Data, measured 14 September 2026, from the PropDetect analysed base

Auction guide prices against what actually sold

This page rests on 64 auction lots, 32 of them in the North East, and it says so first because the figures below are large enough to be surprising and the sample is small enough to move. Each lot has at least 30 matched sold comparables, a median 100. Beside every measure sits the same measure for the 2,956 ordinary listings in the analysed base with the same evidence, so the auction figure is read against something.

The guide price sits at a median 71.5% of the comparables' sold median, 28.5% below what actually sold, where an ordinary asking price sits at 97.4%. 46.9% of guides are 30% or more below sold; 10% of asking prices are. Modelled at the guide, 51.6% of lots stack against 16.9% of ordinary listings at asking. And the guide is not the price. Measured 14 September 2026.

Table 1: auction lots beside ordinary listings

Every row is measured across the 64 lots and the 2,956 ordinary listings unless a smaller count follows the figure. Price means the guide for a lot and the asking price for a listing.

The median guide in the sample is £70,000 (middle half £49,750 to £95,000) against a median comparable sold price of £92,950. Part of that gap is the work: the modelled refurbishment runs at 22.1% of the guide against 7.4% of an ordinary asking price, three times the share. The rest is the guide doing its job, which is to fill the room.

Table 2: at the guide, what stacks and at what discount

The 64 lots by the verdict the report gave them at the guide, with the median guide as a share of the comparables' sold median within each group.

The lots that stack at the guide are guided at 64.1% of sold; the ones that do not are guided at 89.5%. A guide a third below the comparables leaves room for the hammer to climb and the deal to survive; a guide a tenth below does not. Among the 33 that stack, the winning strategy is flip 16, buy, refurbish, refinance 14, hold 2, short let 1. The report solves the maximum bid at which each lot still stacks, with the auction fees inside the purchase costs; that number, not the guide, is the one to take into the room.

What the model adds for an auction lot: an assumption, not a finding

Auction fees are inside every auction figure on this page and they are seeds the report lets the buyer replace from the legal pack. Modern method: a reservation fee of 4.2% including VAT, minimum £6,000, on winning. Traditional auction: an administration fee of about £1,500. Either way, £400 for a solicitor to read the legal pack before bidding. When the listing does not say which terms apply, the dearer modern method fee is assumed. Fees that some houses add on top, such as a buyer's premium, are not seeded and must come from the pack.

How these figures are produced

A listing is an auction lot when the portal categorises it as one (64 of the 64) or the pipeline detects auction terms in the price text, description or agent (1). The guide is stored as a guide, never as an asking price. It is compared with the median sold price of the lot's matched comparables from HM Land Registry data, the same basis as the below-market page. Stacking, refurbishment and flip margin are modelled with the purchase at the guide and the seeded fees inside the purchase costs. Lots by region: North East 32, Yorkshire and The Humber 16, Scotland 5, West Midlands 4, South East 3, North West 2, South West 1, Wales 1.

How the numbers behind this page are produced, and what the method can and cannot see: does AI property valuation actually work, and the measured back-tests on the accuracy page.

What this data does not say

• It does not say what lots sold for. No hammer prices are in the base yet; every figure is at the guide, and the guide is the floor of the conversation, not the end of it.
• 64 lots, half of them in one region, chosen by the investors who analysed them. The medians are honest for this sample and will move as it grows; the quartiles are printed so the spread is visible.
• One lot in the base (65 auction-like rows in all) lacked 30 comparables and is not counted.
• Figures are as of 14 September 2026 and are refreshed from the committed query, never edited by hand.

One limit applies to every figure on this page: the properties were chosen by PropDetect users to analyse, so this is a sample of investor interest, not a survey of the UK market. Where our users search shapes what the numbers can say.

The bid ceiling for one lot is a question about that lot, its comparables and its work. Run your own property through PropDetect and the same figures arrive for that address in about three minutes. No card needed for the first analyses.