The MyHomePrice methodology

What we believe

Online property estimates are broken: most "free valuations" are lead-capture forms, sold on to agents who call you for weeks. MyHomePrice does the opposite — a 100% online service, no agent and no middleman, built exclusively on sales actually registered with HM Land Registry. Nobody will call you. Your result is ready in a minute, and every figure is traceable.

The data: real sales, not listings

Everything rests on HM Land Registry Price Paid Data: the official record of residential sales in England & Wales — 29.5 million transactions registered between 1995 and 2026 (detached, semi-detached, terraced, flats; commercial "Other" records are excluded), refreshed with every official release. It is published as open data under the Open Government Licence v3.0 (© Crown copyright and database right). No asking price ever enters our calculations: only prices actually paid.

Algorithm 1 — Price prediction (trend extrapolation)

For your postcode sector — falling back to the postcode district, then the postcode area, whenever the local sample is too thin, and telling you when we do — we compute the median sold price for your property type year by year since 1995, then the compound annual growth rate (CAGR) between the earliest and latest year.

That rate is projected over 5, 10 and 15 years and applied to your property's value, framed by three scenarios — neutral, optimistic and pessimistic — anchored to real Bank of England and ONS paths for mortgage rates, CPI inflation and 10-year gilt yields, and refreshed automatically every month. Validated on held-out data, our district-level growth model is off by about 2.3 percentage points a year at a 5-year horizon — better than zero-growth and momentum baselines — while at 1 year it performs close to a zero-growth baseline: short-term moves are mostly noise, and we tell you so. When the local trend is strongly negative we say so too, and the optimistic scenario is floored at 0% growth rather than dressed up.

/prediction →

Algorithm 2 — Repeat-sales indexbeta

Median prices drift when the mix of what sells changes. To measure the true trend we build a repeat-sales index — the method behind the famous Case-Shiller index in the US: among the 29.5 million transactions we identify the pairs where the same property sold twice, exclude quick flips and aberrant price ratios, and run a Bailey-Muth-Nourse (BMN) regression to extract the pure, constant-quality annual movement.

Computed at full scale, the index rests on more than 10 million repeat-sale pairs fed into the BMN regression, and is published for the whole of England & Wales and for each of the 105 postcode areas, on a base of 1995 = 100 — the national index stands at around 438 in 2026. It is benchmarked against the official UK House Price Index to validate its behaviour, and remains a v1 (beta) release.

Algorithm 3 — Machine-learning valuationbeta

Our valuation engine is a machine-learning model trained on Price Paid Data enriched with floor-area and energy features from Energy Performance Certificates (EPC): of 23.6 million certificates, 18.2 million sales (61.8%) were matched to an EPC by address and form the training set. To serve results in under a second, the full gradient-boosted model is distilled into a compact valuation grid covering every local market segment.

Measured accuracy of the current model (gbr-uk-esti-v1): on a held-out most recent year of sales, the median absolute error is 12.7%, against 16.3% for a baseline that simply applies the local median price. That is a median — half of the valuations land closer, half land further — and the engine remains a v1 (beta). We publish our limits as readily as our results.

Nowcast correction. Registered sales reach Price Paid Data one to three months after completion, and our local benchmark blends the last three years of sales — so its effective date sits in the past. To close that gap, every valuation is index-adjusted: we compute the benchmark's volume-weighted effective date, then scale it by the movement of the official UK House Price Index (HM Land Registry/ONS, monthly, local-authority level) from that date to the latest published month. The correction is local — Manchester moves with Manchester's index, not a national average — clamped to a ±10% band to absorb single-month noise, and the index refreshes automatically every month. If the index is unavailable, the valuation falls back to the uncorrected benchmark.

/estimation →

AI survey & report analysis

Unlike the three algorithms above, this service is not statistical: it uses Claude, a large language model, to read the due-diligence pack you upload (PDF, up to 30 MB and 150 pages). The file is processed map-reduce style — read section by section, then consolidated — extracting every defect and its RICS condition rating (1/2/3), indicative repair cost ranges and negotiation points, each with a reference to the page it came from. Figures such as lease years, ground rent or EPC scores are transcribed exactly as written in your documents, never invented.

It is assistive by design: an AI reading can contain errors or omissions, and it replaces neither a RICS chartered surveyor nor your solicitor — only the original documents are authoritative. Your file is deleted from our servers as soon as processing is complete; only the result remains in your account. The service costs £8.99 per analysis, independent of the credit system.

/survey-analysis →

What we don't do

• ❌ No canvassing: your estimate triggers no agent calls.
• ❌ No data resale: your information never leaves MyHomePrice.
• ❌ No asking prices: only registered, completed sales.
• ❌ No promised returns: bounded scenarios, never a guarantee.

The essentials

MyHomePrice applies the methods of institutional indices — trend extrapolation, a BMN repeat-sales index benchmarked to the UK HPI, and machine-learning valuation — to the full public record of England & Wales sales since 1995, and delivers the result in a minute, with no middleman.

MyHomePrice estimates and predictions are provided for information only, from public data. Contains HM Land Registry data © Crown copyright and database right, licensed under the Open Government Licence v3.0. They are neither a surveyor's valuation nor investment advice.