix), and direct customer-demo evidence of differentiated workflow IP built from real deal pain (per demo-dwight-customer-voice). The Foundations tier traces to YouTrack's pain census (access/reliability = 16 of 108 issues, 14.8%) and to Modus's independent 2023 research on the configuration-tax and training burden (per modus-enterprise-insights). The Enablers tier traces to Modus's explicit modularization mandate and the platform's own technical-debt register (per technical-debt-register — draft; per platform-customizability-breadth). And the Outsource tier follows directly from the AI posture this plan has already locked: BTDR-0004 puts commodity diligence-AI in the deliberately-not-build column and mandates customer-supplied models (BYOM) at the enterprise tier (per BTDR-0004-pivotal-bet), while BTDR-0008 outsources authentication to a managed identity provider (per BTDR-0008-auth-provider-migration).

This section answers the capability question the future state (§07) raised: what does the winning platform need to be able to do, concretely, and which of those things are worth owning versus fixing versus enabling versus buying versus not building at all.

End-to-End Lifecycle Integration: The Platform's Core IP

The platform's five pillars — Targets (Pipeline Management), Due Diligence, Programs (Post-Merger Integration), Synergy Tracking, and Legal Workflow — are not five separate capabilities. They are one capability: end-to-end lifecycle integration, running in a single shared data model across all nine M&A lifecycle stages, from sourcing through post-close value realization. A Target flows through the pipeline; winning the deal converts Target to Program, triggering the PMI phase — that conversion is the lifecycle seam no competitor models the same way (per demo-dwight-customer-voice). Naming the pillars individually invites "we have pipeline management too" commodity comparisons; naming them as one integrated lifecycle line makes the differentiation argument: Devensoft caters to all nine stages in one platform, and no rival in the competitive set matches the back-half depth (per competitor-feature-parity-matrix).

The honest lifecycle-fit assessment. Catering to all nine stages is not the same as being a complete solution for them. Against the canonical stage-need bar (the five capability bullets per lifecycle stage), the agreed fit-to-need matrix scores Devensoft v23 with zero natively-covered stages: three stages land in the configurability band (Pipeline Management 72, Due Diligence 75, Integration Execution 75), five in the partial band, and one — Regulatory & Compliance, at 5 — effectively absent. The platform's strongest stages, Integration Planning and Integration Execution, are real but configuration-shaped rather than native ("configuration driven… only for some aspects of this stage needs"; "configuration can achieve a lot" — Diego's 2026-08-18 re-band), and the gradient runs back-heavy (Integration Planning 63, Integration Execution 75, Post-Close Value Realization 62) over front-thin (Opportunity Identification 35, Deal Execution 35, Close & Cutover 40). The honest summary is the matrix's own: v23 fills every stage through configurability and is a complete solution for few — which is precisely what makes the back-half moat claim credible rather than promotional, and what the Diversify family (front-scoped DevenLite first, follow-ons behind it) and the v24 deepening are shaped against (per fit-to-need-canonical — AGREED first pass 2026-08-18; rubric 81–100 native · 66–80 configurability · 31–65 partial · 11–30 marginal · 0–10 absent; the stage-need bar is aspirational, not demand evidence).

Three proof points sit underneath this line. First, Data Requests and TSA (Transitional Service Agreement) workflow modules were built from a real 10,000-line-item deal pain point — they are demand-originated IP, not architecture, and they reach back into Due Diligence from the Integration phase (per demo-dwight-customer-voice). Second, the audit log — which tracks every action with timestamp, user, and detail, sufficient for regulatory audit — is a shipped, uncontested enterprise moat that "ends regulator conversations" and is DORA-aligned today (per competitor-feature-parity-matrix; per demo-dwight-customer-voice). Third, the PMI/Synergy/Legal depth at stages 7–9 is where Devensoft's center of gravity sits — the Activity Map, Financials, Synergies, RAID logs, Key Deliverables, and permissioning down to task level are all shipping capabilities at a depth no competitor's product reaches (per competitor-feature-parity-matrix; per demo-dwight-customer-voice).

This line is settled as Core IP — not by demand convergence (though Due Diligence is the clearest lifecycle friction cluster per youtrack-demand), but by shipment and competitive uncontested status today. It does not need Eric and Thom's ranking; it is the platform's identity.

Reporting & Analytics: Core IP Capability, Outsourceable Engine

Three independent evidence streams converge on the same answer, which is rare enough in this plan to be load-bearing on its own. The 12-month census of the YouTrack LGCYEMA requirements project puts Reporting & Analytics at 18 of 108 issues (16.7%), the single largest cluster in a full, unsampled count (per youtrack-demand). The HubSpot support-ticket stream, drawn from a separate 640-ticket 24-month extraction, independently names reporting as its largest capability-demand cluster at roughly 86 tickets (per hubspot-customer-voice). And Modus's 2023 enterprise interviews name "disappointing, inefficient reporting" as the #2 named pain across 15 stakeholders, driven by a real "granularity gap" (per modus-enterprise-insights). That convergence earns Reporting its Core IP status — it is the one capability every measured signal agrees is the platform's biggest opportunity to close a real gap.

It is also the capability this plan's locked AI posture treats as the credible floor-tier AI bet: BTDR-0004 places reporting-AI (natural-language-to-insight) at the Floor tier specifically because it is demand-attached — though that placement is explicitly conditional on a time-boxed proof-of-concept, not a committed ship (per BTDR-0004-pivotal-bet).

But the capability and the engine are not the same thing, and this plan splits them. The underlying reporting architecture today is SSRS (SQL Server Reporting Services) — a server-rendered engine that requires a Devensoft code-change to modify any report, with PowerPoint export producing image-only (not editable) output, and embedded Power BI dashboards layered on top (per vault-comprehensive-suite — provisional; per demo-dwight-customer-voice). SSRS is the mechanism behind the granularity gap Modus named. So while the reporting capability is Core IP, the SSRS engine belongs in the Outsource / Buy tier: replace it with a modern BI platform (Power BI native, or equivalent) that lets customers self-serve report modification without a code-change. This split prevents the Core IP label from accidentally defending a broken engine, and it gives leadership a clear "buy" decision alongside the "own" decision.

Foundations: The Broken Floor

Before any strategic investment earns trust, the platform has to stop bleeding from two wounds that every evidence stream names independently.

The first is access and reliability basics — login denials, session timeouts, blank-password errors, and the general class of "I can't get into the tool" failures. This is the #1 pain in the YouTrack LGCYEMA census at 16 of 108 issues (14.8%), ahead of every other theme including reporting (per youtrack-demand). It is not a strategic capability gap; it is a floor-level failure that erodes trust in everything else the platform does. These are table-stakes fixes, not investments — they must be closed before any new capability work is credible.

The second is UX and configuration-debt remediation — the cumulative cost of the platform's configuration-driven architecture operating without a design system or component library. The platform's nine configuration levers let one codebase serve boutique-to-enterprise customers, but the cost is that no two clients see identical UIs, navigation has accreted into bloat across 20 years of feature additions, and the configuration tax makes every surface harder to use than its individual features warrant (per platform-customizability-breadth). Modus's 2023 research quantified the commercial impact: 20-plus training sessions are required to reach user proficiency (a CAC problem that makes the sales-led model unscalable), users use only 30–40% of available capabilities, and industry churn data puts a bad first experience at roughly 32% churn risk (per modus-enterprise-insights). This is also table-stakes work — "stop the bleeding" — not strategic investment.

Both items sit in Foundations, not Enablers, because they are remediation of existing problems, not creation of new capability. The component library and design system that would structurally fix the configuration tax belong in the Enablers tier (§05); Foundations is the tactical remediation work that must run in parallel — the polish, the nav cleanup, the first-experience fixes — until the structural enablers arrive.

Enablers: The Strategic Investment Layer

Three capabilities sit in the Enabler tier: none of them sells the platform on its own, but nothing sellable below the enterprise price point exists without them. They are capability arguments (what the platform needs to be able to do); their sequencing and execution belong in §10 (Enablers & Roadmap), and the product surfaces they unlock belong in §11 (Commercial Model & GTM).

Modularization is the first enabler. Modus's 2023 UX/UI modernization engagement produced an explicit structural recommendation: "Productize or create modules. Apply boundaries to the subsets of activities." The mandate names a specific mechanism: replace the platform's bespoke, per-instance "Workstreams" model with standardized "Streams of Value," backed by a reusable component library and a real design system, neither of which exists today (per modus-enterprise-insights). The commercial logic is concrete: modularization is what makes a lower-cost, self-serve product tier commercially legible — Modus's own priced "PE Proof of Concept" recommendation (~$117k, ~6 weeks) is a near-direct template.

Frontend/backend decoupling is the second enabler. The platform's own technical-debt register names the specific coupling that blocks independent deployment: one Angular application (emaspa) with a single router and build serves many mount surfaces, bridged into the legacy MVC shell through a shared handler registry and hard-coded URLs (per technical-debt-register — draft). Modus recommended micro-frontends backed by CI/CD and a shared component library. There is a real architectural fork here — Modus evaluated and rejected an incremental "Strangler Fig" extraction in favor of a cleaner rebuild, while the platform's own modernization plan leans toward the incremental path Modus rejected — but that sequencing choice belongs in §10, not in this capability inventory.

The Deven Connect API surface is the third enabler — the seam everything else rides on. It is a Level-3 REST (Collection+JSON HATEOAS) surface, and a working proof-of-consumer already exists: an Aurelia2 single-page application deployed to Azure Static Web Apps consumes the Deven Connect hypermedia API across thirteen entities today (per EMAKBAI-devenvibe-frontend-container; per vault-comprehensive-suite — Ingestion-Phase/provisional). That same artifact has surfaced concrete backend gaps a real extraction would inherit: missing total-count metadata on collection responses, an inconsistent single-item lookup that returned a 404, and casing drift between API and client models. These are a scoped punch-list, not an open architecture question. The API's commercial packaging — tiered pricing, Zapier distribution — belongs in §11; its technical completeness belongs here.

Open Decision: Core-IP Candidate Ranking

The settled Core IP lines — end-to-end lifecycle integration, Reporting & Analytics, and the audit log — do not need ranking; they are the platform's identity. What needs leadership's decision is three candidate lines that earn their place by competitive differentiation or vision, not by demand convergence or shipment. These candidates compete for different resources and serve different strategic objectives, so the ask is not a flat linear ranking but a fund / throttle / defer sort: for each candidate, Eric and Thom decide independently whether to fund it, throttle it (limited investment), or defer it.

Integration/PMI-seam AI is the strongest candidate by competitive differentiation. BTDR-0004 names this the plan's one genuine "Win"-tier AI bet precisely because no rival in the competitive set currently ships it, unlike diligence-AI, which every serious competitor already offers as a commodity feature (per BTDR-0004-pivotal-bet; per competitor-feature-parity-matrix). It is unshipped today and requires AI-engineering capacity that does not yet exist (the staff-augmentation vendor path).

Agent-led onboarding (DevenMCP) is a candidate by vision. A working consumer of the platform's hypermedia API already exists — the Aurelia2 SPA that navigates Deven Connect's HATEOAS surface across thirteen entities — which makes this more than speculative (per EMAKBAI-devenvibe-frontend-container; per vault-comprehensive-suite — Ingestion-Phase/provisional). But it remains a vault-documented thesis, not a validated one. It should be funded (if at all) as a bet, not a committed line. Its concrete prerequisite — validation rules not being introspectable through the modern API — is named in §08 (Unresolved Fog).

Security & Compliance maturity is a candidate by aspirational differentiation. The full AI-governance ADR suite (ADR-0031..0040) and the three-tier isolation hierarchy (ADR-0040) could constitute a procurement moat that most competitors cannot produce. But this line is conditional: the underlying 53-question client AI-governance report that the entire ADR suite cites as its originating evidence has not been directly opened or verified in this plan's evidence trail (per vault-comprehensive-suite — Ingestion-Phase/provisional). Until that source is confirmed, the procurement-moat claim is directionally likely, not verified. Furthermore, most of this line — SOC2/ISO certifications (currently "by-inheritance/claimed"), the ADR enforcement layer (target architecture, not shipped) — is aspirational, not shipped. Only the audit log (already settled Core IP) is proven today.

This is deliberately not pre-decided here. The ask to Eric and Thom is: for each of these three candidates — fund, throttle, or defer — using five criteria: (1) capital required, (2) engineering capacity (post-Tim, with staff-aug vendor for AI work), (3) time-to-market, (4) differentiation durability (how long the moat lasts), and (5) downstream unlock (what the capability makes possible that is not possible without it). Each accepted candidate gets an explicit funding line, captured as a new BTDR once decided.

Outsource / Non-Build Tier

The mirror-image decision is which capability lines this plan should deliberately avoid owning. The tier distinguishes three structurally different postures, because each carries a different cost model, risk profile, and commercial implication.

Buy — capabilities where we pay a vendor to provide something we choose not to own. Two lines sit here. First, the SSRS reporting engine is a Buy candidate: the reporting capability is Core IP, but the SSRS engine (which requires a code-change to modify any report and produces image-only PowerPoint exports) should be replaced with a modern BI platform. Second, authentication is a locked Buy decision: BTDR-0008 outsources auth to a managed identity provider (Clerk or Auth0), replacing the platform's dual legacy auth (Forms Auth + OAuth client-credentials-only) (per BTDR-0008-auth-provider-migration). This is F1 on the §10 critical path — a hard prerequisite for all Diversify products.

Don't build — capabilities we deliberately choose not to have, at any cost. BTDR-0004 places commodity enterprise diligence-AI in this column outright: every serious competitor already ships it, so building a "me-too" version would spend capacity to match a feature no evidence stream asks us for (per BTDR-0004-pivotal-bet). Zero cost, zero capability, zero competitive gain.

Customer-supplies — capabilities where the customer owns the capability and we route but do not host. BTDR-0004 mandates bring-your-own-model (BYOM) at the enterprise tier: the customer's own model serves platform AI uniformly, which keeps Devensoft out of the compliance and sub-processor obligations that come with hosting a model directly (per BTDR-0004-pivotal-bet). Zero inference cost to Devensoft; the customer-onboarding burden shifts to the customer.

Most of this tier is already locked. The BYOM/no-hosted-model boundary, the no-commodity-diligence-AI boundary, and the auth-provider migration are all decided by BTDR-0004 and BTDR-0008. The one open ask for Eric and Thom is the SSRS engine replacement: which modern BI platform, and on what timeline. That decision, once made, gets captured as a new BTDR.

Unresolved Fog: Boundaries Named, Not Settled

Several boundaries in this section are named, not settled, and should stay named rather than be quietly assumed away.

The validation-not-introspectable API gap is a concrete, unresolved dependency. The platform's validation rules are enforced on write but are not introspectable through the modern Deven Connect API (per platform-customizability-breadth). This is a direct blocker for any agent-assisted or dynamically-generated frontend — including the DevenMCP candidate (§06) — and it is not yet decided whether that gap gets fixed, worked around, or fed out-of-band. Until it closes, the agent-led-onboarding candidate carries a known hard dependency.

The 53-question AI-governance report is the load-bearing source the entire AI-ADR suite (ADR-0031..0040) cites as its originating evidence, and it has not been directly opened or verified in this plan's evidence trail (per vault-comprehensive-suite — Ingestion-Phase/provisional). Because Security & Compliance maturity is now a named Core-IP candidate (§06), this conditional travels with the promotion: until the report is opened and confirmed, the procurement-moat claim is directionally likely, not verified. If the report confirms the moat, the condition lifts; if it does not, the candidate downgrades.

The Strangler-Fig vs incremental-extraction fork — Modus's rejection of incremental extraction in favor of a cleaner rebuild versus the platform's own modernization plan favoring the incremental path — is a real architectural disagreement. Both paths reach the same decoupled end state, but which is safer to execute with today's team is unanswered. This is a sequencing decision and belongs in §10, not here; M08 treats FE/BE decoupling as an Enabler regardless of which path is chosen.

Diversify product breadth — whether additional products beyond DevenLite (Targets+Tasking) and DevenDiligence (DD carve-out) are needed, and in what sequence — is an appetite-driven call for Eric and Thom, informed by fake-door market validation. DevenLite and DevenDiligence are resolved as separate products (the M11 decision review, 2026-08-11); the broader question of additional products (Devenfy, SDE Calculator) remains open and belongs in §11 (per BTDR-0004-pivotal-bet).

Naming these boundaries here is the point. Three of the four are deferred to other sections (§10, §11) or to future evidence verification; only the API gap is a within-M08 dependency.

M&A Journey - Intro Summary

Devensoft owns the full enterprise M&A lifecycle — stages 1 through 9 — but that coverage is not uniform, and the friction is not evenly distributed. The platform's center of gravity sits squarely in the back half: Integration Planning, Integration Execution, and Post-Close Value Realization (stages 7-9), where the PMI/Synergy/Legal depth is the competitive moat no rival matches. Due Diligence (stage 3) is the clearest lifecycle-specific friction cluster in the demand signal. Sourcing and Pipeline Management (stages 1-2) are real, used standalone by clients, but are not CRM-native — a positioning question that belongs to the market and future-state sections. The middle stages (4-6: Deal Execution, Regulatory & Compliance, Day One Readiness) are thin in the demand record, and that silence is itself a finding. A parallel, genuinely different lifecycle exists for the Lower-Middle-Market — five stages, not nine, with a different terminal point — but its friction is demand-unvalidated and remains a skeleton here. Reporting is not a stage-specific capability; it is the plan-wide cross-cutting underlay (#1 demand across YouTrack and HubSpot) that spans every stage and is detailed in the current-state and capabilities sections.

The credibility of this friction map comes from what it does not do. It does not list features and call them coverage. Every friction point named here traces to a demand signal — a YouTrack issue, a HubSpot support ticket, a customer-voice demo moment — and where the demand signal is absent, the section says so rather than filling the gap with an asserted need. The two independent demand streams (108 YouTrack PRD issues, census-clean; 640+ HubSpot support tickets, lower bound) agree on where the lifecycle hurts most, and both corroborate the demo transcript's center-of-gravity read: the Integration Execution stages where Dwight Fontilla — a 35-year M&A integration veteran who led the platform's adoption at the largest client — spent the majority of a 57-minute demo (per demo-dwight-customer-voice). The competitive gradient — strong where rivals are absent (PMI/Synergy/Legal), present where they are strong (sourcing), thin where the demand is thin (stages 4-6) — is cited from the parity matrix, not invented (per competitor-feature-parity-matrix).

