docs(plan): close Step 9 — recommender simulator deferred
Some checks failed
Build & Push Docker Images / lint (push) Failing after 36s
Build & Push Docker Images / build-and-push (push) Has been skipped

NocoDB inspection (via MCP) confirms the legacy Interests table carries
only the current Sales Process Level value plus point-in-time event
timestamps as text fields — no dedicated stage-change history table.
That isn't enough resolution to replay stages-over-time through the
recommender's tier-ladder + heat-score weights. Simulator deferred
until ~10+ real wins accumulate under the new pipeline, then we can
simulate against actual CRM history.

The existing /admin/berth-recommender heat-weight tuning UI is
sufficient for v1 launch.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-05-14 15:55:17 +02:00
parent 3c2826635d
commit a8607ecc9e

View File

@@ -148,12 +148,25 @@ Mechanical. Run last to minimize merge churn. ~2h.
Dedicated commit. Email template URL builder, page-side fragment readers, Better Auth integration test. ~1h. Dedicated commit. Email template URL builder, page-side fragment readers, Better Auth integration test. ~1h.
### 2.9 Step 9 — Optional: NocoDB inspection + recommender simulator ### 2.9 Step 9 — NocoDB inspection complete: simulator DEFERRED
Pre-flight: use the NocoDB MCP to inspect what stage-advancement / win history exists in the legacy interest records. NocoDB `Interests` carries only the current `Sales Process Level`
single-select + a handful of point-in-time event timestamps
(`EOI Time Sent`, `Time LOI Sent`, `clientSignTime`,
`developerSignTime`, `EOI_Completed_At`, `finalized_document_sent_at`)
scattered as text fields. There is **no dedicated stage-change
history table** — only the most recent stage value survives.
- **If recoverable**: build the "simulate against history" admin tool that replays past wins through current recommender weights. ~half day. The recommender simulator's tier-ladder + heat-score logic depends on
- **If not**: defer until production accumulates ~10+ wins. Update `BACKLOG.md` to reflect. "how long did this deal sit at each stage" and "which stage did past
deals make it furthest to before falling through." Without an
advancement timeline that's not recoverable: every imported interest
collapses to one data point.
**Decision (2026-05-14):** defer the simulator until production
accumulates ~10+ won deals under the new pipeline — then the simulator
can replay against real CRM history. The existing per-port heat-weight
tuning UI in `/admin/berth-recommender` is sufficient for v1 launch.
--- ---