Vintage loss curvesCR-02
Is the book we are writing now worse than last year’s?
How much of this is measured: Curve SHAPE between the published points is modelled, not measured — ACF publishes peaks and 12M averages, not a full MOB series. The DWH fact must be built at cohort × channel × scheme × month-on-book or this chart cannot be produced for real (03 §6).
- Formula
- cumulative loss by MOB, by cohort/channel/scheme
- Unit / direction
- pct · lower better
- Source system
- DWH
- Refresh
- daily_t1
- Target
- TBC
- Target status
- PROPOSED BY IKIGAI (T-01)
- Baseline in context/
- TW ~47%, Cell Phone ~40% (mgmt, ⚠)
Both proposed by Ikigai and awaiting ACF confirmation (DEC-02).
- Clause that fired
- no Layer-1, regulatory or agreed-target signal in the registry today
- Share of financial
- 1.3%
- Of total attention
- 0.53% at the approved weighting
The band is derived, not chosen — a published rule over facts already in the registry, re-runnable with tools/assign_weights.py. It is what the strategy simulator splits a perspective’s weight by, which is why it is not cosmetic: change this band and the shortlist changes. Q-28 asks who at ACF owns the decision.
No commentary recorded for this metric. A red tile with no explanation is the failure mode the current monthly pack already has.
Commentary is attached to the metric and period, not to a slide. It survives the month — which is what makes a variance reviewable at the next meeting instead of re-argued. Prototype — read-only fixtures; at fullstack this is the COMMENT table with author identity from AD.