ZIP-level market metrics are served from stored Redfin ZIP features when available and fall back to live fetches during migration.
FHFA HPI repeat-sales data is used when available to stabilize ZIP-level price momentum versus raw median sale price.
When the public ZIP feed omits price-cut activity, the model derives a conservative seller-concession proxy from sale-to-list ratio, bidding intensity, and days on market.
Median list versus median sale pricing is now used as an additional public-data concession signal when available, so seller softening is not inferred only from price-drop counts.
Listing quality is still public-data-first: direct market fields and explicit proxies are available today, while withdrawals, expireds, relists, and explicit concessions remain roadmap items for MLS or vendor feeds.
Unemployment remains county-level context because ZIP-level labor-market data is not generally available from the same public sources.
Freddie Mac PMMS mortgage rates are national averages applied as financing context rather than ZIP-specific pricing.
County building permits and ZIP vacancy context add future-supply signals that can loosen buyer conditions before that inventory fully hits the resale market.
Seller lock-in is modeled as a proxy that blends the current mortgage-rate gap with local owner-occupancy and owner-tenure context from ACS, not observed loan-level seller data.
Affordability blends a current-rate financed payment with ACS local owner-cost and property-tax context, falls back to state-level property-tax and home-insurance benchmarks when needed, and also reflects whether that burden has improved or worsened versus last year.
This ZIP report closes the biggest granularity gap with the competitor sample while preserving the improved leverage signals.