How PropDetect works: the methodology behind every figure

PropDetect pulls up to 100 sold comparables around the property's postcode, built on HM Land Registry sold price data via PropertyData. Each comparable is scored on how well it matches the subject: same property type, similar bedroom count, distance, and floor area where known. Poor matches are excluded, and the report says how many were excluded and why, in plain English. A typical methodology line reads: "based on 22 recently sold properties near this 3-bedroom end-terrace house (829 sqft). 49 sales were excluded from the pool."

The concluded figure is a conservative trimmed median of the surviving comparables, cross-checked against the price per square foot those sales imply for the subject's floor area. Where the two disagree, the report shows both rather than hiding one. When the comparable pool is thin, the report widens its range and lowers its confidence instead of manufacturing precision, and it says so on the page.

The full comparable table is visible on every report: address, sold date, price, distance, floor area and whether each sale was used in the valuation.

Room by room, from the listing photographs and floor plan. Each room is judged on what is actually visible: a clean, recently fitted kitchen is costed at zero, a dated but functional bathroom is costed as a refresh rather than a rip-out, and a wreck is costed as a wreck. Rates are regional, calibrated against real builder quotes, so a bathroom in Bolton does not price like a bathroom in Brighton.

Energy efficiency works are costed separately from condition works, so an EPC upgrade plan is visible on its own line. Overheads such as scaffolding, a damp survey or a full rewire are added only when the evidence in the photographs or the EPC suggests them. A 15 percent contingency is always carried. Every figure is editable, and edits recalculate everything downstream.

On accuracy: two customers have compared estimates against real builder quotes on the record. One estimate of 14,800 pounds came in against an actual quote of 14,500 pounds. Another customer, who prices deals with builders regularly, reports estimates within 500 to 1,000 pounds of actual quotes across their deals. Early evidence from two customers is not a measured error rate, and we make no wider claim until one exists.

From lettings comparables matched on bedroom count and property type near the subject. The report shows two figures side by side: the market rent today, which is the median of those matched comparables, and a post-refurbishment estimate, which sits at the 65th percentile of the same comparable set to reflect a renovated, well-presented property letting above typical stock. The premium between the two is stated as a percentage rather than hidden.

The rent confidence score is derived from the evidence: how many genuinely comparable lets existed, how tightly they agreed, how far away they were, and how many were let agreed rather than asking prices. Thin or scattered evidence produces a low score and the report says why in a sentence.

When a deal fails at the asking price, PropDetect does not stop at no. It works backwards from the end value, the refurbishment cost, the finance costs and a target return to the highest price at which the deal would work, and states it: "This stacks at 107,500 pounds. That is 17,500 pounds below the 125,000 pounds asking price, a 14 percent reduction." A no with a number is a negotiating position rather than a dead end.

Confidence is shown per component, separately for the end value, the rent, the EPC match and the refurbishment, each with the count of evidence behind it. It is computed from the evidence rather than asserted: comparable counts, how closely the comparables agree, and distance all move it. A wide spread across few comparables produces a low score, and the report explains what drove it rather than presenting a bare percentage.