Monthly stream of product updates, model releases, and fixes. Structural changes carry an explicit API label so integrators can scan for the ones that require their attention.
Per-metro momentum is now a 70/30 blend of the trailing 4-quarter YoY and the last 2 quarters annualized — faster reaction to direction changes without losing stability. Zillow For-Sale Inventory joins as the fifth bounded per-MSA feature alongside permits, unemployment, GDP, and the rate drift.
The model →Drop two CBSA codes into the URL — /compare?a=12060&b=38060 — and we'll show forecasts, signal, risk grade, and a price-history sparkline side-by-side. Free, no sign-up. Linked from every market page.
Compare →Every /markets/{cbsa} page now shows how our 12-month forecast for that metro has moved across runs (with the 80% confidence band) and which inputs are driving the current number — momentum, mean-reversion, rate drift, and the per-feature nudges. No black box, no gating.
Example →/movers (top forecast revisions this run), /markets/{cbsa}/forecast-track (track of our calls), /forecast-backtest (latest accuracy snapshot). All in the OpenAPI spec; all served with edge cache headers.
API reference →Every monthly forecast run now compares the prior year's predictions to today's realized prices and writes one aggregate row of MAE / MAPE / hit-rate / 80%-coverage — beats vs. naive YoY-persistence baseline. Powers the new headline tile on /performance.
Performance →Every /request-brief submission lands in /app/leads with conversion-by-source rollup and CSV export. Tagged with the ?ref query param so the operator can see which channel converts.
Portfolios with rolled-up metrics, custom signal rules, walk-forward backtesting, scheduled reports, Slack/Teams/PagerDuty destinations, per-key policies, live SSE event stream, SAML/OIDC SSO, what-if scenario builder, integration marketplace, onboarding checklist.
Pricing →Typed spec for every public endpoint. Pair with /docs/api for the rendered reference, or pull into Postman / openapi-codegen directly.
API reference →Auto-generated deltas, pairwise correlation, volatility, max drawdown. Preset crop windows (Since GFC, Post-COVID, last 5y/10y). Multi-select from Markets drops straight into Compare.
Compare →Baseline forecast moves from a linear trend to a mean-reverting momentum blend (local YoY × national drift × national mean). Comparative skill scoring against the v1 model lands with the first production backtest.
Methodology →The March 2026 FHFA release added a new column; our ETL now tolerates schema additions without halting. Backfill completed for the affected quarter.
Every /api/v1/* response now includes X-RateLimit-Limit, X-RateLimit-Remaining, and X-RateLimit-Reset. 429 responses always carry Retry-After in seconds.
Expanded scoring from the top 150 to every CBSA in the FHFA HPI series. Minimum history requirement: 20 quarters (existing metros unaffected).