How a report is produced

Last updated: October 21, 2026.

This page explains exactly how Should You Bid arrives at a number, and where that number stops being trustworthy. If you are about to bid real money on the strength of a report, you should read it.

Should You Bid does two things, and they are produced differently. A bid analysis ($3) reads one listing you have already found. A car search ($7, or three for $18) goes and finds listings you have not. The first half of this page is the analysis; the car search is explained further down.

The four figures, defined

A report gives four prices. They mean specific things, and the difference between them is the whole point.

• Fair from / Fair to — the range this car is worth in today's market, judged against sales of genuinely comparable cars. It describes the market, not you.
• Bid this — the most we would advise paying for this particular car, after subtracting for the risks the analysis found. This is the number to act on.
• Walk above — the point at which the purchase stops making sense on any reading, and you should stop.

"Bid this" can legitimately sit below "Fair from." That is not an error. It means the model is worth the money but this example carries unresolved risk: an unanswered question about the title, a service history nobody can produce, tyres a decade old. A car worth $70,000 with $20,000 of unknowns in it is a $50,000 bid.

All figures are hammer prices — what the bidding reaches, before the auction house's fee. Bring a Trailer adds 5 percent to the buyer, capped at $7,500. Budget for it separately; the report does not fold it in.

What the analysis reads

1. The listing itself. The seller's written description, the specification table, the title status, and the stated history.
2. Every comment on the thread. Not a sample. The comment thread is usually where a listing's real problems surface, because other bidders ask the questions a seller would rather not answer, and the questions that go unanswered are themselves evidence.
3. The photographs, up to seven, examined for what they show and for what is conspicuously missing. An absent undercarriage shot on a car of a certain age is a finding.
4. Comparable sales, drawn from completed auction results for the same model.

How comparable sales are chosen, and when they are refused

This is the part that decides whether a report is worth anything.

Comparables are matched on marque and model, and a set is rejected outright unless most results name that marque or model and the median model-year gap is twelve years or less. A near miss is not accepted as better than nothing.

When no good comparables exist, the report says so and prices from model knowledge with a deliberately wider range. It does not quietly substitute loose ones. This rule exists because the loose version produced a real failure: an AMG GT Roadster priced against a marque index full of G-Wagens and a 190SL returned $32,500 for a car that was worth about $70,000. Priced against eighteen actual AMG GT sales, the same analysis returned $70,000. An empty comparables section is an honest answer; a wrong one is worse than silence.

What a report cannot see

• Anything the seller did not photograph or mention. No one inspects the car. Undisclosed damage, a concealed repair, a tired gearbox or a failing differential will not appear in a report because they do not appear in the listing.
• Title and odometer fraud. We read what the title is claimed to be. We cannot verify it.
• The reserve. Where a listing has one, it is private. A report estimates the market, not the seller's floor.
• Who else is bidding. Two determined bidders routinely take a car well past any defensible figure. That is a fact about auctions, not a flaw in the estimate.
• Mechanical condition. A report is not a substitute for a pre-purchase inspection, and in most cases it will tell you to get one.

How often it is wrong

Automated analysis makes mistakes, including confident ones. The underlying listing can be inaccurate, incomplete or deliberately flattering, and a report reasons from it in good faith. Treat the figures as a well-researched second opinion that costs three dollars, not as a valuation you can take to a bank. A report is not an appraisal and is not prepared to any appraisal standard.

If a report looks wrong to you, it may be. Tell us at support@shouldyoubid.com with the listing, and we will look at it.

Which sites can be read

Bring a Trailer, Cars & Bids and most major auction and classified sites. Coverage changes as those sites do, and you are only ever charged when a report is produced.

How a car search is run

A car search answers a different question: not "what is this worth" but "which of these should I be looking at". It costs $7 because it reads the live market rather than one page, and opens every car it shortlists.

The brief

You describe the car in plain English and Should You Bid turns that into a brief — make, model, year range, budget, mileage ceiling, how far you will travel. That brief is shown back to you, editable, before anything is charged. This is deliberate: the expensive failure here is searching confidently for the wrong car, and a budget read as $35,000 when you meant $25,000 is not worth discovering afterwards. Your original wording is also passed through, so a preference the parser has never heard of still reaches the analysis.

What it searches

Two kinds of source, because they hold different cars:

1. Dealer and retail inventory, through an aggregator that covers CarGurus, cars.com, Carvana, Autotrader and others.
2. Private-party listings on Craigslist, which no aggregator can embed and where the cheaper and more interesting cars tend to sit.

Both are read with a real browser, because both serve nothing useful to anything else. Listings are then parsed, deduplicated across sources, and filtered against your brief before any judgement happens — a car well over budget or of the wrong model is dropped rather than explained away.

Facebook Marketplace is not searched. It serves a sign-in wall to anything without a session, and we are not willing to drive a logged-in account to get around that.

What it then produces

• A shortlist of five, chosen for value against the market's own quartiles and for a spread of price, mileage and seller type — because you are choosing between kinds of car as much as between cars.
• The market it was priced against: how many cars are in range, the median asking price, the middle half of the range, median mileage. These are arithmetic over the listings found, not estimates.
• The generation you are buying into: chassis code, engine code, where those model years sit in the production run, the faults those years genuinely have and roughly what they cost, and the recalls to check against a VIN.
• What you can stop worrying about. A model's most notorious fault very often belongs to the generation before or after the one you are buying. A report names those explicitly, because a seller will otherwise price one in and you will pay for a repair that cannot happen to your car.
• A read of each listing's photographs — what they show, and what staged dealer photography is arranged not to.

What a car search cannot see

Everything in "What a report cannot see" above applies, and two more things:

• Asking prices are not sold prices. A search reads what sellers want. What cars actually change hands for is a different number, and usually a lower one.
• No vehicle history report is read. We cannot see a Carfax or an AutoCheck, and a report never implies otherwise. Where history is what settles the question, it tells you which record to demand and what it would show — which is the right order anyway, since you run a history report on the one car you have chosen, not on five.

A search is refunded if nothing in range is found. A brief you have already paid for re-runs free.