`tk10873/cross_brand_gbp_check.md`

27 lines
# TK-10873 cycle 9 — is the Zoffany GBP-as-USD bug fleet-wide? (READ-ONLY diagnostic)

**Question:** cycle 4 found `zoffany_catalog.price_retail` stored UK £-prices as USD (~2.54×
underprice on 209 rows). The sibling SDG brands were harvested by the same pipeline — do they
share the bug? Method = compare each brand's `*_catalog.price_retail` vs its independent S1
feed harvest `retail_usd`, by mfr code, looking for the ~2.54× bimodal signature.

## Result

| brand | catalog | overlap | ratio~1.0 (clean) | ratio~2.54× (GBP-as-USD bug) | verdict |
|---|---|---|---|---|---|
| **Harlequin** | harlequin_catalog | 726 | **726** | **0** | ✅ CLEAN — catalog matches feed exactly; NO bug |
| Zoffany | zoffany_catalog | 244 | 30 | **214** | ❌ the known bug (fixed in v2/v3) |
| Morris | william_morris_catalog | 0 | — | — | ⚠️ INCONCLUSIVE — catalog uses `saw####` codes, feed uses `DJA…/DMY…/DBPW…`; incompatible key systems, can't cross-check this way |
| Sanderson / Scion | (no catalog on this mirror) | — | — | — | not checked |

## Conclusion
**The GBP-as-USD error is NOT a systemic pipeline bug — it is Zoffany-specific.** Harlequin, the
LIVE pilot brand (726 SKUs), is priced correctly against its own feed. This *bounds* the problem
to Zoffany and de-risks the epic: the sibling brands' prices do not need the ~2.5× correction.
Caveats: Morris couldn't be joined (key-scheme mismatch — worth a separate mapping if Morris is
ever repriced off its feed); Sanderson/Scion have no catalog on this mirror to check.

Note: join requires extracting the trailing mfr code from the catalog's slugified mfr_sku
(e.g. `houndstooth-...-haw0289-04` → `HAW0289-04`) and uppercase-matching the feed. Read-only;
no writes to any sibling catalog (ownership respected).