How our value estimates work
Every building page shows our estimated market value next to the county's own numbers. This page explains what feeds that estimate, how the model works, and exactly how accurate it measured on recent sales. It is an estimate, not an appraisal, and we publish the error numbers because you deserve to know how much to trust it.
What feeds it
Public records only: King County Assessor tax rolls (appraised land and improvement values, 16 years of history), recorded real property sales (we keep verified arm's-length, single-parcel market sales and exclude family transfers, foreclosures, portfolio deals, and other non-market deeds), county GIS parcel data, and the county's own building records (size, age, use). No private data feeds the model.
How the model works
The core models the ratio of real sale prices to county values, adjusted for time, land share, building age, size, lot, scale, type, and city, blended with the most similar verified sold buildings nearby. On top of that, a component model values the land and the improvements separately, and a price-per-square-foot model cross-checks the result. The three are blended per segment with weights chosen on held-out sales, and a segment keeps its blend only where it measured at least as accurate as the core alone. Estimates refresh with each county data release.
The range, honestly
The range we show is a measured error, not a guess: it comes from the actual spread between our predictions and real later sale prices. About half of recent verified sales landed inside their range. When a segment has too few sales to measure, we widen the range or hold the estimate entirely.
Income view
For parcels with real building square footage and enough sale support in their segment, we also show a separate income view. It models annual rent per square foot, vacancy, operating costs, annual NOI, and a going-in cap rate, then divides NOI by that cap rate. This income value is a cross-check beside the MakeMe Estimate; v1 does not blend the two into one headline. If rent support, cap-rate support, or real building square footage is missing, the income view does not render.
Income view tested on 403 recent verified sales the income model had never seen. Median error 26.2%; 40% landed within 20%; 50% landed inside the shown income range.
| Segment | Sales tested | Median error | Within 10% | Within 20% | Inside range |
|---|---|---|---|---|---|
| multifamily | 201 | 23.8% | 20% | 42% | 50% |
| retail | 76 | 33.2% | 22% | 38% | 50% |
| industrial | 70 | 25.0% | 23% | 41% | 49% |
| office | 56 | 29.0% | 20% | 34% | 50% |
Measured accuracy (through 2026-05-29)
Tested on the most recent 12 months of verified sales the model had never seen (456 sales). Median error 15.1%; 32% of estimates landed within 10% of the real price and 63% within 20%; 50% of sales fell inside the shown range.
| Segment | Sales tested | Median error | Within 10% | Within 20% | Inside range |
|---|---|---|---|---|---|
| multifamily | 201 | 11.7% | 43% | 74% | 50% |
| retail | 76 | 19.7% | 24% | 53% | 50% |
| industrial | 70 | 18.3% | 20% | 51% | 49% |
| office | 56 | 18.7% | 25% | 54% | 50% |
| other commercial | 53 | 17.9% | 23% | 60% | 51% |
Off-market commercial estimates run wider than home-value tools: there are fewer sales, and public records do not contain rent rolls or in-place NOI. We publish our numbers anyway, and they improve with every county refresh and every recorded sale.
What this is not
Our estimate is not an appraisal and no appraiser is involved; where a county value appears it carries the county's own label. Everything on this site comes from public records and our models, provided as is, without warranty. It is not investment, legal, or tax advice. It is a starting point for a conversation between principals. County values are per the King County Assessor.
More about the platform: about makeme1031.