Seven things a sold price cannot tell you

The sold-price record is the best housing data Britain publishes and it is a short list of fields: a price, a date, an address, a property type, a tenure and whether the home was new. Everything people want to know beyond that is absent, and most of the ways this data gets misused come from not noticing which is which. Here are seven limits, with the size of each where it can be measured.

1. It cannot tell you what condition the home was in

There are no attributes in the register at all: no state of repair, no kitchen, no rewiring, no damp. Two identical houses on the same street, one refurbished and one not, appear as two prices with nothing to separate them. This is the single largest source of variation inside any local median, and no amount of statistical work recovers it, because the information was never recorded.

2. It cannot tell you what the seller wanted

A sold price is the end of a negotiation. The asking price, the number of reductions, the time on the market and the number of offers are all absent, so the register cannot tell you whether a home sold quickly at a fair price or slowly at a reduced one. Those are different markets and they look identical here. Asking-price indices exist for exactly this reason, and they answer a different question.

3. It cannot tell you whether the sale was ordinary

Not every transfer is a sale between strangers at market value. The register marks a category for this, and this site uses standard sales only, which excludes transfers to family, repossessions and portfolio transactions. What it cannot exclude is a sale that was technically arm’s length and practically not: a probate sale, a divorce, a seller who needed to move by a date. Those sit inside every median and they are part of why the middle 60% band is published beside it.

4. It cannot tell you how long the lease had left

Tenure is recorded, so a sale is known to be freehold or leasehold. The number of years remaining is not, and on a flat it is one of the largest single determinants of value: a lease with 70 years left and one with 950 are different assets at the same address.

The consequence is a real limit on what can be published. Splitting a district by tenure halves each sample, and a median needs at least 30 matched sales to be published at all.

A leasehold median is publishable in 1,034 of 2,150 districts, which is 48% of the districts that have any leasehold sales at all. In the rest, the honest answer is that there is not enough to say, and the page says so rather than printing a figure built on eleven sales. Where both are publishable, the gap is worth knowing.

5. It cannot tell you the home was the same size when it sold

Floor areas come from Energy Performance Certificates, and a certificate records a measurement on the day it was issued. Match a recent certificate to an old sale and the home is being priced at the size it is now, extension included. The error runs one way, because homes grow more often than they shrink, and it makes historic rates look too low and therefore growth look too fast. How large that is, and how much of the market can be measured at all, has its own article.

6. It cannot tell you a precise number

A median is an estimate from a sample, so it has an interval, and the typical district median rests on 191 matched sales and carries a 95% interval about 4.38% wide either side. That is narrow enough to be useful and wide enough that two districts a few per cent apart are not reliably different. Quantiles need a larger sample still, at least 50 matched sales, which is why some districts publish a median with no middle-60% band beside it.

There is also a fence: a small share of sales produce a price per square metre so far from the local distribution that they are excluded from the rate, while staying in the counts and the transaction lists. The fence is applied per property type, because a fence that removes a similar share of each type is filtering errors and one that does not is filtering a type. The spread is .

7. It cannot tell you what happens next

Every figure here is a description of transactions that have already completed and been registered. There is no forecast anywhere on this site and there will not be. The register also lags: a sale appears weeks to months after completion, so the most recent period is always incomplete and always revises upward as registrations arrive.

So what can a sold price tell you?

Read as what it is, the record answers a narrow question extremely well: what did homes like this, in this place, actually sell for, and how much did that vary. That is enough to judge an asking price, to compare two areas, and to see whether a market has moved. It is not enough to value a specific home, and every service that claims otherwise is filling these seven gaps with an estimate and not telling you which parts are estimated.

The practical version: use the data for the arithmetic and a human for the condition. Figures on this page are read live from the published dataset and are current to June 2026.