Home Dashboard Download this document Terms & Conditions

User Guide & Methodology

Real Data Warehouse · Version 1.2 (draft) · October 2026

This guide explains exactly what the numbers on this site mean: where the raw data comes from, how it is filtered and processed, how often it updates, and the limitations you should understand before relying on any figure. It is written to be reproducible — a motivated reader with the public source files should be able to reconstruct our headline counts and totals exactly.

1. Data sources

DatasetPublisherWhat we take from itLicence
Price Paid Data (PPD)HM Land Registry Every registered residential sale in England & Wales: price, date, address, property type, new-build flag, tenure, transaction category Open Government Licence v3.0
Energy Performance of Buildings RegisterMHCLG Floor area, current/potential energy rating and score, certificate date Published open-use terms
ONS Postcode DirectoryONS / Ordnance Survey Postcode → region assignment and coordinates for maps and radius search Open Government Licence v3.0

2. Which sales are counted — the filter

Headline statistics deliberately describe the ordinary residential market. From the raw Price Paid file we include a sale only if all of the following hold:

  1. Transaction category A (“standard price paid”). Category B — repossessions, buy-to-let portfolio transfers, sales at other-than-market value — is excluded.
  2. Price of £10,000 or more — excluding nominal-consideration transfers that are not market transactions.
  3. Property type Detached, Semi-Detached, Terraced or Flat. The “Other” category (land, blocks, commercial mixtures) is excluded.
  4. A postcode is present. Land Registry publishes a small number of sales (roughly one in five hundred) with no postcode. Without a postcode there is no postcode sector to place the sale in, so these sales are excluded from every chart, table and download. They remain in the public file, so an unfiltered reconciliation will count them.

The Pro data table and all downloads apply the same filter, so charts and exports always describe the same set of sales. Raw-file reconciliations must apply this filter first; unfiltered Land Registry totals will always be higher than ours.

3. How EPC data is joined

Floor areas (and therefore all £/sq ft figures) come from matching each sale to an EPC certificate at the same postcode. Matching is automated: addresses are normalised and compared with fuzzy string matching; the best-scoring certificate above a similarity threshold is accepted, with a stricter threshold where many candidate certificates share the postcode.

Coverage: across 2023–2026, about 92–93% of qualifying sales match an EPC. Metrics that need floor area are computed over matched sales only; transaction counts and price totals are computed over all qualifying sales. These denominators differ — that is by design, not an error.

4. Derived metrics — exact definitions

MetricDefinition
Average priceArithmetic mean of sale prices in the slice
Median priceMiddle value (mean of middle two for even counts)
Median £/sq ftEach sale's price ÷ its EPC floor area, over matched sales; the middle value when sorted (mean of the middle two for an even count). Every figure labelled "Median" on the site is this per-sale median. A median is unaffected by a few extreme sales or floor-area errors, which is why it is the headline rate rather than an average.
Area-weighted average £/sq ftSum of prices ÷ sum of floor areas, over matched sales. The public dashboards show the median only; the Pro dashboard has an "Area-weighted average" switch that shows this figure instead, everywhere £/sq ft appears, with the labels changed to match. Also in the API as avg_price_per_sqft.
Floor areaEPC total floor area, converted at 10.7639 sq ft per sq m
New build vs resaleLand Registry's new-build flag on the sale record
Trend lines (rolling 3-month median)Each month's point is the median over every sale in that month and the two before it, pooled: £/sq ft over matched sales, price over all qualifying sales. A single month's figure is never plotted on its own.
Price growthThe last 3 months' median £/sq ft compared with the same 3 months one year earlier, expressed as a percentage change. Shown only where both windows contain at least 30 sales with a floor area; a median over fewer sales moves in steps and its growth would be noise.
Growth trendOn a region, town or borough view, a line for each of the last 12 months showing the rolling 3-month median £/sq ft against the same month a year earlier, for the area, its region and the country. Points with fewer than 30 sales in either window are left blank.
Sales volume by yearCounts and values for the last three full calendar years plus the current year to date; the window rolls forward each January.
Top 5 transactionsFor a neighbourhood or radius search: the five highest prices paid within the last-3-months window, with property type, floor area and £/sq ft.

5. Geography

6. Update schedule

ProcessWhenWhat it does
Weekly syncEvery Monday, ~01:00 UK Re-reads the current and previous calendar year of Price Paid Data and brings our copy into line with it: new registrations are added however late they arrive, records Land Registry has amended are corrected, and records it has removed are removed. Aggregates are refreshed for every month that changed.
Quarterly update1 Jan / Apr / Jul / Oct The same process over every year since 2023, so amendments to older history are applied too.
Dashboard cacheUp to 7 days Popular views are pre-computed; a figure can therefore lag the database by up to a week. The “last update” note on each dashboard shows the data month actually displayed.
Data releaseShown on each dashboard Land Registry republishes its files monthly. The footer of each dashboard names the release date of the file our data reflects, and when it was synced, so a reconciliation can be made against the same release.

7. The registration lag — read this before judging recent months

A sale enters Land Registry data only when the registration completes, not when contracts complete. The gap is typically several weeks and can be months where applications queue. Two practical consequences:

  1. The most recent month is always an undercount — commonly by a substantial fraction — and will be revised upward by later weekly syncs.
  2. Dashboards anchor to the last complete quarter for headline figures, precisely to avoid presenting a half-filled month as a market fall.
Official sources for the current lag: HM Land Registry publishes its processing times at gov.uk/guidance/hm-land-registry-processing-times, and each monthly Price Paid release notes the share of late-arriving records. We review these with each quarterly update.

8. Known limitations

9. Verifying our numbers

Counts and price totals are exactly reconcilable against the public Price Paid file once the section-2 filter is applied: filter the raw file, group by month and area, and compare. Two automated reconciliations run after every weekly sync: one recomputes every dashboard figure from our raw transaction rows, the other recomputes the counts and totals from the public file itself and compares them with the live dashboards. The script for the second check is published with the site's source code, so the same test can be run by anyone. The only expected difference is registrations Land Registry has published since the release date shown on the dashboard. Discrepancies can be reported to info@realdatawarehouse.com — please include the view, the filter state, and the figure you expected.