The capabilities the prior section categorized — Core IP, Supporting, Outsource — are what serve this journey. The question this section answers is not what capabilities exist (that is M08) or how they are built and sequenced (that is M10), but whether the platform owns the lifecycle those capabilities serve, and where the per-stage friction actually lives. The answer — yes, stages 1-9, with honest depth variation and a parallel LMM lifecycle that needs validation — is the bridge from capability inventory to the execution roadmap that follows.

Two Parallel Lifecycle Views

The M&A journey this platform serves is not one lifecycle but two, and conflating them was the original analytical bug. The enterprise 9-stage lifecycle is the primary view — it is the funding base, the installed base, and the lifecycle every current client pays for. The Lower-Middle-Market 5-stage lifecycle is a genuinely different stage set, not a depth subset of the enterprise nine. Conflating the two — treating LMM as "enterprise-lite" — was the error the early framing carried, and it produces wrong product decisions: features that serve enterprise deal complexity (Regulatory Compliance, Deal Execution governance) do not exist in the LMM lifecycle at all, and features that matter in LMM (Valuation/SDE-to-GAAP conversion) have no enterprise equivalent.

The enterprise 9-stage lifecycle (the canonical list) runs:

1. Opportunity Identification
2. Pipeline Management
3. Due Diligence
4. Deal Execution
5. Regulatory & Compliance
6. Day One Readiness
7. Integration Planning
8. Integration Execution
9. Post-Close Value Realization

The platform's center of gravity — demonstrated by a customer-turned-advocate who led integration at the largest client — sits in stages 7-9 (Integration Planning through Post-Close Value Realization), with reach back into Due Diligence (stage 3) via the Data Requests workflow and into Regulatory (stage 5) via the audit log (per demo-dwight-customer-voice § Capability → M&A-lifecycle). The terminal point is deal-centric: did this deal hit the synergies and value-creation targets the integration plan committed to?

The LMM 5-stage lifecycle is a structurally different path (per lmm-thesis §4):

1. Search / Sourcing (marketplace-to-pipeline, sourcing velocity)
2. Diligence (frictionless VDR, shared-folder feel + SOC2 audit trails)
3. Valuation (SDE-to-GAAP/EBITDA automated modeling — the wedge feature)
4. Integration / PMI (playbook-driven, repeatable)
5. Multi-Asset Scale / Governance (centralized performance data, "Investor View")

The differences are not cosmetic. Enterprise adds Regulatory/Compliance (stage 5) and Deal Execution governance (stage 4) because $50M+ deals carry regulatory clearance and structured-execution complexity that $5-50M retiring-founder transactions skip. LMM adds Valuation (stage 3) — the SDE-to-GAAP conversion that bridges a retiring owner's cash-basis books to an institutional LP's GAAP/EBITDA reporting — because that is where the equity upside is secured, and no enterprise M&A platform ships it. The LMM terminal point is portfolio-centric: not "did this deal work" but "are all the deals compounding into LP-returnable EBITDA." The LMM research's own self-placement explicitly names enterprise Devensoft inside the "Enterprise Trap" — a cold, complex, rigid platform for institutional buyers — and concludes that enterprise-Devensoft and a LMM product are distinct products, not the same platform at different depths (per lmm-thesis §3, §10).

This distinction is load-bearing for the rest of the plan. It means the stage-to-product mapping (which lifecycle stages become standalone carve-out products) has two shapes, not one. It means competitive comparisons that rate Devensoft on enterprise lifecycle depth are correct but incomplete — the LMM lifecycle is a different arena with different competitors (Excel/Pipedrive/marketplaces, not Midaxo/DealRoom/Datasite). And it means the friction map below is enterprise-grounded — the demand evidence is enterprise — with LMM friction explicitly deferred until the demand is validated.

Coverage Claim — Stages 1-9 Owned, With Honest Depth Variation

The platform covers all nine enterprise lifecycle stages. That is a factual claim — Targets and Pipeline Management are real modules that clients use standalone for sourcing, not placeholders (clients run Targets/Pipeline without engaging PMI at all, and DevenLite — the planned Diversify carve-out — is built on exactly this surface). Due Diligence is owned through the Data Requests workflow and target-portal infrastructure. Integration Planning and Execution are the platform's deepest modules. Post-Close Value Realization is owned through Synergies tracking, KPI dashboards, and the recently requested "Value Realization" status field. Regulatory & Compliance is served through the audit log — a capability a customer-advocate called "the end of the conversation with the regulators" (per demo-dwight-customer-voice ~17:51).

But "covered" is not "dominant," and the competitive gradient is honest about where the platform leads, where it is present, and where it is thin:

Strong — the moat (stages 7-9: Integration Planning, Integration Execution, Post-Close Value Realization). The platform's PMI/Synergy/Legal depth is the lifecycle segment no competitor matches. The parity matrix records this as the "WIN" row: back-half lifecycle depth, the audit log as a unique enterprise moat, and the tenant isolation hierarchy as a productizable advantage competitors' multi-tenant SaaS does not replicate (per competitor-feature-parity-matrix § WIN/BEHIND/OPENING). No rival ships AI in this segment — the competitive convergence is on diligence-AI, not integration-AI — which is the opening M07 and M06 develop.

Present — real, not CRM-native (stages 1-2: Opportunity Identification, Pipeline Management). The platform's Targets module is a genuine sourcing tool — clients use it standalone — but it is not a CRM-native pipeline manager. The parity matrix rates Devensoft's Pipeline as partial ("Targets, not CRM") versus Midaxo and DealRoom, both of which ship full pipeline management with native CRM-like behavior (per competitor-feature-parity-matrix § D1 Lifecycle). This is a positioning question for the market and future-state sections, not a coverage gap: the platform has sourcing; the competitive question is whether sourcing should be deepened or whether that space belongs to the tools clients already pair it with.

Thin — demand-sparse (stages 4-6: Deal Execution, Regulatory & Compliance, Day One Readiness). These stages are owned — the audit log serves stage 5, the TSA management workflow spans stages 4/7/8/9, permissioning serves stage 6 — but the demand signal from them is light. No concentrated cluster of YouTrack issues or HubSpot tickets names stage 4 or stage 6 as a primary friction source. That thinness is partly structural (these are short, high-stakes windows where the tool is used intensively but briefly) and partly honest signal: the demand concentrates elsewhere, and this section does not fabricate friction the evidence does not support.

The competitive interpretation of this gradient — who wins which stage, and where Devensoft's opening is — belongs to the market and future-state sections (M03/M06). This section states the coverage fact: nine stages owned, depth varying honestly, with the strongest position exactly where the installed base and the demo voice both say it is.

Evidence-Backed Friction Map

The friction map is built from where the demand signal is loudest, and it is honest about where the signal is silent. Two independent demand streams — 108 YouTrack PRD issues (census-clean) and 640+ HubSpot support tickets (lower bound) — agree on where the lifecycle hurts, and both are corroborated by the customer-voice demo transcript. The friction clusters where the evidence clusters; the silence in stages 4-6 is not a gap to fill but a signal to report.

Due Diligence (stage 3) — the clearest lifecycle-specific friction cluster. YouTrack's Data Import/Export theme (11 issues, 10.2%) maps directly to DD mass-document upload/download workflows — including one of two dataset-wide Show-stoppers (LGCYEMA-447, "Wipro Import Load Screen") and a Critical (LGCYEMA-469). The HubSpot stream surfaces at least 24 DD-specific tickets with unusually rich customer language: target-portal timeouts losing work, follow-ups sent one at a time with no bulk option, documents auto-syncing to the deal room before internal review, and bulk-download gaps for external counsel and bankers (per hubspot-customer-voice § Due diligence workflow friction). The YouTrack Notifications/Email cluster (7 issues, 6.5%) adds CTL/NDA/Request-Update template failures and link-expiration windows — friction that hits at exactly the moment external parties are interacting with the platform under time pressure. The demand here is concentrated, specific, and actionable; it is the lifecycle stage where a Diversify carve-out (DevenDiligence) is most justified by existing-usage data.

Integration Execution (stages 7-8) — the center of gravity, with workflow-level friction. The demo transcript grounds this: the Data Requests workflow was built from a real 10,000-line-item deal pain point ("automate this damn thing" — Dwight quoting a coordinator whose hair was on fire, ~28:42), and the TSA management workflow lifecycle (costing → SLAs → decisions → change requests → terminations) was built from Accenture-scale integration experience (per demo-dwight-customer-voice § Pain-point → Resolution). The HubSpot stream shows Integration Planning/Execution as "the next major locus" after DD — activity-map behavior, workstream status, ownership, dependencies, and reporting asks cluster here. The friction is real but different in character from DD: DD friction is about external-party interaction under time pressure; Integration friction is about internal workflow complexity at scale — coordinating hundreds of tasks across workstreams, maintaining RAID integrity, and producing executive-ready status reporting on a compressed cadence.

Stages 5-6 (Regulatory & Compliance, Day One Readiness) — served lightly, no concentrated demand signal. The audit log serves Regulatory; permissioning and task-level access control serve Day One Readiness. But neither stage generates a concentrated friction cluster in the demand streams. This is honest signal, not a hole: these are short, high-stages where the platform's existing capabilities (audit trail, permissioning hierarchy, document repository) are used intensively but briefly, and where the pain is absorbed rather than escalated to support or PRD. The Programs and Legal pillars serve these stages, but neither produces the ticket-volume or issue-density that would indicate concentrated, addressable friction.

Stages 1-2 (Opportunity Identification, Pipeline Management) — present but not the friction locus. The demand evidence places sourcing friction well below DD and Integration in the support/PRD streams. Clients use Targets and Pipeline standalone — the demand is about the platform working (access, reliability, reporting), not about the sourcing workflow being broken. The positioning implication (Targets is not CRM-native; competitors ship stronger pipeline) is a market question, not a friction question, and is handled in M03/M06.

The pattern is clear: friction concentrates at the two ends of the lifecycle where the platform is most actively used — Due Diligence (external-party interaction) and Integration Execution (internal workflow complexity) — and the platform's moat sits in the back half where that workflow complexity is deepest. The silence in the middle is structural, not a coverage failure.

Data Requests + TSA Management — Load-Bearing Workflow Evidence

Two workflow modules anchor the Integration Execution friction story, and both trace to specific customer-voice evidence that makes them load-bearing for this section's stage 7-8 friction claim. These are not feature-listing entries; they are the capabilities a 35-year M&A integration veteran demonstrated as built-from-real-pain, and they are what differentiate the platform's back-half depth from competitors who ship generic project-management surfaces.

The Data Requests workflow was built from a real 10,000-line-item deal. The demo transcript captures the origin story verbatim: a coordinator at the client was managing data requests across hundreds of line items manually, reached the breaking point, and said "automate this damn thing" (per demo-dwight-customer-voice ~28:42). The resulting workflow module — which manages the structured request, response, follow-up, and document-exchange lifecycle between the integration team and external parties (targets, counsel, auditors) — reaches across lifecycle stages 3 (Due Diligence), 5 (Regulatory), and 7 (Integration Planning). Its design is enterprise-scale: task-level permissioning, audit trail, and the ability to control the release of new documents before internal review — a control the HubSpot demand stream explicitly asks for ("Once it is added to the released DD request, it automatically syncs with the deal room — it doesn't wait for another 'Send to Target' release," per hubspot-customer-voice § Due diligence workflow friction). The friction here is not "the module doesn't exist"; it is "the module exists, clients use it intensively, and the specific pain points (auto-sync timing, bulk-download, follow-up mechanics) are the actionable demand."

TSA (Transition Service Agreement) management was built from Accenture-scale integration experience. The demo walks through a full TSA lifecycle: costing → SLA definition → decisions → change requests → terminations (per demo-dwight-customer-voice ~36:12). TSAs are the contractual arrangements where the divesting entity provides temporary services to the carved-out business post-close — they are inherently stages 4 (Deal Execution, where they are negotiated), 7-8 (Integration, where they are managed), and 9 (Post-Close, where they are terminated). The TSA workflow is one of the platform's most differentiated capabilities: no competitor in the parity matrix ships a dedicated TSA management module. The HubSpot stream surfaces a related demand signal — a customer request for "TSA Exit" and "Value Realization" as structured status-report fields (per hubspot-customer-voice Source Ledger, ticket 43289976895) — confirming that clients are using the platform for TSA lifecycle management and want the reporting surface to reflect it.

These two workflows are load-bearing for the stage 7-8 friction claim because they demonstrate that the platform's Integration Execution depth is not generic project management. It is domain-specific machinery — built from the specific, named pain points of large-scale M&A integration — that no competitor replicates. The Data Requests workflow is also the architectural pattern validated for DevenDiligence's external-facing surface (an external DD respondent interacting via a scoped link, not a full auth session), which is the carve-out feasibility argument that belongs in the stage-to-product mapping below.

Stage-to-Product Mapping — The Tier-1 Vehicle, Pending the Choice

The stage-to-product mapping is intentional, and naming its shape explicitly is what prevents it from looking like an accidental coverage gap. The enterprise lifecycle's front end — sourcing (stages 1–2) — plus deal-execution capture is where the first standalone carve-out product would land: the Tier-1 slot is the LMM Vehicle Choice (per BTDR-0010 Am.2) — a deferred leadership decision among the LMM vehicles, with the mapping below carrying each vehicle's facts as options. The middle (stages 4–6: Deal Execution governance, Regulatory, Day One Readiness) is enterprise-complexity that LMM transactions skip entirely, and it is deliberately not carved out. A candidate full standalone (Devenfy) would fill the remaining gap, but its demand is unvalidated and pending the demand-signal instrument.

DevenLite = Targets + Tasking, front-scoped — the larger Tier-1-live option. DevenLite is a Tier-1-live vehicle option under the LMM Vehicle Choice (BTDR-0010 Am.2; its Am.1 facts survive as option properties) — the frontend-only carve-out over the DevenConnect API that productizes the platform's Targets module (sourcing/pipeline: stages 1–2, scored 62/75 in the agreed fit-to-need matrix) and tasking as deal-execution capture (stage 4, scored 52). Its scope is deliberately front-only: the PMI stages (6–9) are out of scope by design (scored 1/0/0/5 — Diego-agreed 2026-08-17), which retires the earlier "barbell" framing (stages 1–2 plus 7–9) that BTDR-0010 Amendment 1 carried — the honest scope is narrower than the barbell implied, and the BTDR/glossary sync amendment is owed on that basis. The caveat language is canonical: a sliver — the first step, not a complete solution — and standalone viability rides on the follow-ons (per fit-to-need-canonical, AGREED first pass 2026-08-18; per lmm-thesis §4 — LMM buyers need Search/Sourcing and integration help, not Deal Execution governance or Regulatory Clearance, because $5-50M retiring-founder transactions do not carry that complexity).

DevenDiligence = stage 3 (Due Diligence), the smaller Tier-1-live option. The DD carve-out is the most demand-justified Diversify product, because its demand is derivable from existing usage — the platform's Data Requests workflow, target-portal infrastructure, and document-exchange mechanics are already used for DD at the installed base. DevenDiligence productizes this surface as a standalone micro-product: external respondents interact via a scoped link (the external-facing-surface pattern validated by RequestUpdate extraction), the integration team manages the request/response lifecycle, and the data flows into the full platform when the deal closes. Its former Microsoft Clarity demand gate is retired (BTDR-0010 Am.1): under the choice encoding it is a Tier-1-live option in its own right — and, on the DevenLite branch, the follow-on required (with DevenFiles) for that vehicle's market viability, on the demand basis that users of DevenLite alone will struggle with multiple systems and want integrations or build-out (per BTDR-0007; per BTDR-0010 Am.1/Am.2). The Data Requests architecture (one-page form for external task reporting, scoped link, no full auth session) is the validated pattern for DevenDiligence's external-facing surface.

DevenFiles = the self-serve VDR, follow-on on the DevenLite branch. The third family member (lmm-candidate C5, named 2026-08-15) rides the vehicle choice: on the DevenLite branch it is the integration-reliant document layer that family needs for viability; if DevenLite is not chosen, its position returns to the signal-gated candidate set.

Stages 4-6 = enterprise-complexity, deliberately not carved out. Deal Execution governance, Regulatory & Compliance clearance, and Day One Readiness are the stages where enterprise deal complexity lives — the regulatory clearance, the structured execution frameworks, the cross-border governance that $50M+ transactions require. LMM transactions ($5-50M retiring-founder buys) skip these stages entirely; there is no LMM equivalent. Carving them into a standalone product would produce a product with no LMM buyer — it is enterprise machinery by definition. These stages stay in the enterprise platform. (DevenLite's deal-execution capture — tasking over documents and status — is not Deal Execution governance; the distinction is why stage 4 scores 52 for capture but stays uncarved as a governance product.)

Devenfy (candidate) = the gap-filler, pending validation. A full standalone LMM product would fill the remaining gap — a PLG-frontend product covering the full LMM 5-stage lifecycle (Search → Diligence → Valuation → Integration → Governance) rather than the front-scoped first step. But its demand is genuinely unvalidated — the LMM thesis is NotebookLM-derived ideation research, not primary market research — and the demand-signal instrument (fake-door, survey, or leadership-decided method) is the gate before any build commitment (per lmm-thesis §10 caveats; per BTDR-0007).

Reporting — The Cross-Cutting Underlay

Reporting is not a stage-specific capability — it is the plan-wide cross-cutting underlay that spans every stage, and it is the #1 demand in the evidence base (18/108 YouTrack, census-clean; ~86 HubSpot tickets, lower bound). Any per-stage friction map that omits reporting is incomplete by construction: reporting pain is not lifecycle friction at one stage, it is the underlay that makes every stage's output visible. The reporting capability itself — its current state, its gap, and the resolution path — is detailed in the current-state (M05) and core-capabilities (M08) sections.

