Methodology
How the numbers work
Section titled “How the numbers work”Every number this system emits is supposed to satisfy four rules: produced by exactly one code path, versioned, explainable, and falsifiable against realized outcomes. This section documents how close the current model (p5v4-2026.07) gets, including where it falls short. Start with limitations and accuracy. In the current point-in-time segment scorecard, MdAPE among issued Tier-1 values ranges from 12.2% to 16.4%, but 41.7%–78.9% of attempted Cook subjects are suppressed and Lake valuation coverage is zero. Error and coverage must be read together.
The three places a number can come from
Section titled “The three places a number can come from”| Source | What it produces | Cadence |
|---|---|---|
| Tract signals | 26 precomputed metrics per census tract (1,492 tracts, Cook + Lake), percentile-ranked county-wide | monthly refresh |
| Comp engine | Per-address comp sets pulled live from Cook County sales records, filtered and gated at request time | per request |
| Underwrite solvers | Deterministic deal math (offers, DSCR, carry, economics gate) on top of the first two | per request |
Nothing else generates numbers. When a request can’t be served from these three honestly, the answer is “unscored” or “insufficient comps” with a reason string — never a filler value.
- Scores — the 0–100 scale, its known compression, the deal-score formula, unscored states
- Flip score · BRRRR score — factor weights and penalties
- Risk score — scored separately from opportunity, flag by flag
- Valuation — comp approach, income approach, reconciliation, refusal behavior
- Rent model — SAFMR-anchored bedroom rents, the renovated multiplier
- Underwriting — offer solvers, carry model, the economics gate
- Data sources — every feed with cadence, vintage, and coverage gaps
- Data coverage — the honest per-source matrix: what’s fully valued, context-only, or not tracked at all
- Accuracy — the published backtest
- AVM quality control — the five interagency QC factors mapped to current practice
- Invariants — the eight standing integrity assertions and their live status
- Limitations — read first
Data licensing roadmap
Section titled “Data licensing roadmap”Inventory (what’s for sale or rent) and valuation evidence (comps) come from different legal baskets, and that basket is shifting on purpose. Today: public active listings come from the licensed RentCast feed; legacy scrape lanes fail closed. The system’s off-market parcel engine uses county records rather than listing content — see data sources for the current mix and coverage for what each source actually reaches. Comp evidence for valuations is always county sale records (public record), never scraped listing data, regardless of which inventory sources are active. Next: once the operator’s Illinois broker license and MRED participation are both active, licensed MLS data (via the RESO Web API) becomes the inventory source of record after contractual display review. See brokerage disclosures for what activates on that day. Nothing about the valuation math changes when that happens; only where “what’s currently for sale” comes from does.
Every API and MCP response is stamped with model_version; changes that move numbers are logged in the changelog with what changed and why.