← Back to AuctionLens

How AuctionLens calculates a bid range

Last updated: 24 September 2026

Most price tools do one of two things: quote the auction house's own estimate back to you, or average whatever numbers they can scrape. AuctionLens does neither. This page sets out the principles behind the recommended bid range in the extension: what counts as evidence, what does not, and why the tool will sometimes tell you that no honest range exists.

1. Real sales, not asking prices

An estimate is the auction house's expectation, sometimes its hope. Estimates are routinely set low to draw bidders in, or high to please a consignor, and a lot that passes unsold at $10,000–15,000 tells you the market declined that price. Treating estimates as evidence of value is how confident-looking "recommended bids" get built from lots nobody bought.

AuctionLens therefore separates the two completely. Realised prices, what a lot actually fetched, are the only inputs to the range. Estimates, and the estimates of lots that passed, are listed so you can see what the houses asked, clearly labelled, and never allowed to shape the number.

2. Every price is verified at its source

A search-result snippet is not a reliable price. Wherever it can, AuctionLens reads each comparable's own lot page and takes the outcome from there: whether the lot sold, what it sold for, and in which currency. On platforms that label every past lot "Sold at Auction" whether or not it sold, the status comes from the record, not the headline. A price that cannot be traced to its source is not used.

Every price that contributes to a range is shown to you with a link to the listing it came from. Nothing in the number is hidden.

3. Like-for-like, or not at all

A search for "Gandharan Buddha head" returns stucco heads, life-size heads, books about Gandharan sculpture and news stories about record prices. None of those is evidence for a small schist head. AuctionLens filters comparables on the characteristics that actually drive value for the kind of object in front of you, and it is deliberately strict about it: a different material, culture, object type or scale means a sale is set aside.

Set-aside sales are still listed, marked "not counted", with the reason in plain words, so you can disagree with the tool if you know something it does not. What you will never see is a price silently included that should not have been.

4. When the evidence disagrees, so does the range

Sometimes the genuine comparables that exist are simply too far apart to describe with one figure: a modest example and a museum-quality one, sold years apart, in different rooms. A tool that averages them produces a range so wide it cannot be bid on, and one that hides the spread produces a number that looks precise and is not.

AuctionLens has a rule for this: when the sales it has found are not coherent enough to support a range, it declines to give one. It tells you the sales are too far apart, shows you the span, and lets you weigh each sale on its own. In our testing this happens on a meaningful share of lots, and it is the correct answer for them.

5. Confidence is stated, not implied

Every range carries a confidence level that reflects how many genuine sales stand behind it. A range built on a handful of sales says so. A range with broad, consistent support says that too. Where the market splits, for instance between what the major houses achieve and what regional houses and marketplaces do for the same kind of object, AuctionLens shows the two markets separately rather than blending them into a meaningless middle.

6. What the range is not

It is not an appraisal, a valuation for insurance, or a prediction of what a specific lot will fetch on the day. It is a disciplined summary of what genuinely comparable objects have sold for, with every input shown. Condition, provenance, the strength of the room and the reserve can move a result well outside any range. Verify independently before bidding.

7. Attribution checks

Separately from pricing, AuctionLens looks at the lot's images without reading the seller's description and compares what it sees with what the listing claims: the material, the technique, the period, whether the object is what it says it is. A concern is raised only when the two clearly disagree, and the visual evidence behind it is shown alongside, so you can judge it yourself and take it to a specialist.

8. How we keep it honest

The engine is run continuously against a bank of real lots with known outcomes, and every change is scored against it before it reaches you. If a change would make the tool more confident but less right, it does not ship.