Divestiture — Resolved as Cross-Cutting Deal-Type VariantDecide

The divestiture question — "is divestiture a 10th lifecycle stage, or is it out of scope?" — is resolved: divestiture is a cross-cutting deal-type variant, not a linear stage and not out of scope. A divestiture reuses Integration Planning and Integration Execution (stages 7-8) in reverse — separation instead of integration, divestiture-ready data rooms instead of integration-ready data rooms, TSA management for outgoing services instead of incoming services. The machinery is the same; the direction is opposite. This resolution reconciles the tension the original framing carried and is recorded as a BTDR candidate to file on sync.

The evidence supports this resolution from two directions. The customer-voice demo explicitly confirms divestiture is enterprise-functional: the transcript captures divestiture workflow discussion at ~04:19 and ~36:18, with the TSA management workflow serving both integration and separation scenarios (per demo-dwight-customer-voice § Capability → M&A-lifecycle). The LMM thesis confirms the market-scope boundary: the LMM research is buy-side only — divestiture is absent from the LMM thesis, and the $5-50M retiring-founder transaction type that defines the LMM sweet spot does not carry divestiture complexity (per lmm-thesis §4 — "Divestiture = NOT a stage; lifecycle ends at Multi-Asset Governance").

In domain language, divestiture lives in the platform's own pillar/persona vocabulary — it is a deal-type that exercises the same Programs, Synergy, and Legal pillars as an acquisition, in the reverse direction. It does not need its own stage number because it does not introduce a new capability surface; it reconfigures existing capabilities. Treating it as a 10th linear stage would break the lifecycle's sequential logic (a divestiture runs within the same stage sequence, not after it); treating it as out-of-scope would deny the enterprise-functional reality the demo confirms. The cross-cutting variant resolution is the honest read.

This section does not produce a new BTDR of its own — the resolution is documented here and flagged as a BTDR candidate to file when the workspace syncs back to the vault. The functional scope is settled; the market scope (N/A for LMM) is settled; no leadership decision is owed on the divestiture question itself.

Flag — Diversify Product Breadth Beyond the Tier-1 Vehicle ChoiceFlag

The stage-to-product mapping above frames a choice, not a commitment: the Tier-1 slot is the LMM Vehicle Choice (per BTDR-0010 Am.2) — a deferred leadership decision among the LMM vehicles, with DevenLite (Targets+Tasking, front-scoped — the first step, not a complete solution; its viability riding on the DevenDiligence (DD, stage 3) and DevenFiles (self-serve VDR) follow-ons on that branch) and DevenDiligence (in its own right) the Tier-1-live candidates. What remains genuinely open is the breadth question: which products beyond that choice should the Diversify track build?

The two-tier structure, as amended by the org preview and re-encoded by the choice (BTDR-0010 Am.1 → Am.2), answers part of it. Tier 1 = the LMM Vehicle Choice: leadership decides among the vehicles (DevenDiligence's former Clarity gate is retired; DevenLite's demand basis is leadership-asserted, with the lost-lead census as preferred data). Tier 2 (validate then build, need external signal) holds the remaining candidate products — C3 Valuation Engine, C4 SDE Quick-Val, C6 Sourcing (DevenFiles/C5 rides the vehicle choice) — gated on the demand-signal instrument (fake-door, survey, or meetings — leadership-decides on method), not on internal data (per BTDR-0007; per BTDR-0010 Am.1/Am.2).

The full standalone candidate — Devenfy — would fill the remaining gap by covering all five LMM stages rather than the front-scoped first step. Its demand is not validated. The LMM thesis research is NotebookLM-derived ideation — 10 pre-transformation research sources synthesized in April/May 2026 — not primary market validation, and the research itself flags this limitation explicitly (per lmm-thesis §10 — "Devensoft-originated ideation research, not primary market research; figures directional, not confirmed").

This stays a flag rather than a decision because the demand-signal instrument has not run and the Tier-1 vehicle itself awaits leadership's choice; the CEO's stance on the method (against fake-door; prefers having the product itself as the proof) is a leadership-decides item. The stage-to-product shape (a front-scoped first step as the leading option, with a candidate gap-filler) is correct regardless of how many products ultimately ship; the breadth (how many candidates graduate from Tier 2) depends on demand evidence that does not yet exist.

Flag — LMM Friction Deferred (Demand Unvalidated)Flag

The LMM lifecycle view in this section is a skeleton: five stages named, terminal point identified (Multi-Asset Governance), and stage-to-product mapping applied (the front-scoped vehicle option — DevenLite if chosen via the LMM Vehicle Choice, with the DevenDiligence/DevenFiles follow-ons; see Stage-to-Product Mapping — The Tier-1 Vehicle, Pending the Choice). What is deliberately absent from the LMM view is a per-stage friction map. The friction map above — the evidence-backed treatment of DD friction, Integration Execution friction, and honest silence for thin stages — is enterprise-grounded. It draws from YouTrack (108 PRD issues from the enterprise LGCYEMA project) and HubSpot (640+ enterprise support tickets). Neither evidence stream contains LMM demand signal, because Devensoft has never sold to the LMM segment.

Fabricating LMM friction from an unvalidated thesis would be the kind of stance drift the validation protocol guards against. The LMM thesis research does name theoretical friction points — sourcing velocity without looking institutional, accrual-conversion value leakage at the Valuation stage, ERP-migration delays post-close, one-and-done churn risk from the finite search window — but these are research-hypothesized friction points from NotebookLM-derived ideation, not demand-evidenced friction from actual LMM users (per lmm-thesis §2, §7). The research itself flags this: "Demand is unvalidated — thesis rests on the displacement-gap macro + the False-Choice framing; neither has conversion evidence" (§10).

The theoretical LMM friction belongs to the customer-segments and capabilities sections (M04/M08), where the LMM persona thesis and the capability requirements for a LMM product are developed. M09's role is to present the lifecycle shape honestly: enterprise friction is evidence-grounded; LMM friction is demand-unvalidated and deferred. When the demand-signal instrument runs — whether it confirms, narrows, or kills the LMM thesis — the LMM friction map will populate from real demand data, not from a research hypothesis.

This is named as a flag rather than a gap to fill because the deferral is the correct posture, not an oversight. The M12 kill-conditions (four conditions that must all hold for the LMM thesis to beat defend-enterprise) are the structured gate; M09 feeds those conditions by providing the lifecycle shape but not by pre-answering whether LMM demand exists.

Enablers + Roadmap — Intro Summary

Yes — the plan is executable inside the ~9-month horizon, but only along a single critical path whose geometry the team does not get to choose. The dependency DAG (the truth model of what unlocks what) runs F6 → F1 → T2 → T3 → T7: the DI/runtime bridge completion (F6) gates the auth-provider migration (F1), which gates the DevenConnect API surface (T2), which gates every extracted frontend and Diversify product (T3), which gates the GTM motion (T7). At realistic allocation, F6 + F1 alone consume 14–21+ serial dev-weeks — roughly four to seven months of the nine — before the first extracted surface can ship (per fundamentals-sizing-brief F6/F1; per BTDR-0008). The capacity envelope — modeled live in recommendation-matrix.html at a three-tier effective-FTE ramp with an agentic multiplier — currently resolves to 105.3 dev-weeks total, of which ~43.8 dev-wk is committed to the Retain/fundamentals floor and ~61.5 dev-wk remains as headroom for Grow and Diversify. That envelope holds; what it does not hold is the illusion that everything can be done at once.

The credibility of this roadmap rests on a property the prior, stale directives lacked: every dependency edge traces to a structural prerequisite with an effort-calibrated sizing, not to a phase label on a slide. The fundamentals clusters (F1–F7) are bottom-up sized from analogous completed YouTrack epics with actual spent-time data, calibrated by the empirically observed 1.5× overrun ratio for large work (per fundamentals-sizing-brief §2–§3). The critical path is not a recommendation — it is the longest serial dependency chain through the DAG, and it is the binding constraint whether or not the plan acknowledges it. The phase gates are not calendar milestones — they are hard/soft gates mapped to the capacity ramp, protecting the serial chain while allowing parallel work at risk. And the fundamentals-first sequencing rule per BTDR-0004 G4 — capacity is smaller than ambition, so the Retain/fundamentals floor sequences ahead of Grow/Diversify when the two compete for the same dev-week — governs the entire roadmap, coexisting with the 90/10 attention firewall that guards focus on a separate axis (per BTDR-0004 G4).

That brings the plan from the M&A journey (§09) to its execution mechanics. The prior section established what the capabilities serve across the lifecycle; this section establishes what unlocks what, in what sequence, against what capacity constraint. The remainder of this section presents the constraint model — the DAG, the tracks, the critical path, the phase gates, the capacity ramp, and the enabler stack — that M14's recommended path runs against. M10 owns the dependency truth; M14 owns the recommended judgment. The next turn — the commercial model that funds this roadmap — follows in §11.

The Dependency DAG and the Seven Execution Tracks

The roadmap's truth model is a dependency directed acyclic graph (DAG) — a network of work items and the structural prerequisites that connect them. The DAG states what unlocks what. It does not state what to build first — that is a judgment call, owned by M14's recommended path. The DAG's value is precisely that the two are separable: leadership can debate sequencing, capacity allocation, and appetite against a fixed constraint surface, rather than negotiating the constraints themselves (per the M10 decision review).

The DAG decomposes into seven execution tracks, each carrying an R/G/D lens:

The DAG edges — each evidence-grounded, not asserted — define the unlock relationships between these tracks. F6 (DI/runtime bridge completion) gates F1 (auth migration) because auth/session refactoring requires clean dependency injection (per fundamentals-sizing-brief §5). F1 gates T2 (DevenConnect) because scope-based authorization requires the managed identity provider's native scope/role model (per BTDR-0008). T2 gates T3 because every Diversify product is a frontend over the DevenConnect API seam (per BTDR-0007; per BTDR-0006). T2 gates T4 because BYOM routing and reporting-AI ride the same API surface (per BTDR-0004). T6 informs T3's remainder: the demand-signal work (T6; method leadership-decides) indicates which Diversify products beyond the Tier-1 vehicle get built — the Tier-1 slot itself is the LMM Vehicle Choice, a leadership decision rather than an instrument verdict (the Tier-1-live candidates C1/C2 are not signal-gated) (per BTDR-0010 Am.2). T3 gates T7 because GTM needs a product to sell.

Two structural nuances the DAG surfaces. First, the in-horizon AI wins — the DevenMCP sliver (G5a), BYOM routing (G2), and the reporting-AI PoC (G3) — run above and parallel to the floor with no F1 dependency. They consume the existing DevenConnect surface, not the F1-gated future one (per BTDR-0004). This split is the key insight: not all AI work is behind the critical path. Second, within T3 the extraction sequence is RequestUpdate (POC) → the chosen Tier-1 vehicle (late-window, pending the LMM Vehicle Choice) — a tiny one-page form SPA validates the extraction pattern at minimal blast radius before any Diversify product ships, and whichever vehicle leadership chooses delivers late-window behind the fundamentals floor and Phase 1: DevenDiligence (~3–5w) is the smaller extraction; DevenLite (~32w minimum-guess) the larger, with the DealSetupProcessing coupling (24/69 dependencies) dominating its LOW-confidence sizing and carried openly (per the M10 decision review; choice encoding per BTDR-0010 Am.2). The DAG does not hide these constraints; it makes them visible as structural facts rather than failures of will.

The Critical Path — F6 → F1 → T2 → T3 → T7

The critical path through the DAG is F6 → F1 → T2 → T3 → T7 — the longest serial dependency chain, and the binding constraint on the 9-month plan. It is not a recommendation; it is a structural fact. Every item on this chain must complete before the next can begin, and no amount of parallel headcount clears a serial dependency faster (per the M10 decision review; per fundamentals-sizing-brief).

The chain's geometry, node by node:

F6 — DI/runtime bridge completion: 5–8 dev-weeks remaining. LGCYEMA-511 Phase 2 (Static → Non-Static) is roughly 70% spent, but the unspent 30% is the harder back half — six unstarted children (EmaHangfire, EmaServices, ImExData, EmaControllers, wwwroot, BusinessObjects-Followups), all unestimated in YouTrack, including the high-blast-radius EmaControllers request-handling layer. Sized by analogy to completed children (3d–1w3d actual each) × the 1.5× calibration ratio for large work (per fundamentals-sizing-brief F6). F6 gates F1 because auth/session refactoring requires clean dependency injection — you cannot untangle CurrentState (325 symbols, 69 direct dependents) or introduce per-request service resolution while the static→non-static bridge is half-built.

F1 — auth-provider migration: 9–13 dev-weeks (pre-re-sizing). Replaces Forms Auth + OAuth client-credentials with Clerk.com or Auth0 — buy-not-build (per BTDR-0008). The sizing brief scoped F1 at the narrow enforcement surface (9 controllers, 16 endpoints, one class-level [Authorize]), but the actual scope is a full provider integration plus dual-system migration against a larger controller/endpoint count — re-sizing is a live dispatch item (see §08). F1 gates both the floor's access-pain resolution and every Diversify product. It is the critical-path intersection where Retain and Diversify tracks share a single gate: if F1 slips, both slip.

F6 + F1 = 14–21+ serial dev-weeks. At the realistic capacity allocation of the ramp curve (~2.0 effective FTE in months 1–2, ramping to ~3.5 FTE in months 3–6), this serial chain alone consumes ~4–7 months of the 9-month horizon. This is why the in-horizon AI scope had to be bounded so narrowly, and why the DevenMCP full runtime core (G5b) is a post-horizon north-star rather than a deliberately-not-build (per BTDR-0004). The chain is not headcount-fixable — two developers or five, the serial dependency takes the same calendar time because each node cannot start until its predecessor finishes.

T2, T3, T7 follow in strict sequence. T2 (DevenConnect MVP) cannot begin its build until F1 provides scope-based auth. T3 (Extraction) cannot ship its first extracted surface until T2 provides the API seam. T7 (GTM) cannot sell what T3 has not shipped. Each node's start date is the predecessor's end date — there is no compression available without cutting a node, and cutting a node breaks the chain's structural integrity.

The critical path tells leadership one thing plainly: the first four to seven months of the plan are spent on prerequisites, not on visible product. The in-horizon AI wins that are visible — the DevenMCP sliver, BYOM routing, the reporting-AI PoC — are precisely the ones that run parallel to this chain, above the floor, on the existing API surface. Everything that depends on the chain waits its turn. The phase gates (next) are the mechanism that protects the chain while allowing parallel work at risk.

Phase Gates and the Capacity Ramp

The roadmap progresses through three phase gates, each mapped 1:1 to a stage of the capacity ramp curve. The ramp is not a flat FTE assumption — it is a three-stage curve that reflects the reality of staff-aug onboarding drag and pizza-team coordination cost as the team scales (per the M10 decision review; modeled live in recommendation-matrix.html).

Phase 0 — Stabilize & Validate (≈ months 1–2, Ramp-in ~2.0 FTE)

The team enters at Diego + Roman, carrying a −30% transition drag from the vendor handoff. Parallel work in this phase: F6 completion (Diego), F2/F3/F4 quick wins (Roman + ramping contractor), T6 fake-door launch, T5a CRM hygiene, T2 DevenConnect scoping and specification.

• Hard gate (must-meet to advance): F6 complete — unblocks F1. Observability baseline live (F2 dependency checks shipped). Team ramped to ≥3 FTE effective.
• Soft gate (at-risk-ok): Fake-door demand signal captured. CRM hygiene pass complete.

Phase 1 — Modernize & Connect (≈ months 3–6, Build ~3.5 FTE)

Two agenic seniors have ramped past onboarding drag. Parallel work: F1 auth-provider migration (the primary critical-path item), DevenConnect MVP build, F5 infra resilience, T5b baseline census, T4 reporting-AI floor PoC (conditional on semantic-model gate), T3a RequestUpdate extraction POC.

• Hard gate: F1 complete — unblocks extraction and PLG self-serve. DevenConnect MVP live — unblocks Diversify and the AI Satisfy tier.
• Soft gate: Reporting-AI ship-or-kill decided. At least one extraction surface in progress.

Phase 2 — Diversify & Scale (≈ months 7–9, Accelerate ~4.5 FTE)

Full team at velocity. Parallel work: T3b DevenDiligence extraction ships, T4 AI Satisfy tier, T7 GTM motion, remaining fundamentals (F5, retention work), T5c evidence-checklist scorecard (month 0 → month 9 delta).

• Hard gate: Diversify product shipped + validated. Contractual floor sustained.
• Soft gate: AI posture credible (floor + Satisfy live). GTM motion active.

The hard/soft distinction is load-bearing. Hard gates protect the serial critical-path chain — F6 must complete before F1 starts; F1 must complete before extraction ships. These are structural; slipping them slips everything downstream. Soft gates allow parallel at-risk work to continue if a gate is missed — the reporting-AI PoC can ship-or-kill late without blocking the extraction track, and the fake-door signal can arrive in month 3 instead of month 2 without stalling Phase 1's hard gate. The gates are not calendar deadlines; they are dependency-structural checkpoints mapped to the ramp stages.

The capacity envelope (current snapshot)

The three-tier effective-FTE model (core 1.5 FTE with no onboarding drag; contractors from 1.5→3 FTE carrying onboarding drag; contractors past 3 FTE carrying additional pizza-team coordination drag), run at the Diego+Roman preset (5.5 FTE, 39w, 30% onboarding drag, 0.4 pizza-team drag above 3 FTE, 1.6× agentic multiplier, 2h/8h non-engineering overhead) and modeled live in recommendation-matrix.html, currently resolves to:

This is the constraint surface M14's recommended path reasons against. The in-horizon net-new draw is branch-dependent on the LMM Vehicle Choice: a DevenLite choice — the AI scope plus the DevenLite build and its DevenDiligence follow-on — commits ~47–53 dev-wk, which fits inside the ~61.5 dev-wk headroom barely: the residual ~9–15 dev-wk is what remains to absorb the onboarding-drag and pizza-team-coordination penalties the three-tier model builds in, and DevenLite's coupling-dominated ~32w minimum-guess sizing is the dominant uncertainty in that arithmetic. The envelope is dynamic — leadership owns the headcount and appetite call that sets it — and the recommendation-matrix.html tool makes the trade-off visible in real time rather than hiding it in an asserted ratio. If the envelope compresses (lower FTE, optimistic agentic multiplier, or the DevenLite guess proving optimistic), fundamentals-first sequencing per BTDR-0004 G4 dictates that the Retain floor is protected and the Grow/Diversify scope compresses first — the chosen vehicle leading that compression — not the other way around (per BTDR-0004 G4; BTDR-0010 Am.2).

