HouseValuation

How the valuation works

The valuation is an indicative figure derived from real local sales, not an estate-agent guess and not a paid AVM black box. Everything below is fully transparent.

1. Two datasets, joined

We work from two open datasets that cover England and Wales:

We join the two on postcode + house number/name. The Land Registry doesn't record floor area, which is what we need for £/m² benchmarks. The EPC register does. About 84% of recent sales join cleanly.

2. Pulling the candidate pool

For a given address, we look up every Land Registry transaction in the same postcode outcode that completed in the last five years, capped at 2,000 candidates. Each candidate is joined to its EPC record (if available) for floor area + energy rating.

3. Composite scoring

Every candidate gets a composite score that mixes five signals. The weights are deliberately conservative:

SignalLogicWeight
RecencyExponential half-life of 2 years30%
ProximitySame postcode > same outcode tier25%
Size matchFloor area within ±15% of target20%
Property typeDetached / semi / terrace / flat with adjacency rules15%
EPC bandBand match, with ±1 tier softening10%

4. Estimated value

From the ten highest-scoring candidates, we take the medianprice-per-square-metre and multiply by the target's EPC floor area. The median is more robust than the mean against single high-end outliers.

What this is not

Treat the figure as a useful starting point for thinking, not the final answer.