How the Frontdoor Estimate works

Every estimate begins with real, completed sales from HM Land Registry Price Paid Data — the official record of what homes in England and Wales actually sold for. We pull sales within roughly three quarters of a mile of your postcode, going back around eight years. Postcode coordinates come from postcodes.io (Ordnance Survey data).

We do not use asking prices. An asking price is what someone hoped for; a sold price is what someone paid.

Not every nearby sale tells us the same amount. Each one is weighted by three things:

• Proximity — weight falls away sharply beyond about a quarter of a mile, because streets differ.
• Recency — a sale's influence halves roughly every five years.
• Similarity — a sale of the same property type counts roughly twice as much as a different one.

We then take a weighted median rather than an average, so one unusual mansion or one repossession can't drag your figure around.

A sale from 2019 isn't worth its 2019 price today. We fit local growth from the sales we actually pulled — a log-linear fit of price against year — rather than assuming a national rate. That local rate is shown to you on your own estimate.

If you tell us what you paid and when, we uplift that by the same local growth rate and blend it with the comparable-sales figure. A recent purchase carries more weight (up to half the final baseline) than one from fifteen years ago, because your own transaction is the single most relevant sale there has ever been for your home.

This is where most online estimates stop, and it's the biggest source of error. We apply explainable adjustments — every one visible under "How we got this" — for:

• Bedrooms and floor area relative to a typical local home
• Overall condition
• Kitchen and bathroom age — condition is the single biggest factor most online estimates never see
• Bathroom-to-bedroom ratio, and how many are en-suite
• Open-plan kitchen/dining space, and whether there's also a separate snug
• Garden access — bi-folds, sliding doors or French windows
• Tenure, and remaining lease length where relevant
• Parking and outside space
• Work done since you bought, capped deliberately to avoid double-counting

We publish a range, not just a single number, because a single number implies a precision no model has. The range narrows as our evidence improves: we score the number of comparable sales, how recent they are, how tightly they cluster, how much you've told us, and whether we have your own purchase price. High confidence gives a ±5% band; low confidence gives ±16%.

Where we genuinely don't have enough evidence, we say so rather than guessing.

When a home we've estimated actually sells, we store the sale price next to the estimate as it stood before the sale — not one recalculated afterwards. That's what the accuracy figures at the top of this page are built from, and it's what drives changes to the model. Valuations you've been given by estate agents feed the same loop.

Published figures are aggregate and anonymous. No individual home is identifiable.

The Frontdoor Estimate is an estimate, produced by a computer that has never been inside your home. It is not a professional valuation, a mortgage valuation, or a survey, and it should not be the only thing you rely on when deciding what to list at. For that, get a local agent through your door.