The Three-Layer Enabler Stack

Execution capacity is not just headcount. It stacks across three layers, each of which must be in place for the roadmap to hold (per the M10 decision review).

E1 — Organizational

The organizational layer carries three enforcement mechanisms:

Staffing ramp. The capacity ramp curve (Phase 0 → Phase 2) is the staffing model: Ramp-in at ~2.0 effective FTE in months 1–2 (Diego + Roman, −30% transition drag), Build at ~3.5 FTE in months 3–6 (two agenic seniors ramped), Accelerate at ~4.5 FTE in months 7–9 (full team). The ramp replaces the incumbent offshore vendor (~$650K sunk) with agenic seniors at materially lower cost. A vanity-hire veto stands: the 2020 modernization relapsed in part because it overstaffed with the wrong skill mix, and this plan does not repeat that pattern (per the M10 decision review; per BTDR-0004 — G4).

90/10 attention firewall. The firewall guards attention and energy, not dev-week budget. It keeps the team's focus on transformation rather than relapsing into reactive client work — the exact failure mode that derailed the prior modernization cycle. It operates through four mechanisms (per the M10 decision review): Diego owns enforcement; Eric overrides — the documented floor spec is the reference. Escalation is severity-triggered: prod-down, data-loss, and security incidents go to the floor immediately; everything else queues behind the transformation work. A monthly drift check measures how much attention has drifted to legacy. A circuit-breaker: if drift exceeds 30%, a stop-the-line review is triggered. The firewall coexists with — and does not replace — the fundamentals-first sequencing rule per BTDR-0004 G4: the firewall guards focus (are you working on transformation at all?); fundamentals-first sequencing governs capacity direction (when capacity is smaller than ambition, which sequences first?). Together they answer the two questions a leadership audience will press first: are you protecting the rebuild from client-urgency drift? (yes — the firewall) and when capacity is scarce, what sequences first? (the fundamentals — per BTDR-0004 G4).

ADKAR as a lens. Change management is applied as a soft-gate checklist at each phase transition — not a formal Prosci program. The five ADKAR stages (Awareness → Desire → Knowledge → Ability → Reinforcement) are checked at each gate: does the team understand why the phase matters (Awareness), want to execute it (Desire), know how (Knowledge), have the capability (Ability), and will sustain it (Reinforcement)? A failed ADKAR check does not block a hard gate, but it flags organizational risk before a phase begins (per the M10 decision review).

E2 — Tooling / Process

The tooling layer carries three workstreams, deliberately deduped from the fundamentals cluster — no F1–F7 overlap:

Azure migration completion. KeyVault enablement (currently disabled — DB password, SMTP, OAuth, and Teams/PowerBI secrets committed in plaintext), Blob storage migration, and Azure AppConfig adoption. This is configuration-management refactoring LGCYEMA-2, currently 50% complete (6/12 subtasks done; 2 still New) (per fundamentals-sizing-brief F4). The plaintext-secrets exposure is a security-audit failure waiting to happen and gates enterprise IT approval at renewal — making this Phase-0 hygiene, not deferred tech debt.

CRM hygiene pass. The HubSpot support pipeline (640+1,200 tickets mined over 24 months) is the primary customer-voice evidence stream, but the portal's data quality is directional, not census-clean (CX/CSAT/NPS fields effectively empty; ~38% of recent tickets are spam/outreach). The hygiene pass is Phase-0 instrument work (T5a) that makes the CRM a reliable M13 measurement instrument, not just an evidence mine.

Test-coverage scaffolding. TernaryLibrary — the custom-MVC debt at the platform's coupling center — has zero automated test coverage. Any refactoring on the critical path (F6's ConnectionManager = 29 direct imports, 275 impacted symbols; CurrentState = 325 symbols, 69 dependents) requires full manual integration testing today. The scaffolding does not attempt full coverage in the 9-month window; it establishes the safety net for the critical-path refactoring that cannot proceed safely without it (per fundamentals-sizing-brief F6).

E3 — Technical Prerequisites

The technical layer is the DAG. The dependency chain F6 → F1 → T2 → T3 → T7 is the technical prerequisite structure — the work that must land before downstream work can begin. E3 is not a separate workstream; it is the acknowledgment that the enabler stack's foundation is structural, and the DAG (presented in §02 and §03) is its complete specification. The plan does not pretend these prerequisites can be shortcut; it sequences them honestly.

Beyond the Horizon — the Nine Deferred Items

The 9-month plan is the next battle, not the war. Nine work items are deliberately deferred — not cancelled, not deliberately-not-build, but positioned as the north-star agenda for the horizon that follows. Each is named here so investor-grade scrutiny sees what the plan chooses not to do and why (per the M10 decision review). (DevenLite's history on this list is a round trip: it left 2026-08-15 when the org preview committed it as the Tier-1 build (BTDR-0010 Am.1), and it re-enters 2026-08-18 — the Tier-1 slot is now the LMM Vehicle Choice (BTDR-0010 Am.2), a deferred leadership decision, so DevenLite sits here again pending the choice: deferred, not struck; it enters the plan only if chosen — see M14.)

The distinction between deferred (Beyond the Horizon) and deliberately-not-build (NB1–NB5, owned by M14) is load-bearing. Deliberately-not-build items are walked away from — the plan recommends against them regardless of horizon. Deferred items are sequenced out — the plan intends to do them, just not in this battle. The critical-path serial chain (F6 → F1) is the reason most of these items sit beyond the horizon: the four-to-seven-month prerequisite window leaves insufficient serial capacity for the items that depend on it. Naming them explicitly — rather than letting them accumulate in silence — is what makes this a plan with a boundary, not a wishlist without one.

F1 Resolved — the Auth-Provider MigrationDecide

F1 — the largest single swing factor on the critical path — is resolved. The M10 decision review and BTDR-0008 settled the question the sizing brief left open: what kind of work is F1?

The decision: outsource auth to a managed identity provider — Clerk.com or Auth0. Replace both Forms Auth (the webmonolith's ASP.NET cookie-based system) and OAuth client-credentials-only (the WebAPI's token-presence-based system, per ADR-0020/0022) with a single managed service. Buy-not-build: delete auth code, integrate a provider. This decision does five things at once (per BTDR-0008):

1. Removes the major auth code from the codebase — the OWIN OAuth server, Forms Auth, token storage, custom auth middleware. Net debt reduction, not net addition.
2. Unifies auth across the webmonolith and WebAPI through one external service, eliminating the dual-system maintenance burden.
3. Provides scope-based authorization natively — the managed provider handles scopes, roles, and claims that the current token-presence-only model cannot enforce (any valid token grants access to all endpoints today; per fundamentals-sizing-brief F1).
4. Is the hard pre-requisite for all Diversify products — DevenLite, DevenDiligence, and Devenfy all need proper user-auth flows (PLG self-serve, multi-tenant, external-facing surfaces) that client-credentials-only cannot support.
5. Makes the OWIN→Core "no drop-in equivalent" problem moot — you are not looking for a drop-in replacement; you are replacing the entire auth model.

This supersedes ADR-0020 and ADR-0022 (both Target-Proposed, 2026-04-16). The question those ADRs posed — "how do we re-implement OWIN auth in ASP.NET Core" — is no longer the question. The question is now "which managed provider do we integrate," which is a vendor-selection exercise, not a build exercise. Both ADRs should be marked superseded by this BTDR.

Why this was a `Decide` and is now `Resolve`. The prior §10 directive carried F1 as an open Decide — a decision to be made. BTDR-0008 (Accepted 2026-08-11) closed it. F1 is no longer whether to migrate auth or how; it is which provider (Clerk vs Auth0), which is a downstream implementation decision not gated by this BTDR. The resolution converts F1 from a roadmap risk into a roadmap fact — albeit one whose sizing still needs updating (see §08).

What does not change. F1's position on the critical path does not change. It still gates every extracted surface and every Diversify product. It is still serial — not headcount-fixable. The managed-provider path may be faster than re-platforming OWIN auth in-house, but the larger controller/endpoint surface offsets that, and the net sizing effect is pending the re-sizing dispatch. What changes is that the team now knows what they are building — a provider integration — rather than debating whether to build an auth server. That certainty is itself an enabler: scoping, vendor evaluation, and integration design can proceed in parallel with F6 completion, rather than waiting for a decision that has already been made.

Flag — Sizing Dispatch ItemsFlag

Two sizing gaps remain open. Neither blocks the roadmap's structure — the DAG, the phase gates, and the critical path are locked. Both affect the roadmap's calendar — the specific month in which each phase gate lands. They are named here as dispatch items, not silently asserted.

F1 re-sizing (dispatch)

The sizing brief's F1 estimate (9–13 dev-weeks) was scoped to the pre-decision surface: narrow scope-enforcement on the existing 9 controllers / 16 endpoints, deferring re-platforming (per fundamentals-sizing-brief F1). BTDR-0008 changed the scope: F1 is now a full managed-provider integration (Clerk/Auth0) replacing both legacy auth systems against an actual controller/endpoint count larger than 9/16. The buy-not-build path may be faster per-surface than building, but the larger surface offsets that — net effect TBD pending the re-sizing.

Why it matters: F1 sits on the critical path. A re-sized F1 that runs longer than 13 dev-weeks pushes the Phase 1 hard gate (F1 complete) later, which compresses Phase 2's window for Diversify extraction. A re-sized F1 that runs shorter (because buy-not-build is faster than the sizing brief's build assumption) pulls the gate earlier and opens headroom. Either way, the phase-gate calendar cannot lock until F1 is re-sized against the real surface.

Dispatch: scope F1 against the actual controller/endpoint inventory plus the Clerk/Auth0 integration surface. Apply the 1.5× calibration rule (≥1w estimates) from the empirically observed effort-calibration data across 1,118 YouTrack issues (per fundamentals-sizing-brief §2).

T2 DevenConnect sizing (dispatch)

T2 (DevenConnect) is unsized. The DevenConnect API surface exists commercially today — the architecture and dual-role are real — but the MVP build scope for Phase 1 (how much API surface exists versus how much needs building to support extraction and the AI Satisfy tier) is not yet scoped (per the M10 decision reviewdispatch items).

Why it matters: T2 sits on the critical path between F1 and T3. If the DevenConnect MVP build is larger than assumed, the Phase 1 hard gate (DevenConnect MVP live) pushes later, delaying the first extracted surface. If it is smaller — much of the surface already exists commercially — the gate may clear quickly after F1, and the headroom opens earlier than projected.

Dispatch: inventory the existing API surface against the extraction + BYOM routing requirements. The DevenConnect commercial architecture defines the commercial shape; this sizing dispatch defines the build shape.

Both dispatch items are fact-finding, not decision-finding. The decisions — F1 = auth-provider migration; T2 = the API seam — are locked (per BTDR-0008). What remains is measurement, and measurement is what makes the phase-gate calendar go from structural to specific. Until the dispatch items close, the phase-gate windows in §04 carry an implicit ±1-month uncertainty band.

Commercial Model & GTM - Intro Summary

The return on this plan comes from three revenue engines, not one. Enterprise ARR on Devensoft v24 is the floor that funds everything else - a seat-license base augmented by a premium "Managed Success" tier that unbundles the twenty-plus onboarding sessions the installed base actually consumes, plus the renewal-defended ~$200K relationship and automotive accounts the prior section named (per BTDR-0007 - Decision 2/3). Deven Connect is the margin play under a tiered model (BTDR-0009) - an API commercial layer whose vault-draft $75k MSRP was theoretical / market-rejected / $0 realized (BTDR-0007 Am.1; see 02_deven-connect-commercial-layer) and is now superseded by tiered pricing: free for current customers (Retain), consumption-based monthly for new enterprise (Grow) and Diversify where economics warrant - earning its seat three ways: as a Retain anti-churn value-add, as an itemized enterprise consumption line, and as the invisible infrastructure under every Diversify product (per BTDR-0009; per EMAKBAI-deven-connect - draft/provisional). ⚠ Am.2: $75k MSRP superseded by tiered API pricing — R=free, G/D=consumption-based monthly. Diversify products - DevenLite (Targets+Tasking), DevenDiligence (DD carve-out), and the candidate Devenfy - are the upside bet, self-serve PLG vehicles the existing enterprise motion cannot reach. The GTM motion splits cleanly with that revenue shape: the enterprise base stays relationship/demo-led with a digital-presence fix and a formalized advisor channel; the Diversify side runs fake-door-gated, community-led PLG through search-fund and PE-association channels, with the fake-door itself starting week 1 in parallel.

The credibility of this commercial model rests on what its tiers trace to. The vault's draft $75k Deven Connect MSRP was theoretical and market-rejected ($0 realized) — not invented, but never contracted (per EMAKBAI-deven-connect - draft/provisional; BTDR-0007 Am.1); BTDR-0009 (Am.2) supersedes it with a tiered model (R=free, G/D=consumption-based monthly; specific rates TBD at M11 resolution), so the margin play is now live but rate-undetermined, not a dead anchor. The enterprise onboarding friction the premium tier unbundles is field-evidenced - twenty-plus structured onboarding sessions per enterprise engagement, thirty-to-forty percent feature utilization at the installed base, a 1.5-out-of-5 maturity score against the Modus dimensions (per modus-enterprise-insights). The Diversify pricing band is benchmarked to CT Acquisitions' actual $10k-$40k/year engagement pricing for diligence software, third-party-captured, not invented (per competitor-raw-capture-2026-08-06). The structural pricing pressure the split product line answers is DealRoom's per-deal, unlimited-user model, which directly attacks the seat-license-plus-onboarding friction this plan inherits rather than denies (per competitor-profile--dealroom). Where the commercial evidence runs thin - the enterprise $/seat, the LMM CAC and LTV/CAC, the PLG infrastructure cost - those gaps are named as flags in this section rather than smoothed over with fabricated precision.

That commercial shape - three engines, two motions, one fake-door gate - is what funds the roadmap the prior section laid out. The enablers and sequencing of M10 only stand up if the revenue architecture carries them: the enterprise floor funds the fundamentals, Deven Connect funds the margin and the API build that the Diversify products depend on, and the Diversify upside - gated behind a real demand signal - is what turns the plan from a defense into a growth story. The remainder of this section substantiates each layer against the evidence, names the decisions locked in BTDR-0007, and flags the unit-economics gaps leadership owes an answer on before the commercial model can be called complete.

Deven Connect - the Commercial Layer

Deven Connect is the OpenAPI commercial surface of the Devensoft platform - the seam through which external systems, integrations, and (in the transformation target) AI-driven execution reach into the platform's data and workflows. The vault records it as a "platform capability, not a managed service" - the commercial principle being that Devensoft exposes the API, the client owns the integration build against it, and integration work is not a Devensoft-delivered service (per EMAKBAI-deven-connect - draft/provisional). That framing matters commercially because it lets a single API layer carry two different value propositions to two different buyers without doubling the engineering.

The architecture and the dual-role strategy are real. The price point is now live and tiered. The vault's draft commercial principles recorded a two-tier structure - $0 early-access at the low end and a $75k MSRP at the enterprise tier, with a 25% C-level discount pattern (same source, status: draft). The market did not accept that price: the $75k MSRP was theoretical, clients rejected it, and zero active clients subscribed on those terms - realized Deven Connect revenue was $0 (per commercial-pricing-baseline §3 - Diego correction 2026-08-12). BTDR-0007 Am.1 downgraded $75k from the working commercial frame to theoretical/aspirational; BTDR-0009 (Am.2) supersedes the dead anchor with a tiered model aligned to the three strategic objectives: free for current customers (Retain anti-churn value-add — a fair-use tier with low consumption limits, enough to enable some integrations but not full syncs, bulk downloads, or high-volume usage; customers needing higher limits upgrade to the paid consumption tier), consumption-based monthly for new enterprise clients (Grow - lower barrier than a $75k upfront MSRP), and consumption-based monthly for Diversify where the economics warrant it (per BTDR-0009). The specific consumption rate structure (unit, tiers, caps) remains TBD at the M11 resolution; what is locked is the model - tiered by R/G/D. ⚠ Am.2: $75k MSRP superseded by tiered API pricing (BTDR-0009): R=free, G/D=consumption-based monthly.

The strategic role of Deven Connect in this plan is therefore larger than any single price point, and that distinction is load-bearing. BTDR-0007 locks it as the dual-role layer (BTDR-0009 preserves the mechanism; only the pricing model changes): the same API is itemized at the enterprise tier under the Grow consumption-based rate (where the client supplies the LLM under the model-governance rule - BYOM = zero Devensoft inference cost, so the consumption revenue is margin-positive once rates land) and simultaneously bundled invisibly into the Diversify PLG tier pricing where the LMM buyer never sees a separate API charge (per BTDR-0007 - Decision 5; per BTDR-0009). At the Retain tier the same API is granted free to current customers as an anti-churn value-add (a fair-use tier with low consumption limits — enough to demonstrate value and prevent churn, not enough to substitute for paid consumption). That dual-role + tiered-pricing combination is the commercial mechanism that lets one API investment serve all three objectives: Retain deepens stickiness at zero friction, Grow itemizes at a credible consumption rate, Diversify products inherit it bundled. It is also the seam where AI earns its seat at the enterprise tier - the layered AI posture in BTDR-0004 places integration/PMI seam-AI as the differentiation anchor, and that anchor runs through Deven Connect by construction (per BTDR-0004 - G1). The Zapier integration (Phase 1 build, M14 G4) is the go-to-market mechanism for the API: once available, Zapier zaps become the self-serve distribution surface - advertised as "all integrations enabled by Zapier" - making the API accessible to non-developers and giving Grow a credible integration story on top of the consumption rate. Zapier itself carries a tiered model across all objectives: a base level allowing very low usage (a handful of personal, user-specific integrations) available free/low-cap, with tiered pricing for higher limits upgradeable across Retain, Grow, and Diversify alike — so Retain customers get a free Zapier tier (low usage) alongside their free fair-use API, while Grow and Diversify customers can tier up Zapier spend alongside their consumption-based API (per BTDR-0009 Am.). The honest framing: Deven Connect is a real, dual-purpose commercial asset whose revenue contribution moves from $0 under the dead $75k anchor to a live tiered model (R=free, G/D=consumption-based monthly; specific rates TBD at M11 resolution).

