Data science at Housometer
Every number we publish traces back to an official record — a completed sale, a certificate, a register entry. This page explains the datasets we hold, the methods we run on them, and the rules that keep our figures honest.
The datasets
All official, all open-licence, refreshed as their publishers release.
Every recorded property sale in England & Wales this century — price, date, address, type, tenure, new-build flag. The backbone of every price figure we publish.
Floor areas, property types and energy ratings — the certificate records we hold matched to individual addresses. Joined to sales at the same-address level, this is what lets us publish honest £ per square metre.
The official mix-adjusted index, including per-property-type averages — our benchmark for long-run and area-level comparisons.
HM Land Registry’s registers of property owned by UK companies (CCOD) and overseas entities (OCOD) — who owns what, and via which jurisdiction.
The official measure of relative deprivation, which we cross with price data for distributional analysis.
Flood zones, coastal erosion, mining hazards, storm overflows, broadband, air quality, schools, crime, transport — the evidence base behind every property report.
The full source list, with licences, lives at data sources.
The methods
Chosen so that a sceptical reader — or a journalist on deadline — can verify the claim.
We match every property to itself across its consecutive sales — 9m+ pairs — so questions like "did the seller make money?", "how long do owners hold?" and "do leasehold houses underperform?" are answered from real completed round-trips, not from an index or a model.
Sold prices are joined to EPC floor areas only at the level of the same individual address, with implausible floor areas discarded. No averaging across buildings, no imputed sizes — if we can’t match a sale to its own certificate, it isn’t in the £/m² figures.
Nearly every figure we publish is a median. One mansion sale can’t mint a "millionaire street"; one distressed sale can’t sink a town. Where an average is the right tool (official HPI series), we say so.
When we compare groups — leasehold vs freehold, flats vs houses — we re-run the comparison inside the same postcode districts, so a north–south mix difference can’t masquerade as an effect.
The honesty rules
The constraints every published figure passes before it goes live.
- Minimum sample sizes everywhere: streets need 5+ sales for a median, 8+ for a national ranking; towns need hundreds of matched resales before they enter a league table.
- Outlier bounds on movement figures: street-level swings beyond +120% or −60% are treated as redevelopment or mix change, not price movement, and excluded.
- Nominal gains are labelled nominal, sample-based figures state their basis, and our complete sales record starts in 2001 — so we never claim "since 1995".
- Thin pages are removed rather than padded: if an area lacks the data for a meaningful table, the page 404s instead of showing noise.
- Every research story carries its own methodology note and its official sources, and figures recompute automatically as new data registers.
The research these methods produce is collected at Housometer Research — original analysis like the fleecehold discount, where flat sellers are losing money and Britain’s new millionaire streets, all free to cite with attribution. For anything deeper — a custom cut, our workings on a specific claim, or comment on a housing-data story — get in touch; turnaround is usually same-day.
This analysis comes from the data behind every report.
Search any address in England & Wales for a free Home Confidence Score — or the full £5 report on the home itself.