What open property data shows before a listing

Most of what a sourcing tool sells as proprietary insight is public, free and published by the government. The advantage is not access. It is knowing which fields matter, which records to exclude, and what the absence of a record means.

Which datasets are actually free?

All three are published under open licences and all three are updated on a published schedule. None of them requires a subscription, and none of them is the thing a portal is charging for — a portal sells asking prices, which are the one number in property with no evidential value whatever.

What Price Paid Data can and cannot tell you

It records the price, the date, the postcode, the property type, the tenure and whether the sale was a new build. It also carries a category flag marking transfers that were not at arm's length: repossessions, portfolio sales, transfers between related parties, and sales where the price was not the market price.

Excluding those matters more than it sounds. Leaving them in a street-level median drags it down, and every deal on that street then looks better than it is. A comparable set built without the exclusion is not conservative — it is systematically optimistic, in the direction that costs money.

What it cannot tell you is condition. Two identical terraces on the same street, sold in the same month, can differ by forty thousand pounds because one had been rewired and one had not, and the register records neither fact. Price Paid Data sets the range. It does not place a specific property inside it.

That matters downstream. Open market value is the denominator in true discount, so an error in the comparable set propagates into every figure computed from it — the margin, the Deal Score, and the walk-away price. A range from registered sales, with the non-arm's-length transfers excluded, is the most defensible starting point available for nothing.

Why the EPC register is the more useful half

The rating is the least interesting field on it. The register also publishes total floor area, which converts a price into a price per square metre — and a three-bedroom terrace can be 70 square metres or 110. That difference is the entire margin on a refurbishment purchase, and it is free, and almost nobody uses it.

It carries construction age band, which predicts the works: solid walls, no cavity, likely rewire, likely damp. It carries the assessor's recommendations, which are a costed list of what the property needs written by somebody who stood in it. And it carries the lodgement date.

How do you tell a prepared sale from a live one?

By joining the two datasets on the date. An EPC is a legal precondition of marketing a residential property, so a certificate exists because somebody intended to sell. If a certificate was lodged eighteen months ago and Price Paid Data records no sale of that address since, something went wrong — the sale fell through, the property failed to attract an offer, or the owner changed their mind.

That is a signal no portal publishes, because it is defined by the absence of a listing rather than the presence of one. It is also the single most useful thing in either dataset, and it is the reason the two are worth holding together rather than separately.

For a rented property the register does a second job. An F or a G rating is a compliance problem as well as an energy one, and a landlord holding one faces improving the property or registering an exemption. That is a reason to sell that has a date attached to it.

The limits, stated

• Registration lags completion. A sale completed last month may not appear for several months, so recent silence is not evidence of no sale.
• The register is not a condition survey. It records what an assessor observed for energy purposes, not what a builder would find.
• Neither dataset carries ownership. Who owns a title is a separate register with its own rules, and it is not free at scale.
• An address that appears in neither is not an opportunity. It is an address about which nothing is known, which is a different thing.
• Both are licensed, and the licence governs what may be done with them. Reading data you have no licence for is a problem created at the moment of taking it, not at the moment of publishing it.

Used carefully these datasets narrow a town to a list worth looking at. They do not value a property, they do not tell you whether anybody wants to sell, and treating a derived signal as a decision is how a sourcing business acquires a portfolio of properties nobody wanted for a reason.