The Product Family and the Ladder

The Diversify side of this plan is not a single product. It is a family of segment-aligned products, each a frontend variation over Deven Connect, all migrating natively upward through a graduation ladder whose top rung is Devensoft v24 (per BTDR-0007 - Decision 7). Three products are confirmed in scope; one is a candidate pending demand validation. The family's build order is now framed by the Tier-1 choice rather than anchored to one product: the LMM Vehicle Choice (per BTDR-0010 Am.2) holds the first step as a deferred leadership decision — DevenLite and DevenDiligence the Tier-1-live candidates — with DevenDiligence + DevenFiles named as the follow-ons the DevenLite branch requires for viability (Am.1's facts surviving as option properties).

The ladder connects them: Devenfy -> DevenLite -> Devensoft v24, with DevenDiligence cross-cutting the lower two tiers as a standalone diligence-thesis product available independent of the wider platform (per BTDR-0007 - Decision 7). Confirmed products: DevenLite (Targets+Tasking carve-out), DevenDiligence (DD carve-out), and Devensoft v23->v24. Candidate products pending the demand signal: Devenfy (full LMM standalone) and an SDE Calculator micro-product. The two-tier amendment lineage shapes the family narrative: the org-preview amendment (BTDR-0010 Am.1, 2026-08-15) had put DevenLite in the Tier-1 first-step role with DevenDiligence + DevenFiles (the self-serve VDR; alias of lmm-candidate C5, named 2026-08-15) as the named follow-ons required to make DevenLite market-viable — users of the first step alone will struggle with multiple systems and want integrations or build-out, which is exactly what the follow-ons supply — and the LMM Vehicle Choice re-encoding (BTDR-0010 Am.2, 2026-08-18) now holds the Tier-1 slot as a deferred leadership decision: that follow-on structure survives as the DevenLite branch's viability condition, with DevenDiligence also a Tier-1-live option in its own right.

How the ladder works

Five prescriptions govern the graduation mechanism, and each addresses a specific commercial risk in the Diversify thesis (per the M11 decision review):

• Data continuity. Each tier's data migrates natively upward: a DevenLite tenant's Targets and Tasks become the same records inside a full EMA instance; a DevenDiligence tenant's Q&A dataset becomes the same dataset inside full EMA diligence. The buyer who outgrows the lower tier does not start over.
• Graduated pricing credit. Spend at the lower tiers credits toward the upgrade. The LMM buyer who converts to enterprise is not penalized for having started small.
• Feature gates via Deven Connect. Each tier exposes more API surface - the product is the scope of API access. The same API layer (now tiered by R/G/D per BTDR-0009 — ⚠ Am.2: $75k MSRP superseded by tiered API pricing: R=free, G/D=consumption-based monthly) is what unlocks each rung, with the lower tiers exposing a scoped subset of the same capabilities.
• Self-serve -> high-touch transition. Lower tiers are PLG (self-serve, low-friction); the enterprise tier remains relationship/demo-led. The ladder is the bridge between two motions that today do not connect - the LMM buyer who would never survive twenty-plus onboarding sessions enters low and graduates into the high-touch motion rather than being turned away at the door.
• Tenant-per-product. Each product deploys as its own tenant via the existing tenancy-toggle mechanism (multi-tenant CORE plus single-tenant-via-toggles). No new multi-tenancy engineering - the toggle model already exists.

Why the ladder fits Devensoft

The ladder is not an aspirational PLG overlay; it is the commercial expression of an architectural capability Devensoft has already built. The tenancy-toggle model - multi-tenant CORE with single-tenant-via-toggles - is the mechanism that makes a per-product tenant economically plausible without per-product infrastructure. PMI, Synergy, and Legal depth at the enterprise tier is the top of the ladder - the destination that gives the lower rungs somewhere to graduate to. The twenty-plus onboarding / seat-license friction that constrains today's enterprise motion only applies at the top; the lower rungs sidestep it entirely, which is what makes the LMM segment reachable at all. And the fake-door validation in the next file is engineered specifically to test the bottom of the ladder where demand is least verified - so the ladder is built evidence-first from the rung that carries the most commercial risk, not assumed end-to-end.

The DealRoom Pricing Pressure

The most direct commercial threat to Devensoft's installed base is not a feature gap - it is DealRoom's pricing model. DealRoom ships a per-deal, unlimited-user subscription that structurally targets the seat-license friction Devensoft and Midaxo both carry (per competitor-profile--dealroom). Where Devensoft's enterprise economics couple revenue to seat count and recover onboarding cost through a multi-year structured-onboarding cycle (twenty-plus sessions per engagement, per the Modus synthesis), DealRoom decouples the buyer's cost from headcount entirely - the deal is the unit, not the user. For the Integration Management Office champion already aggrieved by per-seat friction and a heavy onboarding tax, that is a structurally easier yes.

The threat is sharper than pricing alone. DealRoom pairs the per-deal model with the first live MCP connectors in the competitive set - native integrations to Claude, ChatGPT, and Copilot that Devensoft does not yet ship - plus Looker BI with editable PowerPoint export, which goes directly at Devensoft's image-only PPT export weakness on the reporting dimension that the customer-voice evidence identifies as the #1 capability demand (per competitor-feature-parity-matrix). The combination - per-deal pricing, MCP live, BI that beats Devensoft's export - is what makes DealRoom the most dangerous structural rival in the set, not the closest functional analog (that remains Midaxo) (per competitor-profile--dealroom).

Devensoft's answer to that pressure is not to retrofit per-deal pricing onto the enterprise model. BTDR-0007 explicitly rejects retrofitting and tiered-hybrid blurring in favor of a split product line: Devensoft v24 keeps seat-license at the enterprise tier (where the high-touch motion and PMI/Synergy/Legal depth justify it), and the Diversify products adopt PLG/flat-tier pricing aligned to the CT Acquisitions $10k-$40k/year band (per BTDR-0007 - Decision 3, Alternative b). The reasoning: competing on DealRoom's per-deal terms on the enterprise tier would require dismantling a revenue model the installed base funds the transformation through; competing on per-deal on the Diversify tier, against a segment DealRoom does not serve, is where the same PLG motion actually wins. The pricing decision in this section operationalizes that split rather than re-litigating it.

The LMM Fake-Door Validation Plan

The Diversify side of the commercial model is gated by a demand signal that does not yet exist. The fake-door is the instrument that produces it - a non-existent product presented to the market through a landing page, sign-up flow, and waitlist, designed to measure whether the Lower-Middle-Market Searcher persona actually converts on the thesis the ladder rests on (per BTDR-0005). It is the repo's Fake-Door guiding philosophy applied narrowly to the one segment - LMM Searcher - where demand is genuinely unverified, having been narrowed from the broad-downmarket pre-build test that the CEO declined to run on grounds that the market was "self-evident" (per BTDR-0005 - scope narrowing, locked 2026-08-10).

The plan runs in three phases with two landing-page variants, against a budget of ~$3-6k - doubled from BTDR-0005's original $1.5-3k ceiling during the M11 decision review to allow broader persona coverage and champion-seeding incentives (per the M11 decision review; per lmm-notebooklm-execution-plan-validation). Critically, the campaign starts week 1, in parallel with the fundamentals work - it has zero engineering dependency (no product needs to exist for a landing-page test), which makes it both the cheapest pivot point if the demand signal invalidates and the earliest signal source if it confirms (per the M11 decision review).

What the fake-door answers is the Diversify product-breadth question carried as Decision 1 in M12: whether additional Diversify products beyond DevenDiligence (which is the only Diversify product shippable inside the nine-month horizon, per the M10 critical-path constraint) are worth building at all - Devenfy full-LMM standalone, an SDE Calculator micro-product, others. The fake-door targets exactly the bottom of the product ladder where that question lives. The four kill-conditions and four invalidation signals detailed in the next section (M12) are the criteria by which the signal the fake-door produces gets read; this section establishes only that the validation runs and what it costs. The campaign cost itself carries a residual open - PLG infrastructure cost is a real unknown flagged as an M10 dependency - which is named in the flag files below rather than concealed.

Split Product-Line Pricing (Decision - BTDR-0007)

This is a Decide directive. The decision is locked in BTDR-0007 (Accepted 2026-08-11). This file substantiates the decision against evidence; it does not re-litigate the resolution.

The decision: different products in the family carry different pricing models, deliberately and structurally. Devensoft v24 (enterprise) keeps a seat-license base, augmented by a premium tier - "Managed Success" / "EMA Enterprise+" - that unbundles the twenty-plus onboarding sessions, structured enablement, dedicated CSM, priority support, and QBR cadence the installed base actually consumes as a named value proposition above the base subscription, rather than absorbing their cost into the seat price (per BTDR-0007 - Decision 3, 6; per modus-enterprise-insights for the onboarding burden). The premium tier is simultaneously a retention lever for the renewal-contingent accounts the prior section flagged (the ~$200K relationship and the automotive account) and a differentiator against DealRoom's self-serve-only posture - the buyers who need high-touch are paying for it as a recognized product, not receiving it as a hidden cost Devensoft absorbs.

The Diversify products flip the model entirely. Pricing is free trial -> flat subscription tier, lifecycle-mapped to the buyer's journey (Search -> Diligence -> PMI), aligned to the CT Acquisitions $10k-$40k/year engagement band (per competitor-raw-capture-2026-08-06). Free-trial-to-paid was chosen over freemium deliberately: finite one-to-two-year search windows mean perpetual-free users may simply never convert before their deal closes, and one-and-done churn is precisely the failure mode the ladder's data-continuity and pricing-credit mechanisms exist to recover. Per-deal pricing was rejected because competing on DealRoom's terms on the Diversify tier fights the price war on the attacker's home ground; flat subscription tiers tied to lifecycle stage let the product family carry a coherent PLG motion instead (per BTDR-0007 - Alternative c).

Two structural asymmetries ride inside this split, both locked by prior BTDRs and named here so the pricing model is honest about its mechanics:

• AI monetization asymmetry. Enterprise customers operate under BTDR-0004's model-governance rule with BYOM mandated - the client supplies the LLM, and Devensoft carries zero inference cost. Under the tiered model (BTDR-0009), the Deven Connect consumption-based rate at the Grow tier is therefore margin-positive once rates land (BYOM = zero inference cost; consumption revenue flows at near-zero marginal cost), and free at the Retain tier is pure stickiness at zero marginal cost — noting that "free" here is a fair-use tier with low consumption limits (enough to enable some integrations but not full syncs, bulk downloads, or high-volume usage); customers needing higher limits upgrade to the paid consumption tier (per BTDR-0009 Am.). ⚠ Am.2: $75k MSRP superseded by tiered API pricing (BTDR-0009): R=free, G/D=consumption-based monthly. Diversify products absorb the hosted-LLM cost into the subscription, bounded by a fair-use cap and tier-gating that prevents the cost from running away (per BTDR-0007 - Decision 4; per BTDR-0009). This asymmetry is the commercial cost of the model-governance rule, accepted consciously rather than disguised.
• Deven Connect dual role + tiered pricing. The same API is itemized at the enterprise tier under the Grow consumption-based rate (where the buyer expects to see API pricing as a line item), bundled invisibly into the Diversify PLG tier (where the LMM buyer never sees a separate API charge), and granted free to current customers as a Retain anti-churn value-add (per BTDR-0007 - Decision 5; per BTDR-0009; per EMAKBAI-deven-connect - draft/provisional). ⚠ Am.2: $75k MSRP superseded by tiered API pricing (BTDR-0009): R=free, G/D=consumption-based monthly. One API investment serves all three objectives; what changes is whether it is itemized (Grow), bundled (Diversify), or granted (Retain). The Zapier GTM enabler (Phase 1 build, M14 G4) inherits the same tiered logic: a base tier (very low usage — a handful of personal, user-specific integrations) sits free/low-cap across all objectives, with tiered Zapier pricing to unlock higher limits available to Retain, Grow, and Diversify alike (per BTDR-0009 Am.).

The split rejects two alternatives explicitly: retrofitting per-deal onto EMA's enterprise economics (breaks the model that funds the transformation) and a tiered-hybrid that blurs products into a continuum (destroys the clarity the PLG motion depends on). The full alternatives analysis lives in BTDR-0007.

GTM Motion and Fake-Door Campaign (Decision - BTDR-0007 / BTDR-0005)

This is a Decide directive. GTM motion is locked in BTDR-0007 (Decision 6); fake-door campaign scope is locked in BTDR-0005. This file substantiates both against evidence.

Enterprise motion - keep what works, fix what's broken

The enterprise GTM motion stays relationship/demo-led. The customer-voice evidence is unambiguous that this motion works for the installed base: Dwight Fontilla's demo choreography - end-to-end M&A lifecycle walked through by a user who has lived the tool, narrative-anchored in pain points (never named as such), positioned as "deal strengthener / critical infrastructure to the deal" rather than another tool - is the proven conversion pattern (per demo-dwight-customer-voice). The plan does not disrupt that motion; it amplifies it with two additions:

• Advisor channel formalization. M&A advisors and consultants - Accenture (Dwight's firm), CT Acquisitions-class firms - reach buyers at the moment of need, during a live deal, not during a generic buying cycle. Formalizing advisor referrals as a partnership motion (rather than letting them remain ad-hoc) puts Devensoft in front of the Integration Management Office champion at the highest-intent point in the journey (per BTDR-0007 - Decision 6).
• Digital-presence fix, staged. The current digital presence is near-invisible: G2 carries a 4.6/5 score across only twelve reviews, Capterra does not list Devensoft, TrustRadius is empty (per competitor-raw-capture-2026-08-07 - section F). The fix is staged: champion-seed now - hand-pick two-to-three Dwight-class advocates for selective G2/TrustRadius reviews while building the content/SEO/community layer - and broad review cultivation deferred behind the fundamentals. The deferral is non-negotiable: customers will not advocate for a 1.5-out-of-5 product, and asking dissatisfied users for reviews compounds the credibility gap rather than closing it (per BTDR-0007 - Decision 6).
• DevenConnect distribution via Zapier (Phase 1 build, M14 G4). Zapier zaps = self-serve integration distribution for DevenConnect; advertised as the integration story ("all integrations enabled by Zapier") once available. This is the GTM enabler that makes the consumption-based API accessible to non-developers — the enterprise buyer's integration story on top of the Grow tier rate (per BTDR-0009).

Diversify motion - fake-door gate, then community-led PLG

The Diversify GTM motion is fake-door-gated: no GTM investment precedes the demand signal. The validation plan in the prior file runs in parallel starting week 1, and the four kill-conditions in M12 are the criteria by which the signal is read. If the signal confirms, the motion that follows is community-led PLG: the #ETA (Entrepreneurs Through Acquisition) community, Stanford GSB and HBS search-fund programs, AMAA and ACG associations - the low-CAC distribution arm specifically because LMM buyers aggregate in these communities in ways enterprise buyers do not (per BTDR-0007 - Decision 6). The institutional channel is the partnership motion that makes kill-condition #4 (distribution reach) testable rather than assumed.

Campaign design

The fake-door campaign itself runs three phases against two landing-page variants at a budget of ~$3-6k, doubled from BTDR-0005's original ceiling during the M11 decision review to broaden persona coverage and fund champion-seeding incentives (per the M11 decision review). Additional budget is allocated for customer-appreciation and targeted feedback sessions with the installed base - the cheapest source of qualitative signal available, and the seed from which the champion-seed tactic above draws its advocates. The campaign design is intentionally cheap because the question it answers - whether the bottom of the product ladder carries real demand - is the binary gate on which the entire Diversify commercial story either proceeds or collapses.

Flag - Pricing Data: Substantially Resolved (Realized Side Still Open)Flag

This is a Flag directive. Substantially resolved by the 2026-08-12 pricing baseline; the residual gap is surfaced for leadership and M13's baseline census rather than smoothed over.

The section once rested on a single price point - the Deven Connect $75k MSRP - with no published rate card. The picture is now materially better. Real list/guidance pricing exists: Pipeline Only at $135/seat/mo, Pipeline+PMI at $165/seat/mo, and a one-time $5k–$20k onboarding/implementation fee that scales by org size and complexity - and these are explicitly "minimums," since every deal is individually negotiated on both structure and price with no public rate card (per commercial-pricing-baseline §1 - Diego guidance, provisional). Realized negotiated ACVs now exist too: the median closed-won renewal runs ~$83–88K/yr (Stryker $200K at the top, ~2.3× the median), though "Unlimited"/"Premium" deal structures defeat naive per-seat division (same source §2 - internal projection, provisional).

Two caveats keep this a flag rather than a clean close. First, the $75k Deven Connect MSRP — downgraded to theoretical by Am.1 — is now superseded by the tiered model (Am.2 / BTDR-0009): free for Retain, consumption-based monthly for Grow and (if needed) Diversify, replacing the dead anchor with a live commercial strategy; the specific consumption-based rate structure is TBD at the M11 resolution (per commercial-pricing-baseline §3; per BTDR-0009; see 02_deven-connect-commercial-layer). ⚠ Am.2: $75k MSRP superseded by tiered API pricing (BTDR-0009): R=free, G/D=consumption-based monthly. Second, discount-adjusted per-seat realization is still not computable - "Unlimited"/"Premium" structures and individually-negotiated deals mean clean enterprise $/seat needs finance-system data. The CT Acquisitions $10k–$40k/year band that benchmarks the Diversify flat-tier remains third-party-captured advisor-reported pricing (CT Acquisitions is an advisor, not a software vendor) (per competitor-raw-capture-2026-08-06), and competitor pricing across the feature-parity matrix is marketing-page-reported - vendor revenue and ARR are private (per competitor-feature-parity-matrix).

