How sold house-price data really works

Actual completed sales, not asking prices

The gold-standard source isHM Land Registry Price Paid Data: the official record of property sales that have actuallycompleted and been registered in England and Wales. Every entry is a real transaction at the real price paid — not what a seller hoped for, not what a portal’s model predicts. This is the data PostcodeLens uses for the sold prices on every postcode page.

That’s a meaningful distinction from the estimates you see on property portals. An automated valuation (the “we think it’s worth…” number) is a model’s guess, and an asking price is a seller’s hope — both can sit some way from what a home eventually sells for. Price Paid tells you what buyers and sellers actually agreed and settled on, which is the number that matters when you’re gauging a market.

Recent months are always incomplete

Because Price Paid records registered sales, there’s a built-in lag. A sale completes, then the transaction has to be lodged and registered with the Land Registry, and only then does it appear in the data — a process that commonly takes weeks to months. The practical consequence: the most recent months are always incomplete, because plenty of sales from that period simply haven’t been registered yet.

So a dip in the very latest month is usually an artefact of registration lag catching up, not a real fall in prices. PostcodeLens works in 12-month windows rather than fixating on the newest data point, precisely so a half-registered recent month doesn’t masquerade as a trend.

Why medians beat averages for prices

When you summarise a set of sold prices into one number, themedian (the middle value) is almost always more honest than the average (the mean). House prices are “skewed”: a single £4m sale on a street of £300k homes drags the average sharply upward, producing a figure that describes no real house on the street. The median — the price with half the sales above and half below — shrugs that outlier off and stays close to the typical home. That’s why PostcodeLens reports medians.

The small-sample trap

A median is only as trustworthy as the number of sales behind it. With just a handful of transactions, even the median can be pulled around by a few atypical sales — producing a figure that istechnically true but unrepresentative of the ordinary market.

The textbook case is SW1A 1AA, which sits in the same postcode sector as Buckingham Palace: its few registered sales are enormous and nothing like a normal home, so a median built on them would be correct and completely misleading at once. PostcodeLens guards against this directly. When a median rests on fewer than 30 sales, the page flags it — telling you it’s“based on a small number of transactions… treat the shape of this line with caution” — rather than presenting a thin-sample figure as settled fact.

Postcode, sector, or wider area — and why we say which

A single postcode often doesn’t have enough recent sales to produce a reliable figure at all. Rather than show nothing, PostcodeLenswidens the area until there’s enough data — and always tells you which area the figure actually covers. The order is:

• This postcode, last 12 months — if there are enough sales;
• otherwise the postcode sector (the wider group your postcode belongs to), last 24 months;
• otherwise the local district, last 36 months;
• and if even that is too thin, we say the figure is suppressed rather than publish an unreliable one.

This is where honest attribution earns its keep. A median labelled “this postcode” and one labelled “postcode sector” are different claims about different areas, and quietly presenting a wider-area figure as if it were postcode-specific would overstate how local the number is. PostcodeLens never shows a widened-area figure without saying so — and where even the district is too sparse, it says plainly that there are“too few sales nearby… to publish a reliable median”, which, as with crime, means the data is undisclosable at that scale, not that the area has no market.

You can see the attribution in action onUB7 9AB,M1 2HF, andBS1 5TR — each shows the area its median is based on alongside the figure itself.

In short

• Land Registry Price Paid is actual completed sales — not asking prices or portal estimates.
• The most recent months are always incomplete because of registration lag — don’t over-read the latest point.
• Medians describe the typical home; averages get dragged around by outliers.
• A median on very few sales (under 30) can be true but unrepresentative — PostcodeLens flags this.
• Check which area a figure covers (postcode, sector, or district) — it’s always stated.

Next: how to check flood risk before buying, or what police.uk crime data actually means. For the full detail of how figures are aggregated and suppressed, see themethodology page.