`data/next-cohort-scan.md`

48 lines
# TK-10875 — Next-recoverable-cohort scan (read-only) · 2026-08-30

Applied the Wolf-Gordon recovery method (corrected join + cost-population check) to the
top unbuyable vendors, to find the next bounded recoverable cohort. **Result: there is
none — Wolf Gordon was the exception.**

## Method
For each top-unbuyable vendor: (1) locate its `<vendor>_catalog` staging table + cost
column, (2) test `shopify_products.sku = catalog.dw_sku` AND `mfr_sku = mfr_sku` join
yield with cost>0, (3) check whether the catalog's cost column is even populated.

## Findings

| Vendor | Unbuyable | Staging cost populated? | Join yield | Verdict |
|---|---:|---|---:|---|
| **Wolf Gordon** | 269 | ✅ 5,135 rows | **224** (sku=dw_sku) | **RECOVERED + shipped** |
| Phillipe Romano | 8,274 | ❌ 0 rows (empty col) | 0 | cost-sourcing |
| Phillip Jeffries | 2,338 | ❌ empty | 0 | cost-sourcing |
| China Seas | 1,776 | ❌ 0 rows (quadrille_house) | 0 | cost-sourcing |
| Coordonné | 1,299 | ❌ price_trade empty (+DWDC↔DWCO2 prefix gap) | 0 | reconcile + cost |
| Pierre Frey | 659 | ❌ empty | 0 | cost-sourcing |
| Rebel Walls | 494 | (rebelwalls_catalog) | 0 | reconcile/cost |
| Arte International | 489 | ❌ 0 rows | 0 | cost-sourcing |
| **Scalamandre** | 474 | ✅ **4,877 rows** | **0** | **identity-reconcile** (keys diverge) |
| Quadrille | 443 | ❌ 0 rows | 0 | cost-sourcing |
| Alan Campbell | 434 | ❌ 0 rows | 0 | cost-sourcing |
| CMO Paris | 344 | ❌ empty | 0 | cost-sourcing |
| Malibu Wallpaper | 339 | ❌ empty (malibu_catalog) | 0 | cost-sourcing |

### The Scalamandre exception (has cost, but keys don't reconcile)
- live `shopify_products`: `dw_sku='DWA-89131'` (DWA prefix), `mfr_sku='wp88373'`
- `scalamandre_catalog`: `dw_sku='DWSC-111446'` (DWSC prefix), `mfr_sku='SCWP885230003'`

Different prefix families AND mismatched mfr numeric cores (88373 vs 88523…) — the live
Scalamandre line was onboarded under a different SKU scheme (likely the "Architectural
Wallcoverings" DWA- rebrand) than the catalog scrape. Recovering these 474 needs a real
identity reconciliation (pattern+color match or re-scrape), not a join fix.

## Conclusion
**TK-10875 bounded recovery is COMPLETE** (Wolf Gordon 224, shipped). The remaining
~24,990 unbuyable split into two SEPARATE initiatives, each warranting its own ticket:
1. **Cost-sourcing** — most top vendors have empty staging cost columns → vendor scrapes /
   vendor-price-ingest (largely gated, large scope).
2. **Identity-reconciliation** — Scalamandre (has cost) + Coordonné/Rebel Walls (prefix
   gaps) need SKU/prefix reconciliation before their cost can join.

Neither is a bounded quick-win loop increment.