The implication for this section is that the pricing shape is right and now grounded in real list + realized figures - split product line, dual-role Deven Connect under a tiered model (R=free, G/D=consumption-based monthly; specific rates TBD), lifecycle-mapped Diversify tiers - but the specific per-seat realization and the consumption rate structure both remain leadership-owned refinements, the former gated on the M13 baseline census producing discount-adjusted enterprise $/seat and Diversify conversion data, the latter on the M11 resolution.

Flag - Mid-Market Unit Economics: Narrowed (CAC/LTV Still Absent)Flag

This is a Flag directive. Narrowed by the 2026-08-12 pricing baseline on the enterprise side; the LMM-side gap is named as one of the three leadership decisions surfaced in M12 (Decision 1) and depends on the M13 baseline census + the fake-door signal.

The Diversify commercial model rests on unit economics Devensoft does not currently possess. There is no measured customer acquisition cost (CAC) for the LMM segment - the segment has never been sold to, so there is no historical funnel data; the institutional-channel and community-led PLG motions in this section are designed from analogous segments, not from Devensoft's own experience. There is no lifetime value (LTV) measurement, and therefore no LTV/CAC ratio on which to judge the Diversify bet's economics; the data-continuity and pricing-credit ladder mechanisms are designed precisely to convert one-and-done LMM churn into lifecycle LTV, but whether they will is hypothesis, not measurement. The CT Acquisitions $10k-$40k/year diligence-pricing band is the closest analog and is explicitly third-party-reported, not Devensoft-verified (per competitor-raw-capture-2026-08-06; per enterprise-market-opportunity - section 6, pricing as strategic lever).

On the enterprise side, the gap is now narrower than it was. The HubSpot CRM hygiene pass (resolved 2026-08-11) established clean denominators - twenty-five active companies, eleven active deals, thirty active tickets, sixty-one active contacts - against which seat-license revenue and renewal-cycle figures can in principle be computed (per hubspot-hygiene-pass-results). The 2026-08-12 pricing baseline adds the missing dollar dimension: list minimums of $135/$165 per seat/mo + $5k–$20k onboarding and realized negotiated ACVs (median ~$83–88K/yr; Stryker $200K top) now enable a first-cut enterprise unit-economics model (per commercial-pricing-baseline §1–§2 - guidance/provisional). What is still not computable is clean per-seat realization ("Unlimited"/"Premium" structures defeat naive division), discount-adjusted ARR, and CAC-by-channel - that work belongs to the M13 baseline census.

The hybrid approach this section takes - directional ranges from benchmarks plus explicit named gaps for leadership - is therefore not a deferral but a scoping act: BTDR-0007 locks the commercial architecture; the unit-economics gaps are surfaced for leadership as Decision 1 in M12, with the M13 census and the fake-door signal as the two inputs that will close them (per BTDR-0007 - Decision 8, follow-ups).

Flag - Campaign and PLG Infrastructure Cost Is OpenFlag

This is a Flag directive. Fog item named, not resolved. The PLG infrastructure cost is flagged as an M10 dependency; the fake-door budget is doubled but acknowledged as a floor.

Two campaign-cost items remain genuinely open. First, the fake-door budget was doubled from BTDR-0005's original $1.5-$3k ceiling to ~$3-6k during the M11 decision review, to broaden persona coverage and fund champion-seeding incentives alongside the landing-page variants (per the M11 decision review; per BTDR-0007 - Decision 6). The doubling is acknowledged as a floor on what a credible test costs, not a validated total; the campaign could require more if the early signal is ambiguous and a follow-on round is warranted. Additional budget is allocated for customer-appreciation and targeted feedback sessions with the installed base - the cheapest qualitative signal source available - but the magnitude of that allocation is a leadership appetite call.

Second, the PLG infrastructure cost - the hosted-LLM inference cost the Diversify products absorb under BTDR-0007's AI monetization asymmetry - is a real unknown. The fair-use cap and tier-gating mechanisms are designed to bound it, but no Devensoft build actuals exist against which to size it: the LMM segment has never been served, and the hosted-LLM cost profile under a PLG motion is by definition a new cost line (per BTDR-0007 - Decision 4). This cost is flagged as an M10 dependency: the enabler-stack work (E2 tooling/process investments in M10's enabler model) needs to surface it before any Diversify product is committed to a flat-tier price that bakes it in. The plan does not invent a hosted-cost estimate; it names the gap.

M12 — Risks, Assumptions & Decisions: Intro Summary

The plan's downside is not a list of independent risks but a fragile base that shapes how a two-front hedge is sequenced. The 2026-08-12 financial baseline showed a Retain base too small, too concentrated, and too eroding to fund a balanced two-front hedge: 82.7% closed-renewal retention, roughly 22 active companies (Stryker's $200K alone is ~2.3× the median realized ACV), $700.8K closed-won against a $1.92M projection, and Deven Connect's $75k MSRP resolved to $0 realized revenue (per fy26-financial-baseline, provisional — internal projection, not recognized ARR; per BTDR-0007 Am.1). A base this concentrated means a single lost renewal materially changes the cash position that funds everything else — which is exactly why the posture stays BTDR-0004's "Hedge, sequenced" (a two-front hedge supporting both Retain and Diversify, time-boxed and sequenced), and why its G4 rule sequences fundamentals first: capacity is smaller than ambition, so the Retain/fundamentals direction leads (per BTDR-0004 · G4; capacity is a dynamic leadership-headcount exercise modeled in recommendation-matrix.html; the 90/10 attention firewall coexists on a separate axis). The fragile base makes the Retain leg of the hedge the sharper near-term risk, not a mandate that Retain wins every capacity contest. On that sequencing the risk surface stacks: a capacity envelope sized at ~105.3 dev-wk total, with the Tier-1 net-new draw branch-dependent on the LMM Vehicle Choice — a DevenLite choice commits ~47-53 dev-wk of the ~61.5 headroom: it fits, barely (per recommendation-matrix.html — 3-tier effective-FTE model, Diego+Roman preset: 5.5 FTE, 39w, 30% drag, 0.4 pizza drag >3 FTE, 1.6× agentic, 2h/8h overhead; a DevenDiligence choice draws materially less, ~18–26 incl. follow-ons; no choice = Tier-2 signal work only), of which ~43.8 dev-wk holds the Retain/fundamentals floor and ~61.5 dev-wk remains as Grow/Diversify headroom; an in-horizon AI win narrowed to a DevenMCP sliver (~4–6 dev-wk, HIGH confidence) layered above the existing DevenConnect surface — the full runtime core (~22–34w full / ~12–19w MVP) is deferred post-horizon on the F1 serial gate, not headcount-fixable, the north-star for the next battle rather than deliberately-not-build (per the AI-interface sizing brief; F1 serial gate per the M10 decision review); a Diversify track whose Tier-1 slot is the deferred LMM Vehicle Choice - DevenLite (~32w minimum guess, LOW confidence, coupling-dominated by DealSetupProcessing's 24/69 dependencies) and DevenDiligence (~3–5w, LOW confidence) the Tier-1-live candidates, no vehicle committed, with the DevenDiligence + DevenFiles follow-ons as the DevenLite branch's viability condition (per BTDR-0010 Am.2); and a newly split commercial model whose unit economics are not yet Devensoft-verified (per BTDR-0004 G4; per the M10 decision reviewramp + critical path; per BTDR-0007 split line + unit-economics gap; per lmm-thesis DST-LMM-THESIS, NotebookLM-derived - provisional). Leadership owes three concrete decisions before capacity commits further: the Tier-1 vehicle itself — the LMM Vehicle Choice — plus which Diversify products carry the bet beyond it (the D3 instrument's verdict covers C3/C4/C6), whether reporting-AI ships early in the horizon or lands later (the F7 floor now costed at ~5.5 dev-weeks plus ~4-6 dev-weeks for the PoC layer), and how the capacity-vs-ambition conflict resolves - cut scope, extend the timeline, or add capacity. None of these are pre-decided by this plan; each is surfaced with its decision criteria so Eric and Thom can make the call.

This section holds the same bar the plan applies to its own current-state read (M05): name the gaps plainly, don't launder them into softer language for leadership comfort. Where evidence is thin — the LMM kill-conditions' quantification, the fourth invalidation signal, the F1 auth-migration re-size, commercial unit economics — it is flagged as fog rather than filled in with invented numbers, per the "decide, don't list" and "risks named honestly" guiding principles.

Picking up from the commercial model and go-to-market motion (§11), this section turns from where the return is to what could go wrong, and what must be decided before we commit further capacity to it — the risks and decisions the bet hinges on.

LMM Kill-Conditions and Invalidation Signals

The Diversify (LMM) bet is explicitly conditional, not a foregone conclusion. The underlying research holds that a downmarket LMM product beats "defend and deepen enterprise" only if all four of the following are true at once: the Enterprise Trap is immutable (the existing platform genuinely cannot be rebranded or simplified for LMM without alienating the enterprise base); the displacement gap is a volume play (the $5M–$50M retiring-founder transaction wave outsizes the enterprise share we could instead defend or grow); the infrastructure is a force multiplier (the software is the only, or best, way searchers scale into serial acquirers); and CAC is genuinely lower via the #ETA / Stanford GSB / "Search Twitter" community channel (per lmm-thesis · DST-LMM-THESIS §6, NotebookLM-derived — provisional; per the LMM divestiture decision review, Q3, NotebookLM-derived — provisional). The thesis is Devensoft-originated research (Diego-authored, 2026-04/05), but none of the four conditions has market validation — no fake-door signal, no customer data — and the four rest on that single synthesis rather than independent corroboration.

The same research names four arguments against the bet, each with a mechanism that would sink it: over-engineering (searchers are price-sensitive enough that every software dollar erodes their 25–30% equity stake, making Excel/Pipedrive a durable "good enough" baseline); one-and-done churn (the search phase is a finite 1–2 year window, capping lifetime value unless a searcher converts into a serial acquirer); cultural alienation (Main Street sellers read financial-services polish as a "vulture capitalist" signal, breaking the succession-solution trust the pitch depends on); and marketplace encroachment (sourcing value depends on BizBuySell/SearchFunder/Acquire.com not building native VDR or modeling features).

Honesty flag: the source names four risks but defines only three measurable fake-door signals in total, and only two of the four risks map directly to a named signal — cultural alienation → low conversion on "Institutional Readiness" messaging; marketplace encroachment → resistance to "Sourcing Velocity" tooling versus manual Pipedrive. A third signal — failure of "Valuation Arbitrage" (SDE→GAAP) messaging — is defined in the source but not attributed to any of the four named risks. Over-engineering and one-and-done churn have no defined trigger at all. This gap is carried forward rather than papered over — it's the plan's own evidence that the "4 kill-conditions / 4 invalidation signals" framing is aspirationally complete, not yet operationally complete.

The Defend-Enterprise-vs-Attack-LMM Risk Surface

The pivotal bet (BTDR-0004) resolved the Retain-vs-Diversify tension as a hedge — support both, time-boxed and sequenced — rather than a binary choice. The 2026-08-12 financial baseline did not contradict any number BTDR-0004 cited (it cited none), but it exposed the assumption the hedge rested on: that the Retain base is large enough and stable enough to fund a balanced two-front effort. It is not. Closed-renewal retention ran 82.7%; closed-won landed at $700.8K against a $1.92M projection; the active base is roughly 22 companies, concentrated (Stryker $200K alone is ~2.3× the median realized ACV); and Deven Connect's $75k MSRP resolved to $0 realized revenue (per fy26-financial-baseline, provisional — internal projection, not recognized ARR; per BTDR-0007 Am.1). A base this small, this concentrated, and eroding at this rate is exactly why the posture stays BTDR-0004's "Hedge, sequenced" rather than collapsing to a single front — and why its G4 rule sequences fundamentals first: capacity is smaller than ambition, so the Retain/fundamentals direction leads (per BTDR-0004 · G4; capacity is a dynamic leadership-headcount exercise modeled live in recommendation-matrix.html). The hedge stays two-front — it does not mandate a hard "Retain wins every capacity contest" rule — but the fragile base is what makes the Retain leg the sharper near-term risk: a single lost renewal materially changes the cash position that funds everything else. The 90/10 attention firewall coexists on a separate axis (attention/energy, not capacity). The risk surface that follows is wider than any single capacity question, and the base the Retain leg protects is itself the first risk.

Retain side — the fragile base the hedge's Retain leg exists to protect. The base being defended is itself eroding: every lost renewal from a ~22-company concentrated base shifts the cash position, and the competitive surface offers no relief — Devensoft ships zero diligence-AI while every profiled rival (Midaxo, DealRoom, Datasite, Intralinks) ships it, and the #1 demand signal — reporting — is directly out-featured by Midaxo and DealRoom on real-time dashboards, scheduled delivery, and external BI, where Devensoft depends on SSRS code changes and exports as image-only PPT (per competitor-feature-parity-matrix · DIST-COMPETITOR-FEATURE-PARITY-MATRIX, D2/D4). The locked posture (BTDR-0004 · G1) accepts this exposure deliberately: reporting-AI is the floor tier, conditional on a time-boxed proof-of-concept (now costed at ~5.5 dev-wk for the F7 foundation plus ~4–6 dev-wk for the PoC layer), and diligence-AI stays deliberately deferred as a losing catch-up feature. The risk is that "deliberately deferred" becomes a widening competitive gap the longer the horizon runs — and under the fundamentals-first sequencing (BTDR-0004 · G4) the irony is sharpened: the base the Retain leg protects is the base that is eroding, which makes the deferral costlier, not safer.

AI-posture risk — narrower in-horizon than the full vision. The G5 split of the former AI win condition reframes the AI risk surface. The in-horizon AI win is now the DevenMCP sliver (G5a, ~4–6 dev-wk, HIGH confidence) — an AI/MCP layer above the existing DevenConnect API surface, no F1 dependency, prioritized above any Diversify new-build. The full runtime core (G5b, ~22–34w full / ~12–19w MVP) is post-horizon: F1-serial-gated (~mo 6–7), not headcount-fixable, deferred to Beyond the Horizon as the north-star win condition for the next battle — not in the deliberately-not-build category. The risk this split creates is named honestly: the differentiated AI bet is narrower in-horizon than the original Win-tier vision promised. What ships inside the window is a sliver above an existing surface, not a re-architected AI platform; the full AI differentiation is deferred to the next battle. Combined with the BYOM routing (4 dev-wk, M confidence, the Satisfy tier) and the conditional reporting-AI PoC (~4–6 dev-wk, L confidence), the in-horizon AI scope totals ~12–16 dev-wk — real AI presence, but bounded, not the full vision (per the AI-interface sizing brief; F1 serial gate per the M10 decision review; per fundamentals-sizing-brief).

Diversify side — a deferred Tier-1 choice with the capacity consequences named, carried openly. The capacity picture is no longer "two developers." M10 locked a 3-stage ramp — Ramp-in ~2.0 FTE (mo 1–2, Diego + Roman with −30% transition drag) → Build ~3.5 FTE (mo 3–6) → Accelerate ~4.5 FTE (mo 7–9) — against which the fundamentals floor alone (~32.5–49.5 dev-weeks across F1–F7, per fundamentals-sizing-brief) consumes most of the horizon before any Diversify, AI, or GTM track starts. The critical path is F6 → F1 → T2 → T3 → T7: F6 remaining (5–8w) plus F1 — the auth-provider migration to a managed identity provider (Clerk/Auth0, buy-not-build; BTDR-0008, itself needing re-sizing) — is 14–21+ dev-weeks of serial dependency, roughly 4–7 months of the nine at realistic allocation (per the M10 decision reviewQ5). Against that path, the Tier-1 slot is the LMM Vehicle Choice (BTDR-0010 Am.2): a deferred leadership decision among the LMM vehicles — DevenLite (front-scoped, late-window if chosen, its DealSetupProcessing coupling (24/69 dependencies) acknowledged-and-carried rather than hidden as the ~32w minimum-guess's dominant term) and DevenDiligence (~3–5w), the Tier-1-live candidates; no vehicle committed, none recommended. The risk surface this creates is named honestly rather than smoothed over, and it is branch-dependent: a DevenLite choice commits ~47–53 dev-wk against ~61.5 of headroom — it fits, barely — with the coupling-dominated sizing possibly optimistic and DevenLite not entirely viable without the DevenDiligence + DevenFiles follow-ons; a DevenDiligence choice draws materially less (~18–26 incl. follow-ons); no vehicle chosen leaves Tier-2 signal work only. Under compression (G4) on any branch, the chosen vehicle's scope gives ground before the Retain floor. The remaining breadth question (which products beyond the Tier-1 choice) is explicitly not pre-decided by this plan — it is framed as a leadership appetite call because the underlying demand is itself unvalidated (see the kill-conditions in §2).

Commercial side — a second model to manage, and an unverified one. Layered on the posture is a commercial-model risk surface locked by M11 (BTDR-0007) that this section did not carry when first drafted. The split product line means two commercial models to operate — Devensoft v24 on seat-license (+ a premium "Managed Success" tier unbundling the 20+ onboarding sessions) versus Diversify products on free-trial-to-flat-tier PLG. The AI monetization asymmetry is the commercial cost of BTDR-0004's model-governance rule: enterprise BYOM is zero inference cost to Devensoft (pure margin on the Deven Connect add-on if its vault-draft $75k MSRP ever landed — but that price is theoretical/$0 realized, BTDR-0007 Am.1), while Diversify's hosted LLM is baked into the subscription under a fair-use cap + tier-gating — Devensoft absorbs that inference cost, and the cap is an enforcement surface that, if miscalibrated, either erodes margin or triggers churn. Deven Connect's dual role (itemized enterprise add-on [vault-draft $75k, theoretical] / bundled-invisible Diversify infrastructure) is a cannibalization risk if the two faces are not governed deliberately. And the unit economics underneath all of it — enterprise $/seat, CAC in dollars, LTV/CAC — are directional ranges from benchmarks only, unvalidated against Devensoft's own books pending the M13 baseline census (per BTDR-0007 · Decision 8).

