Agreement instrument per TSK-47: seeded scores with provenance; adjust until agreed. The agreed matrix is the evidence spine for the M08 editorial (TSK-41), NLM verification (TSK-48), and the DevenMnA pitch (TSK-49/50). Columns: Devensoft v23 (current, as-is — baseline) · DevenLite (Targets + Lite Tasking) · DevenMnA (greenfield standalone — opinionated, less configurable, not quite a rewrite) · Devensoft v24 (through Phase 3).
Spec: .scratch/00_cross-cutting/issues/47_lifecycle-canonicalization-fit-to-need.md · Synthesis: 01_ingest/2_synthesized/fit-to-need-canonical.md · Data of record: fit-to-need.json (sibling)
Scoring rubric — what a number means
Basis: fit = fills the needs of the stage, where the stage need = the need-bar defined below. DevenMnA is scored against the minimally-necessary subset of that bar for its LMM target user (stated, not smuggled). Bands rebalanced by Diego 2026-08-17 (post-build) for better spreads — supersedes the original TSK-47 Q5/Q11 edges. Provenance: V vault/code-grounded re-derivation (v23 column, clean-room) · S seeded-NLM verbatim (unverified) · B blended-NLM remap · D derived · D! derived low-confidence · A adjusted-by-Diego. All V/D/D! cells re-derived against the need-bar (2026-08-17 definitions pass).
Matrix — 9 canonical stages × 4 versions
Adjust score — /
Current:
Coverage strip — stages at band, per column (the factual story)
Facts, not editorial: v23's strength is the integration core (7–8) — but Diego's 2026-08-18 pass re-banded it config-driven, not native-complete (7: 63 partial, 8: 75 config; strips: zero native stages), with config-band depth at Pipeline/DD (2–3) and the front tracking-not-sourcing (1); middle thin (4). DevenLite = front-only by Diego's design (1–2 partial/config, PMI stages 6–9 out of scope ≈ 0). v24 = v23's shape, deepened — AI-through-Phase-3 shifts partial→config (2/3/7/8) but adds no native stage and leaves the thin middle thin (1 · 4 · 6) and Regulatory unbuilt (5 — platform posture ≠ M&A clearance workflow). DevenMnA = the only column with native stages (4 — incl. DD 93, the matrix's top score, and value-realization 87). Interpretation belongs to the pitch (M09-GRILL Q11 pattern).
Stage need-bar — the definitions scores are scored against
Source: 01_ingest/3_distilled/ma-lifecycle-stage-definitions.md ← Diego-supplied 2026-08-17 ("just a starting point", appendable). Aspirational capability definitions — NOT demand evidence. ⚠ The bar is AI-forward / modern-practice-tilted: scoring against it is deliberately stricter than scoring against enterprise status quo — that strictness defines the DevenMnA/v24 ambition. Each stage: the 5-bullet need bar + v23's core-code gap vs it.
Rows 5–6 are derived, low confidence for non-v23 columns (thin signal).
Matrix AGREED (first pass) 2026-08-18 — all four columns (export 'fit-to-need (4).json'); v23 r7/r8 + v24 r7/r8 adjusted same day.
BTDR-0010 resolved by Diego adjustment 2026-08-17: DevenLite PMI = out of scope (stages 6–9 ≈ 0) — the "barbell" (Lite = 1–2 + 7–9) is retired for DevenLite; glossary "5 of 9" caveat + M09/M14-03 language amend on sync.
Scores vs MVP (Diego 2026-08-18): DevenMnA scores = full target scope / definition of done. The bold-path timing reference baseline (same 9-mo window as Hedge) commits a production-ready MVP = general ~50% of each stage's need-to-fit scope — thin breadth across all 9 stages, not 100% of a few. The agreed scores are not the MVP; scores unchanged.