How far do real sales stray from our fair value? Every month we value recent sales our models have never seen, against the final sale price, and publish the result for every council, postcode district and type of home in England and Wales. Nothing is left out to flatter the figure.
November 2026 edition. Models trained on sales to December 2025, then tested on sales from January 2026 to June 2026.
Scope: open-market sales registered with HM Land Registry in England and Wales (repossessions, company and bulk transfers left out), every type of home, final sale prices only: never asking prices or surveyors' figures.
Each month this test feeds straight back into our valuations. Every valuation corrects the model for the bias measured in its own postcode district and type of home, then blends in recent comparable sales nearby. The figures above are for the model on its own, before those steps.
We test the full valuation too, on the same sales, valued as you would see them with our standard settings: the bias correction taken from an earlier test, and only comparable sales registered before each sale.
| Typical gap | Within 10% | Within 20% | Bias | |
|---|---|---|---|---|
| The model on its own | 9.8% | 48.9% | 75.7% | +1.6% |
| Your valuation | 9.3% | 51.1% | 77.7% | +0.1% |
By type, your valuation's typical gap: detached 9.4%, semi-detached 8.7%, terraced 9.1%, flats 10.9%. A bias near zero means valuations sit at fair value: neither high nor low on average.
| Type | Test sales | Typical gap | Within 10% | Bias |
|---|---|---|---|---|
| Detached | 63,246 | 10.0% | 48.2% | +2.6% |
| Semi-detached | 81,295 | 9.2% | 51.3% | +1.9% |
| Terraced | 77,539 | 9.6% | 49.8% | +1.3% |
| Flats | 38,424 | 11.3% | 43.5% | -0.5% |
| Freehold | 209,186 | 9.4% | 50.4% | +1.8% |
| Leasehold | 51,318 | 11.4% | 43.2% | +0.4% |
Flats and leasehold homes vary more: floor, view, service charges and the lease itself aren't all in the public data.
| Region | Test sales | Typical gap | Within 10% | Bias |
|---|---|---|---|---|
| East of England | 29,407 | 8.6% | 54.3% | +0.8% |
| South East | 43,834 | 8.7% | 53.5% | +0.5% |
| London | 25,753 | 9.0% | 52.2% | +0.1% |
| East Midlands | 23,012 | 9.3% | 51.0% | +1.5% |
| South West | 28,830 | 9.8% | 49.1% | +1.0% |
| West Midlands | 24,008 | 9.9% | 48.5% | +2.3% |
| Yorkshire and The Humber | 25,385 | 10.8% | 44.7% | +2.6% |
| North West | 34,475 | 11.2% | 43.5% | +3.3% |
| Wales | 13,793 | 11.6% | 42.8% | +3.3% |
| North East | 12,007 | 12.3% | 40.2% | +3.3% |
The gap isn't the same everywhere. Across the 3,517 postcode district and type combinations with 30 or more test sales, the typical gap is 6.7% in the steadiest tenth, 9.5% in the middle one and 13.8% in the most varied tenth. Every valuation shows the figure for its own postcode district and type of home.
See the test for any council, postcode district and type of home: every sale against our fair value, the gap district by district on a map, and the full table, the most varied districts included.
Any valuation can quote a single accuracy figure. Before you rely on one, ask these five questions. Here are our answers.
| Ask | Why it matters | Our answer |
|---|---|---|
| Tested against what? | An asking price or an agreed offer already holds most of the answer. Only the final sale price is a fair test. | Final sale prices registered with HM Land Registry. |
| On sales the model had never seen? | A model tested on sales it learned from is measuring its memory, not its judgement. | Yes: trained to December 2025, tested on January 2026 to June 2026. |
| How many sales, from when, and which left out? | A small or hand-picked sample can give almost any figure. | 260,504 open-market sales, every type of home, re-tested and published every month. |
| The size of the error, or an average of errors? | If overestimates and underestimates are averaged, they cancel out: a model 20% too high on half the homes and 20% too low on the rest would average 0%. | We give the typical size of the gap (9.8%), the share within 10% and 20%, and the bias (+1.6%) separately. |
| For your area and type of home? | One national figure hides the difference between a steady street of similar homes and a varied market. | Yes: every postcode district and type of home, from 6.7% in the steadiest tenth to 13.8% in the most varied. |