Capacity — sized, with the Tier-1 draw branch-dependent and the residual risk on the critical path. The capacity side is now sized: the fundamentals floor runs ~32.5–49.5 dev-wk across F1–F7 (per fundamentals-sizing-brief) against a 3-stage ramp (Ramp-in ~2.0 FTE → Build ~3.5 FTE → Accelerate ~4.5 FTE), and the envelope resolves to ~105.3 dev-wk total with ~61.5 dev-wk of headroom after the ~43.8 dev-wk floor (per recommendation-matrix.html — 3-tier effective-FTE model, Diego+Roman preset: 5.5 FTE, 39w, 30% drag, 0.4 pizza drag >3 FTE, 1.6× agentic, 2h/8h overhead). How much of that headroom the Diversify Tier-1 slot draws is branch-dependent on the LMM Vehicle Choice (BTDR-0010 Am.2): a DevenLite choice commits ~47–53 dev-wk of in-horizon net-new (AI scope + DevenLite + the DevenDiligence follow-on) — it fits that headroom barely, the ~9–15 dev-wk residual being what remains for onboarding-drag, pizza-team coordination, and estimation overrun, with the coupling-dominated DevenLite minimum-guess (~32w, LOW confidence) as the dominant uncertainty; a DevenDiligence choice draws materially less (~18–26 incl. follow-ons); no vehicle chosen leaves Tier-2 signal work only. The execution-timeline risk that does remain is the critical path F6 → F1 → T2 → T3 → T7, which alone consumes roughly 4–7 of the nine months at realistic allocation — a sequencing risk layered on top of a branch-dependent margin, not a comfortable headroom — compounded by the non-floor scope (DevenConnect build, extraction, AI tiers, GTM motion) being still unsized and DevenFiles being unsized (a viability caveat carried openly on the DevenLite branch). If capacity ever runs scarce, the fundamentals-first sequencing (BTDR-0004 · G4) governs the direction the scope gives ground — the chosen vehicle first, never the Retain floor — not a Retain-wins-everything rule.

The shared bench. All sides draw against the same ramp, which is the same capacity already running roughly 136% over estimate on in-flight IoC/DI modernization work (LGCYEMA-511) before any new build is added (per technical-debt-register · DIST-TECH-DEBT) — now readable as estimation-risk evidence rather than the closest capacity data point (the floor is sized; F1 is not). Running Retain-hardening, Diversify-exploration, a narrowed AI sliver, and a split commercial model in parallel — inside a horizon whose critical path alone consumes most of it, under a fundamentals-first sequencing that leads with the base — is the risk this section exists to name plainly, not soften.

The Decision List Leadership OwesDecide

Three decisions carry from the pivotal-bet decision review into this plan as surfaced leadership calls (BTDR-0004) — named individually, each with its own owner and forcing question, not folded into a single vague "risks exist" statement. The 2026-08-11 decision wave (the M10/M11 decision reviews + the reporting-floor sizing) sharpened all three without closing any of them; the 2026-08-12 sharpening wave (the G5 split, the Diversify narrowing, and the capacity envelope now sized at 105.3 dev-wk / ~61.5 headroom) sharpened them further — still without closing any.

Decision 1 — the LMM Vehicle Choice, plus Diversify breadth beyond it. The DevenLite-vs-DevenDiligence fork is resolved (the M11 decision review): DevenLite (Targets+Tasking) and DevenDiligence (DD carve-out) are separate products, not alternatives — both frontend-only over Deven Connect. The Tier-1 build question is now the LMM Vehicle Choice (BTDR-0010 Am.2, 2026-08-18): a deferred leadership decision among the LMM vehicles — the plan commits to no vehicle and recommends none (the pitch's "DevenDiligence recommended" framing stays pitch-side). The Tier-1-live candidates carry their facts openly: DevenLite (C2) — ~32 dev-wk minimum-guess at LOW confidence, coupling-dominated (DealSetupProcessing 24/69), front-scoped, late-window if chosen, with DevenDiligence + DevenFiles as that branch's viability follow-ons; DevenDiligence (C1) — ~3–5 dev-wk at LOW confidence (Clarity gate retired). The choice is deliberately not pre-decided by the architect: it is surfaced to Eric and Thom with the branch-dependent capacity consequences named (DevenLite ≈ ~47–53 of ~61.5 — "fits, barely"; DevenDiligence ≈ ~18–26 incl. follow-ons; no choice = Tier-2 signal work only). The breadth question beyond it remains open in its prior form: whether the Diversify bet extends past the chosen vehicle at all — the D3 demand-signal instrument's verdict covers C3/C4/C6 only (Valuation Engine, SDE Quick-Val, Sourcing; DevenFiles/C5 rides the choice), and Devenfy (NB5) waits on the same signal — an appetite-driven call with per-product kill-conditions and the instrument's verdict (method itself a leadership-decides item, fake-door lineage per BTDR-0007) feeding the answer rather than a guess preceding it (per BTDR-0004 · Decision 2, LOCKED 2026-08-06; G2 terminology amended in the product-taxonomy sweep; choice encoding per BTDR-0010 Am.2).

Decision 2 — reporting-AI: must-ship or stretch. Reporting-AI sits at the floor of the layered AI posture (the table-stakes "yes we have AI" tier) and is conditional on a time-boxed proof-of-concept (semantic-model-first, MCP-second). The reporting-data foundation underneath it is now sized at ~5.5 dev-weeks as the F7 cluster (analog LGCYEMA-517 × 1.5; confidence Medium — per fundamentals-sizing-brief), and the PoC layer above it carries ~4–6 dev-wk at LOW confidence. The broader AI-posture picture has sharpened with the G5 split: the in-horizon differentiated AI win is now the DevenMCP sliver (G5a, ~4–6 dev-wk, HIGH confidence) above the existing DevenConnect surface, while the full runtime core (G5b, ~22–34w full / ~12–19w MVP) is deferred post-horizon on the F1 serial gate — so the reporting-AI ship-or-stretch call is now situated within a deliberately bounded in-horizon AI scope (a sliver + BYOM + conditional reporting-PoC, ~12–16 dev-wk of the branch-dependent ~47–53 net-new total), not the full Win-tier vision. Whether reporting-AI is a hard commitment early in the ~nine-month horizon or an explicit stretch goal that may land later is not yet settled — M10 places the reporting-AI floor PoC in Phase 1 as a soft gate (ship-or-kill decided mid-horizon, not day-one) — and it is entangled with Decision 3, since the capacity that would build it is the same capacity the critical path already consumes (per BTDR-0004 · Decision 1 + Decision 4, LOCKED 2026-08-06; the M10 decision reviewPhase-1 soft gate; G5 split per the sizing brief + the M10 decision reviewF1 gate).

Decision 3 — the capacity-vs-ambition resolution. Eric's ~nine-month delivery horizon (CEO 2026-08-07, "win the next battle"; earlier "2-month mandate" retired per BTDR-0004 G4 amendment) sets the planning frame. The realistic delivery capacity is now sized — and, with the org-preview commitments, stretched to a thin margin: the envelope resolves to ~105.3 dev-wk total (per recommendation-matrix.html — 3-tier effective-FTE model, Diego+Roman preset: 5.5 FTE, 39w, 30% drag, 0.4 pizza drag >3 FTE, 1.6× agentic, 2h/8h overhead), of which ~43.8 dev-wk holds the Retain/fundamentals floor and ~61.5 dev-wk remains as Grow/Diversify headroom. A 3-stage ramp (Ramp-in ~2.0 FTE mo 1–2 → Build ~3.5 FTE mo 3–6 → Accelerate ~4.5 FTE mo 7–9) against which the critical path F6 → F1 → T2 → T3 → T7 alone consumes roughly 4–7 of the nine months at realistic allocation, before any Diversify/AI/GTM track meaningfully starts (per the M10 decision review; F1 = auth-provider migration, BTDR-0008, itself pending re-sizing). The in-horizon net-new draw is now branch-dependent on the LMM Vehicle Choice (BTDR-0010 Am.2): a DevenLite choice commits ~47–53 dev-wk (AI scope + DevenLite + the DevenDiligence follow-on) against that ~61.5 dev-wk headroom — it fits, barely: the ~9–15 dev-wk residual must still absorb onboarding-drag, pizza-team coordination, and estimation overrun, and the coupling-dominated DevenLite minimum-guess (~32w, LOW confidence, no re-size) is the dominant uncertainty. What keeps the conflict live is the critical path as a sequencing risk (4–7 of the 9 months consumed before the non-floor scope starts) layered on that thin margin, plus the non-floor scope being unsized (DevenConnect build, extraction, AI tiers, GTM motion) and DevenFiles being unsized. If capacity ever runs scarce, the fundamentals-first sequencing (BTDR-0004 · G4) governs the direction the scope gives ground — the chosen vehicle first, never the floor. Leadership must still resolve this explicitly: cut scope to what the ramp can ship, extend the timeline, or add capacity — the plan surfaces the conflict rather than quietly assuming it resolves itself (per BTDR-0004 · Decision 4, LOCKED 2026-08-06).

None of these three leadership calls is itself locked — that is the point of surfacing them here. What BTDR-0004 locks is the diagnosis each rests on: the AI-posture layering behind Decision 2 (G1, now sharpened by the G5a/G5b split) and the capacity reality behind Decision 3 (G4, now sharpened by the sized 105.3 dev-wk / ~61.5 headroom envelope, the branch-dependent Tier-1 draw (~47–53 on a DevenLite choice), and the fundamentals-first sequencing) are settled facts, but the leadership choices they force — choose-the-vehicle, ship-or-stretch, cut/extend/add-capacity — remain open. Decision 1 (the LMM Vehicle Choice + breadth beyond it) is the most explicitly unresolved of the three: the original vehicle fork is resolved (the M11 decision review), the Tier-1 slot is now a deferred leadership choice rather than a committed build (BTDR-0010 Am.2 — the plan names the options and their sizing facts; Eric and Thom decide), and which additional products (C3/C4/C6 via the D3 instrument; Devenfy) to build remains open.

LMM Kill-Conditions as Go/No-Go Gates

Naming the four kill-conditions (§2 above) is not enough on its own — the decision this section forces is to adopt them formally as go/no-go gates the Diversify bet is checked against, rather than background assumptions nobody revisits. The gate mechanism proposed is the Fake-Door validation already committed under the plan's fake-door scope decision: run the landing-page tests against the four conditions and the three currently-defined invalidation signals — low conversion on "Institutional Readiness" messaging, resistance to "Sourcing Velocity" tooling over manual Pipedrive, and failure of "Valuation Arbitrage" (SDE→GAAP) messaging — before further build capacity is committed to any Diversify product (per lmm-thesis · DST-LMM-THESIS §7–8, NotebookLM-derived — provisional).

The decision this plan asks leadership to lock is threefold: first, that these four conditions and their signals are the checkpoint (not a softer, informal "we'll know it when we see it" read); second, who owns running and reading the Fake-Door test and on what cadence; and third, what "no-go" actually triggers — whether a failed gate pauses the Diversify product-breadth decision (Decision 1, §4), redirects the Diversify investment back toward Retain/Grow, or simply delays commitment without reallocating capacity. None of this is locked yet — it is a pending decision, not yet carried into a BTDR of its own, and is named here as the gap it is rather than assumed resolved by BTDR-0004's hedge framing.

Flag: Risk Quantification Is Mostly Absent

Most of the risks named in this section still do not carry a number — but the capacity side, which was the largest gap when this section was first drafted, has since been sized. The fundamentals floor is now costed: the seven fundamentals clusters (F1–F7) sum to ~32.5–49.5 dev-weeks, with F7 (reporting-data foundation, the "minimal self-serve report creation" item added to the contractual floor) sized at ~5.5 dev-weeks from the LGCYEMA-517 analog × the standard 1.5 calibration (confidence Medium — near-perfect analog shape, but zero build actuals on the Power BI surface; per fundamentals-sizing-brief · F7). Against the 3-stage capacity ramp locked by M10 — Ramp-in ~2.0 FTE (mo 1–2) → Build ~3.5 FTE (mo 3–6) → Accelerate ~4.5 FTE (mo 7–9) — the envelope resolves to ~105.3 dev-wk total (per recommendation-matrix.html — 3-tier effective-FTE model, Diego+Roman preset: 5.5 FTE, 39w, 30% drag, 0.4 pizza drag >3 FTE, 1.6× agentic, 2h/8h overhead), of which ~43.8 dev-wk holds the floor and ~61.5 dev-wk remains as Grow/Diversify headroom. The in-horizon net-new draw is now branch-dependent on the LMM Vehicle Choice (BTDR-0010 Am.2): a DevenLite choice commits ~47–53 dev-wk (AI scope + DevenLite + the DevenDiligence follow-on) — it fits the ~61.5 dev-wk headroom barely, with the coupling-dominated DevenLite ~32w minimum-guess (LOW confidence, no re-size performed) the largest single unquantified swing in the plan (per the M10 decision review). The capacity-vs-ambition conflict (§3–4) is therefore no longer un-sized; what remains un-sized is the non-floor scope (DevenConnect build, extraction, AI tiers, GTM motion), DevenFiles (the unsized third family member), and the sliding cost of the 90/10 firewall's legacy attention drift.

Where quantification is still absent — and this Flag exists to name, not fill:

• LMM kill-conditions and invalidation signals (§2) remain qualitative pass/fail reads on a single NotebookLM synthesis, not modeled probabilities or dollar-at-risk figures.
• Dollar-exposure on the timing clock — the FY26 financial baseline partially narrows this: 82.7% closed-renewal retention and $128K attrition across 7 accounts now provide a directional baseline, and the renewal registry carries real dates + dollar amounts for the next 12 months (per fy26-financial-baseline, provisional — internal projection, not recognized ARR). The hard cash/runway figure remains an open fog (G-M02-RUNWAY; the workbook is a revenue projection, not a cash position).
• Commercial unit economics — enterprise $/seat, CAC in dollars, LTV/CAC — are directional ranges from benchmarks only; they wait on the M13 baseline census + CRM hygiene pass to become Devensoft-verified (per BTDR-0007 · Decision 8).
• F1 (auth provider migration) itself needs re-sizing — the sizing brief's 9/16 controller-endpoint estimate undercounts the actual surface; F1 is the swing factor on the critical path and its re-sizing is a live dispatch item, not a settled number (per the M10 decision review, F1 dispatch).

Two estimation-risk data points survive the sizing work as cautionary evidence rather than models: a single modernization ticket (LGCYEMA-511) running roughly 136% over its original estimate, and the decision-support calculators' own "fundamentals" placeholder, which was flagged as likely undersizing the #1 demand signal before F7 landed (the prototype has since graduated into the architect-facing calculators with a corrected factor model — still a decision-support tool, not a validated sizer) (per technical-debt-register · DIST-TECH-DEBT).

This gap is named rather than filled with invented numbers. Where the plan cannot yet quantify a risk, it says so plainly; the decision list in §4–5 is written to force the remaining sizing work (the Fake-Door signal, the F1 re-size, the M13 unit-economics census, a capacity-costed scope cut) rather than to substitute for it.

Outcomes & Success Metrics - Intro Summary

Success in this plan is defined by a two-tier causal chain, not a flat list of KPIs. Tier 1 is lagging and growth-first — net-new ARR and Net Revenue Retention, the metrics the CEO named as the scoreboard. Tier 2 is leading and operational — reporting-cycle compression and maturity-scorecard progression — the measurable, evidence-grounded signals whose movement in months 1 through 9 predicts whether the Tier 1 numbers will land in months 9 through 18. The causal logic is explicit: if the fundamentals fixes compress the reporting cycle, if the maturity scorecard moves from 1.5/5 toward a measurable target, if the instrumentation tier comes online — then the lagging growth metrics have a predictive basis. Without Tier 2 movement, Tier 1 is a wish. The instrumentation gap — the fact that most of these metrics lack baselines today — is not confessed as a weakness; it is named as a workstream: a CRM hygiene pass (completed), a one-time baseline census, and a feedback-portal capability that converts the absence into actionable deliverables.

The credibility of this metrics section comes from what it refuses to do. It does not set aspirational targets detached from a pivotal bet — target numbers are explicitly deferred to the recommendations section with a dependency link to the bet decision, because a defend-first posture and a Diversify-first posture produce different targets for the same metric categories. It does not cite the vault's unverified marketing claims (48% time savings, 56% adoption, 7x reporting speed) — those are named once in a "do not cite" footnote and excluded from the metric set until verified. It does not claim a baseline it does not have — five of nine instrumentation-tier metrics now have directional baselines (the CRM hygiene pass gave clean denominators: 25 active companies, 11 active deals, 30 active tickets, 61 active contacts; the FY26 workbook gave a revenue dimension: 82.7% closed-renewal retention, $700K closed-won, $128K attrition), and the remaining four are honestly flagged as build-instrument (per hubspot-hygiene-pass-results; per devensoft-financial-baseline — all provisional, internal projection).

The risks and decisions the prior section surfaced — the four LMM kill-conditions, the capacity-vs-ambition collision, the leadership decision list — are what these metrics measure against. The question this section answers is: how will we know it's working? The answer is a dual-horizon scorecard (nine-month leading indicators measured monthly, twelve-to-eighteen-month lagging outcomes measured quarterly, and a maturity scorecard measured at month 0 and month 9) whose categories hold under any bet outcome and whose targets are bet-contingent. That answer becomes the evidence the recommendations section uses to justify the bets it recommends.

The Two-Tier Causal Chain

The metric architecture is a causal chain, not a scoreboard. Four north-star metrics sit across two tiers, connected by an explicit predictive relationship: if the Tier 2 leading indicators move in the first nine months, the Tier 1 lagging outcomes have a defensible basis for landing in the nine-to-eighteen-month window. Without that causal logic, metrics are just numbers on a dashboard — nice to look at, but carrying no predictive value for the leadership team trying to steer a nine-month transformation.

Tier 1 — Lagging (growth-first, per CEO directive 2026-08-07):

• Net-new ARR — the growth scoreboard. Baseline: $700,844 closed-won across 10 deals in Q1 FY26 (internal projection, provisional — booking shape, not recognized ARR; per devensoft-financial-baseline). Target deferred to the recommendations section, contingent on the pivotal bet (defend-first vs. Diversify-first produces different net-new-ARR targets). Cadence: quarterly.

