Nestfair
We valued 2,400 homes as if it were the day they sold, using only what was known that day. Then we compared each estimate with the price actually paid.
We picked 200 postcode sectors at random across England and Wales, weighted by how many homes sell in each, and took 12 sales from each between August 2025 and July 2026. The model was trained only on sales from before that, and none of these sectors were used to tune it or to set its ranges.
For each sale we hid the sale itself and everything after it: later sales nearby, a later resale of the same home, and any energy certificate lodged after the sale. We rebuilt the street adjustments from sales before August 2025. Each estimate only uses what a buyer could have known on the day.
The 95% confidence interval on the median error is 9.7% to 10.9%. We get it by resampling whole sectors, because homes in one sector share evidence and aren’t independent tests.
| Sales | Median error | Within 10% | Inside full range | |
|---|---|---|---|---|
| Earlier sale on record | ||||
| Sold before | 1,856 | 9.4% | 52% | 80% |
| No earlier sale | 544 | 14.0% | 39% | 79% |
| Price | ||||
| Under £250k | 888 | 12.1% | 44% | 76% |
| £250k–500k | 1,054 | 8.6% | 56% | 84% |
| £500k–1M | 387 | 10.8% | 46% | 79% |
| Over £1M | 71 | 19.4% | 27% | 63% |
| Property type | ||||
| Flat or maisonette | 294 | 11.9% | 43% | 78% |
| Terraced | 646 | 10.8% | 48% | 78% |
| Semi-detached | 756 | 9.3% | 53% | 82% |
| Detached | 704 | 10.5% | 48% | 80% |
| Region | ||||
| East Midlands | 264 | 12.0% | 44% | 75% |
| East of England | 300 | 8.8% | 57% | 87% |
| London | 264 | 10.6% | 47% | 80% |
| North East | 60 | 8.8% | 55% | 75% |
| North West | 276 | 11.7% | 45% | 80% |
| South East | 396 | 8.6% | 56% | 80% |
| South West | 264 | 10.2% | 49% | 81% |
| Wales | 144 | 11.8% | 44% | 85% |
| West Midlands | 216 | 12.6% | 43% | 72% |
| Yorkshire and The Humber | 216 | 10.9% | 47% | 77% |
A home with an earlier sale on record is easier to value, because we can carry its own price forward. About three in four of the homes in the test had one.
Expensive homes are harder. We ran a separate test on 465 homes we valued at £1 million or more, in 127 sectors kept out of everything else. The median error was 13.9%, and 83% of the sales landed inside our full range, which is what it’s built for. (The “Over £1M” row in the table looks worse because it groups homes by what they sold for, and a home that sold well above its estimate lands in that row partly for that reason.) Above £3 million we haven’t tested enough sales to say, so treat an estimate there as a starting point and lean on a surveyor.
Homes under £250,000 are harder than the middle of the market too. A few thousand pounds is a bigger share of the price, and condition counts for more.
On a £400,000 home, a 10% median error means half our estimates were within about £41,000 of the price paid. That’s close enough to tell you whether an asking price is in the right area. It isn’t a survey or a mortgage valuation, and it can’t see a new kitchen or a crack in the wall.
Sale prices are noisy as well. Two near-identical houses on one street can sell for quite different prices depending on who came to view.