A Proanex productWe publish how often we're wrong.
A valuation you can't check is a guess. So we test ours the hard way: we value homes that have already sold, as they stood on the day they sold, and compare our figure with the price actually paid.
median gap between our value and the price paid
sale prices fell inside the range we gave
of homes valued within 20% of the price paid
average bias: neither high nor low
Homes valued as they stood on the day they sold: 290 sales across 40 postcode districts in England and Wales, July 2025 to June 2026. Measured 28 September 2026.
A band that separates good from weak.
Every valuation is labelled high, medium or low confidence. The label is read from how widely the evidence varies, and it holds up: high-confidence valuations miss by less.
Median error by band on the 28 September 2026 sample.
Fair tests, run every time.
As it stood on the day
Each home is valued using only what was known before it sold: earlier sales, certificates lodged by then, listings seen by then and the price index up to that month.
The product's own engine
The test runs the same code as the property page, so what we measure is what you get.
Ranges tested out of sample
Ranges were fitted on half the homes and tested on the other half, thirty times over: they held 80.4% of prices.
A check before every change
Before any change to how we value homes is merged, a regression check is run. It fails if accuracy slips beyond the test's noise or the range stops holding about 8 in 10 prices.
Measured, then improved.
Publishing the numbers keeps us honest about our own method, too.
- Ranges that hold. Our earlier ranges held the sale price only 44% of the time on 1,929 recent sales. Calibrated ranges now hold it 8 times in 10.
- Closer values. Like-for-like comparables brought the median error on 600 recent sales from 10.7% to 9.3%, and the share within 20% from 74% to 78%.
- Extensions accounted for. On homes extended since their last energy certificate, the bias fell from about 17% low to about 7% low.
The limits, in plain sight.
- Flats are harder. Their median error is around 12%, and their ranges are wider to match.
- Recently extended homes. Still valued a little low on average when the extension isn't on record.
- Scotland. Market data covers Scotland, but valuations need the Scottish sales register, so we don't offer them there yet.
- Rents are asking rents. Market rents come from comparable lettings as advertised, not rents achieved. We changed the rent method on 1 October 2026 and will publish its accuracy once it has been measured.
Accuracy on your own work.
Inside Viiyoo, each organisation sees how its own valuations, rents and decisions fared against what then happened, and is told if the claims stop holding.
Valuations against sales
When a home you valued sells, the price is set against your valuation.
Rents against lettings
Your rent assessments against what comparable homes then let for.
Decisions against outcomes
How the decisions you signed off have fared since.
Test us on homes you know.
Send us a few addresses and what they sold for. We'll show you how close Viiyoo gets.