• Net Revenue Retention — the retention scoreboard. Baseline: 82.7% closed-renewal retention — computed as closed-won divided by (closed-won + attrition), which is the booking-shape NRR proxy, not a true NRR (no expansion/contraction/down-sell movement is tracked today; per devensoft-financial-baseline — provisional). $127,977 attrition across 7 named accounts (all flagged High-risk pre-loss) gives the concrete downside denominator. Target deferred. Cadence: quarterly.

Tier 2 — Leading (operational, evidence-grounded):

• Reporting-cycle compression — the #1 demand made measurable. Baseline: directional — a customer-voice demo captures the anchor: a status deck that "takes the analyst 2-3 days... here we just did in a matter of seconds" (per demo-dwight-customer-voice ~45:06). One customer voice, not a systematic measurement — accepted as directional baseline; systematic verification is a stretch goal, not a blocker. Target: compress the cycle measurably by month 9 (specific target deferred to M14). Cadence: monthly.

• Maturity-scorecard progression — the product-discipline metric. Baseline: 1.5 out of 5 (2023 Modus assessment across 5 dimensions — Validation 1, Prioritization 2, Alignment 1.5, Communications 1.5, Implementation 1; per modus-enterprise-insights §6). Externally validated (13 enterprise interviews + 2 internal, heuristic evaluation), but three years old — carried as a reference label, not a constraint. Instrument: built (the evidence-checklist self-assessment, detailed in a later component). Target: measurable progression by month 9, measured against the mo-0 baseline run. Cadence: periodic (month 0 + month 9).

The causal logic connecting the tiers is grounded in the plan's own evidence. Reporting is the #1 demand (18/108 YouTrack, ~86 HubSpot tickets) — if the fundamentals fixes and the F7 reporting-data foundation compress the reporting cycle, customers experience the platform's value more directly, which drives retention (Tier 1 NRR) and creates the reference base for net-new growth (Tier 1 ARR). The maturity scorecard tracks whether the organization is shifting from sales-driven, anecdote-based decisions (Implementation 1/5) toward evidence-led product discipline — the organizational precondition for the plan executing at all. Tier 2 is the leading indicator because it measures what the plan is doing; Tier 1 is the lagging indicator because it measures whether the market rewards it.

The Metric-Baseline Inventory

Beyond the four north-star metrics, nine supporting metrics form the instrumentation tier — metrics this plan commits to instrumenting but does not yet set targets for. Their purpose is to give the leadership team operational visibility: these are the gauges that tell you whether the engine is running, not whether the car has arrived. Five of the nine had no baseline when the decision review ran in August 2026; two cross-cutting workstreams (the CRM hygiene pass and the FY26 financial baseline) have since moved five of the nine toward have-baseline status.

The CRM hygiene pass gave the count dimension: 25 active companies, 11 active deals, 30 active tickets, 61 active contacts — trustworthy active-denominators computed against the real active population, not the 19,649 total companies or 33,000+ total contacts that inflate the dormant portal (per hubspot-hygiene-pass-results § Clean denominators). The FY26 financial baseline gave the revenue dimension: 82.7% closed-renewal retention, $700,844 closed-won, $127,977 attrition, ~$1.92M projection — all internal projection, provisional, booking-shape not recognized ARR (per devensoft-financial-baseline). Together these move the active-customer-count metric from build-instrument to have-baseline and partially fill the revenue-backed metrics.

The residual — three metrics still build-instrument (feature-adoption, NPS/CSAT, API usage) plus one with no systematic measurement (competitive win-rate) — is the honest gap this plan converts into a named workstream rather than a confession. The instrumentation build is detailed in a later component.

Dual-Horizon + Dual-Rhythm

The metrics section presents two horizons on two rhythms, and the distinction matters because it prevents the most common failure mode in transformation measurement: confusing activity (leading) with outcome (lagging) and then abandoning the plan when the lagging numbers do not move fast enough. The nine-month horizon is where the leadership team steers; the twelve-to-eighteen-month horizon is where the market verdict arrives. Different cadences, different audiences, different decision utility.

The 9-month leading-indicator scorecard (monthly cadence). This is the steering dashboard. It carries the Tier 2 leading metrics — reporting-cycle compression, maturity-scorecard movement — plus the instrumentation tier's operational metrics (support-trend, utilization, demand-trend, active-customer-count). Monthly cadence because the plan's execution window is nine months and the leadership team needs to see directional movement within that window to course-correct. The leading scorecard is not a verdict on whether the plan worked; it is a check on whether the plan is executing — are the fundamentals fixes landing, is the reporting cycle compressing, is the instrumentation coming online. If Tier 2 is not moving by month 3-4, the problem is execution, and the 90/10 firewall or the capacity ramp is the place to intervene.

The 12-18-month lagging-outcome set (quarterly cadence). This is the scoreboard. It carries the Tier 1 lagging metrics — net-new ARR, Net Revenue Retention. Quarterly cadence because revenue and retention are reported on a financial cycle, and because monthly movement in these metrics is noise, not signal. The lagging horizon extends beyond the plan's nine-month execution window because the causal chain has a time lag: Tier 2 movement in months 1-9 is expected to produce Tier 1 results in months 9-18. Presenting these on the same horizon as the leading indicators would create the false expectation that growth metrics should move inside the execution window — an expectation the causal logic does not support.

The maturity scorecard (periodic cadence: month 0 + month 9). The scorecard runs twice: once at the start (month 0, to establish the real baseline) and once at the end of the execution window (month 9, to measure progression). It is periodic, not continuous, because the dimensions it measures — Validation, Prioritization, Alignment/Decision-Making, Communications/Feedback-Loop, Implementation/Outcomes — do not change week to week. They change when organizational discipline shifts, which is a quarter-scale movement, not a sprint. The 2023 Modus baseline (1.5/5 aggregate) is carried as a reference label — it is three years old, from an externally validated but point-in-time assessment — and the month-0 run establishes the current reality against which month-9 progression is measured (per modus-enterprise-insights §6).

The dual-horizon framing is honest about the time lag between doing the work and seeing the market reward. It gives the leadership team a monthly view of execution health (are we doing what we said) and a quarterly view of market outcome (is the market responding), and it resists the pressure to declare victory or failure on the lagging numbers before the leading numbers have had time to produce them.

Propose-Instrumentation — The Gap Becomes a Workstream

The most important decision in this metrics section is how it handles the absence of baselines. A weaker plan would confess "we do not have the data" and leave the gap as a weakness. This plan converts the measurement gap into a named workstream — a set of concrete deliverables that turn "baselines are absent" into "here is how we are building the instrumentation" — and in doing so makes the instrumentation itself a measurable output of the transformation.

The CRM hygiene pass — completed 2026-08-11. The HubSpot portal held 19,649 companies, 33,000+ contacts, 1,749 tickets, and 1,561 deals — but >99% was non-actionable across all objects. The hygiene pass classified every record across five segments (actively actionable, worth reviving, worth cleaning, safe to archive, safe to ignore) and established clean denominators: 25 active companies, 11 active deals, 30 active tickets, 61 active contacts (per hubspot-hygiene-pass-results). The active-customer-count metric moved from build-instrument to have-baseline. The pass also surfaced a data-ops quick win: the CRM has no structural linkage to tenancy (tenant_url populated for only 1 of 28 tenants), so backfilling tenant_url + active-seat counts across all instances is the next step to a true active-customer-count.

The one-time baseline census — the M13-owned artifact. This is the "baseline register" — the section's evidence appendix that tracks "here are our numbers, where they come from, and what is still missing." The CRM hygiene pass gave the count dimension; the FY26 financial baseline gave the revenue dimension (82.7% retention proxy, $700K closed-won, $128K attrition, ~$1.92M projection — all provisional, internal projection; per devensoft-financial-baseline). The census pulls these into one register, names the remaining gaps (true NRR needs expansion/contraction tracking; CAC/LTV needs channel-cost data that does not exist; feature-adoption needs per-feature telemetry; competitive win-rate needs structured loss-reason capture), and becomes the living artifact that the monthly leading-indicator scorecard reports against.

The feedback portal — recommended from M13, executed in M10/M08. Continuous NPS/CSAT/feature-request capture is the instrumentation that fills the build-instrument metrics (NPS/CSAT #7, feature-adoption #6, partially competitive win-rate #9). The HubSpot CX/survey fields are empty — not because customers were asked and declined, but because the capture mechanism was never built. The feedback portal is the M10/M08 capability workstream that converts these from build-instrument to have-baseline over the execution window. M13 names the need; M10/M08 own the execution.

This propose-instrumentation posture is the section's answer to the honesty principle (principle 4): credibility comes from accurately naming where the gaps are and what is being done about them, not from pretending the data exists or from softening the absence. The instrumentation tier is a named set of deliverables with owners, sequenced into the roadmap, and measurable against the monthly leading-indicator scorecard.

The Evidence-Checklist Instrument — Tier-2 Scorecard Metric

The Tier 2 maturity-scorecard metric has its instrument built: a self-contained interactive self-assessment that converts the five Modus product-strategy dimensions into evidence-gated capability ladders. The instrument ships alongside the plan as a self-contained interactive checklist — a deliberate method shift from interview-based scoring to objective-evidence self-assessment that makes the scorecard reproducible, bias-resistant, and free of a Modus dependency.

The method shift. The 2023 Modus baseline (1.5/5 aggregate; Validation 1, Prioritization 2, Alignment 1.5, Communications 1.5, Implementation 1) was produced by a UX consultancy conducting 15 interviews and a heuristic evaluation — an impression-based assessment by external observers. The evidence-checklist instrument replaces that method with a cumulative 1-5 capability ladder for each dimension, where every level is defined by observable evidence you can point to: documents that exist, processes that run, decisions that are traceable. Two assessors citing the same evidence converge on the same score. The method shift is documented transparently in the instrument itself — the 2023 score is carried as a reference label and score-bar marker, not as a constraint, and the integer-only scoring retires the 2023 half-points (per modus-enterprise-insights §6).

The five dimensions, as capability ladders:

1. Validation — from "no formal validation; decisions from customer requests/anecdote" (level 1, the 2023 baseline) toward evidence-gated investment decisions
2. Prioritization — from "basic criteria; cannot push back on exec requests lacking business case" (level 2) toward weighted, capacity-bound prioritization
3. Alignment / Decision-Making — from "partially aligned; impulsive; does not weigh options" (level 1.5) toward cross-functional alignment with documented trade-offs
4. Communications / Feedback-Loop — from "disconnected; no feedback loop" (level 1.5) toward continuous customer-feedback-to-roadmap cycles
5. Implementation / Outcomes — from "output over outcomes; sales-driven, not based on actual user data" (level 1) toward outcome-measured delivery

The HITL step that remains. The instrument is built and self-runnable, but a real month-0 baseline run — Diego and Roman scoring the organization against the evidence criteria — is a separate human-in-the-loop step. Its text-summary output joins the baseline register alongside the CRM count dimension and the revenue dimension . The month-9 re-run measures progression against that baseline. The instrument does not set targets (those stay deferred to M14, contingent on M04); it provides the measurement surface that makes progression trackable.

This instrument is the concrete deliverable that makes the Tier 2 scorecard metric operational. Without it, "maturity progression" is an aspiration; with it, the organization has a reproducible, bias-resistant self-assessment that two assessors can run independently and converge on — the precondition for the scorecard carrying predictive weight in the causal chain.

Bet-Agnostic Categories, Targets Deferred to M14

The decision this section once carried — "set the metrics, baselines, and targets" — is resolved by splitting it: the metric categories are bet-agnostic (they hold under any pivotal-bet outcome), and the target numbers are explicitly deferred to the recommendations section with a dependency link to the bet decision. This means M13 produces no new BTDR of its own; it provides the measurement framework that M14's recommendations organize around.

The bet-agnostic logic is straightforward. The four north-star metrics — net-new ARR, Net Revenue Retention, reporting-cycle compression, maturity-scorecard progression — are the right metrics regardless of whether the plan leans defend-first (Retain priority), Grow-first (net-new enterprise), or Diversify-first (LMM). What changes under each bet posture is the target for each metric, not the metric itself:

• A defend-first posture sets a high NRR target (the base must hold), a modest net-new ARR target (growth is secondary), and emphasizes reporting-cycle and maturity-scorecard movement as the leading indicators that the base is being defended effectively.
• A Grow-first posture sets an aggressive net-new ARR target, accepts a moderate NRR target (some churn is tolerated if net-new offsets it), and emphasizes competitive-win-rate and demand-trend as the leading indicators.
• A Diversify-first posture adds LMM-specific metrics (fake-door conversion, PLG sign-up rate) that are bet-contingent by definition and would not appear in a defend-first scorecard.

The metric categories — growth, retention, operational efficiency, organizational maturity — are the same in all three. The pivotal bet is now resolved as Hedge, sequenced — fundamentals-first (BTDR-0004 G4: R items take priority over G/D when capacity competes), but the target numbers still depend on leadership's investment commitment, the Diversify go/no-go, and the firewall enforcement level — all of which are leadership-decides items (per BTDR-0004). (An earlier "defend-first" amendment was introduced then reverted 2026-08-13 — it is not the decision of record.)

The format this section therefore uses for every metric is: "baseline [directional or TBD per census]; target [deferred to §14, contingent on pivotal bet §04]." This is not a hedge — it is the correct sequencing. Metrics first, targets when the bet is locked. The instrumentation tier (the 9 supporting metrics) carries no targets at all by design; it is the operational visibility layer, not the scoreboard.

Flag — Baselines Partially Addressed, Residual Stays OpenFlag

When the M13 decision review ran in August 2026, the baseline picture was stark: most metrics had no baseline at all, and the review's response was to reframe the absence as a named workstream rather than a confession. Two cross-cutting workstreams have since moved the picture substantially — but not completely — and the residual is what stays open.

What moved (count + revenue dimensions). The CRM hygiene pass gave the count dimension: 25 active companies, 11 active deals, 30 active tickets, 61 active contacts — trustworthy active-denominators (per hubspot-hygiene-pass-results). The FY26 financial baseline gave the revenue dimension: 82.7% closed-renewal retention, $700,844 closed-won, $127,977 attrition, ~$1.92M projection (per devensoft-financial-baseline — all internal projection, provisional). Together these give five of nine instrumentation-tier metrics directional baselines (4 hard + 1 directional). The 2023 Modus 1.5/5 scorecard (externally validated, 15 interviews) gives the Tier 2 maturity baseline. Dwight's "2-3 days → seconds" demo quote gives the reporting-cycle directional baseline.

What stays open. Five gaps remain genuinely unresolved:

• True NRR — the 82.7% retention proxy is closed-won ÷ (closed-won + attrition); it does not capture expansion, contraction, or down-sell movement. Computing a true NRR requires tracking seat-count changes and expansion revenue at renewal, which the current systems do not do.
• CAC / LTV / LTV-CAC — Devensoft has never sold to LMM (no LMM CAC exists); enterprise channel-cost is not tracked in the financial system. The unit-economics gap is named in the commercial-model section and carried as a leadership-decides item.
• Feature adoption — no per-feature telemetry exists. The platform does not track which capabilities each client actively uses, which means the 30-40% utilization figure (Modus) cannot be decomposed into "which features are used vs. orphaned."
• API / Deven Connect usage — API telemetry is not built. Deven Connect realized revenue is $0 (the $75k MSRP was theoretical/market-rejected), and there is no consumption-tracking infrastructure to measure API usage when the tiered pricing model activates.
• Competitive win-rate / loss-reason tracking — no structured loss-reason capture exists in the CRM. Deals are marked closed-lost but the reason is not systematically categorized, which means the plan cannot measure whether it is "winning again" against DealRoom/Midaxo or merely retaining a base that is not being contested.

These gaps stay flagged rather than filled because manufacturing baselines would be the stance drift the validation protocol guards against (P2/P4 — investor-grade honesty + provenance). The propose-instrumentation workstream (CRM backfill, feedback portal, API telemetry build) is the path to closing them over the execution window. Until then, the baseline register is honest about what it has (directional baselines for 5/9 metrics) and what it does not (true NRR, CAC/LTV, feature-level adoption, API usage, competitive win-rate).

Flag — Unverified Marketing Claims: Flag-and-ExcludeFlag

The vault's business-context note carries three marketing claims that have never been verified by any source in the evidence base: a 48% time savings, a 56% adoption rate, and a 7x reporting speed improvement. These figures appear in internal vault content (the business-context excerpt, captured 2026-08-02) and are flagged as unverified in the Source Ledger proposed-correction log. The Modus assessment — the richest external-demand source, 15 stakeholder interviews — neither confirms nor cites these numbers. The YouTrack demand sweep (108 issues, census-clean) contains zero references to them. The HubSpot customer-voice stream (640+ tickets) does not mention them.

This section's treatment of these claims is flag-and-exclude: they are named once here — in this footnote-equivalent component — and then excluded from the metric set. They do not appear as baselines, they do not appear as targets, and they are not cited as evidence for any capability claim elsewhere in the plan. If a future verification effort (a systematic customer study, an instrumented usage measurement) validates them, they re-enter the metric set with proper provenance. Until then, citing unverified marketing claims in an investor-grade strategy would fail both the provenance check (P4) and the honesty principle (principle 4 — credibility from accurately naming reality, not from asserting marketing numbers).

The 7x reporting-speed claim is the most tempting to cite because the customer-voice demo does capture a dramatic reporting moment ("takes the analyst 2-3 days... here we just did in a matter of seconds," per Dwight ~45:06). But a single demo moment on a pre-configured dashboard is not a systematic measurement of reporting-cycle compression across the installed base. The directional baseline this section uses for the reporting-cycle metric is the demo voice itself — "2-3 days, one customer" — not the marketing claim. The distinction matters: the demo voice is sourced and timestamped; the 7x claim is unsourced and unverified.

This flag is not a permanent exclusion; it is a provenance gate. The metric set is designed to accept verified evidence when it arrives. What it refuses to do is present marketing assertions as measurement.