Type the words into a search box and you get input boxes. Year, make, model, trim, mileage, ZIP code, a row of condition buttons, and then a number, usually with a range around it and a button underneath offering to introduce you to somebody who will buy the car.

The number is not a lie. It is also not a fact about your car. It is the output of a model that has never seen the vehicle, fitted to transactions in cars that resemble it, adjusted by a set of rules the operator does not publish. Used carefully it is genuinely useful. Used as a verdict it costs people money, because the two commonest ways to be wrong about a trade-in are both ways of taking an estimate too literally.

This page cannot give you a figure. We have no auction feed, no listing archive and no transaction file, and inventing one would be worse than useless. What can be done, and what nobody seems to do, is take the estimate apart: name each input, say what the model is doing with it, say which of them you can independently verify before anybody appraises anything, and be specific about the two that no estimate can carry at all.

The best place to start is the input that looks most objective and is quietly the least anchored.

The mileage adjustment, and the average nobody publishes

Every valuation model adjusts a car up or down for its odometer reading. To do that it needs an expectation — a figure for what a car of that age ought to have on it — and it then prices the distance between the expectation and the reading. That expectation is the hinge of the whole adjustment, and no consumer estimator states it.

The market has filled the gap with folklore. Ask around and you will be told twelve thousand miles a year, or fifteen thousand if the person is thinking of a commute. Neither number has a source attached to it. Both are quoted with total confidence, and both are used by owners to decide, before they have opened a browser, whether their car counts as high-mileage.

There is a measured answer. The Federal Highway Administration publishes Table VM-1, Annual Vehicle Distance Traveled in Miles and Related Data, which divides total vehicle-miles travelled by registered vehicles for each class of vehicle, every year. It is not a survey of drivers and it is not a manufacturer’s claim. It is fleet arithmetic, published year after year, and it is the only baseline anywhere in this process that a reader can go and check for themselves.

For report year 2024, the most recent published, the average light-duty vehicle in the United States covered 10,787 miles. Not twelve thousand. Not fifteen. The figures below were retrieved on 7 September 2026 from the Department of Transportation’s open-data mirror of the table; the published page for that report year carries the same rows and was itself last updated in February 2026.

The split between the two light-duty rows is a wheelbase test rather than a body-style one. The table’s own notes put short wheelbase at 121 inches or less — passenger cars, light trucks, vans and sport utility vehicles — and long wheelbase above that, which catches the larger saloons, vans, pickups and SUVs. If you own a full-size pickup, the row that describes your vehicle is the long-wheelbase one, and for report year 2024 it came out slightly lower than the short-wheelbase row, at 10,707 miles against 10,812. The intuition that big vehicles cover more ground did not survive contact with the arithmetic that year.

Read the combined line across years and the folklore looks worse still. In the twenty-five report years from 2000 to 2024, the combined light-duty figure never reached twelve thousand. Its highest was 11,946 miles in 2004 and its lowest 10,143 in 2020; the median across the whole span was 11,443. Twenty-one of the twenty-five years came in at or above eleven thousand, and the four that did not are all recent.

Three cautions belong with those numbers, and they are the reason this section is longer than a table.

The first is that FHWA revises the series. Its own note says the mileage data rest on the latest state reporting available and may not match previously published results, so a figure quoted here is a figure as published now, not a constant. That is a good reason to date it and a bad reason to ignore it.

The second is that the short- and long-wheelbase rows are not a continuous pair over time. Between report years 2006 and 2007 the long-wheelbase average jumps by 4,050 miles while the short-wheelbase average falls by 1,775 and the combined line barely moves, falling by 406. Two rows swapping several thousand miles in opposite directions while their total sits still is what a change in how vehicles are sorted looks like, not a change in how far anybody drove. So the split is safe to read within a single report year and unsafe to read across them, and the line to follow over time is the combined one.

The third is the most important, and it cuts against the use you are probably itching to make of this. The figure is a fleet average across every registered light vehicle in the country, of every age, including the second car that leaves the driveway twice a week and the fifteen-year-old pickup that only tows. It is not a model-specific expectation and it is not a per-driver figure. A car that is somebody’s only car, driven to work daily, will run above it and is not thereby abnormal.

What the number is good for is calibration. It tells you that the baseline in your head is high, that a car doing eleven thousand a year is an ordinary car rather than a hard-used one, and that the word “high-mileage” is being applied in the used market to vehicles that are, by the only published measurement, average.

Where your own reading actually sits

The practical form of that average is a ladder. Take the report-year 2024 figure of 10,787 miles and lay it out by vehicle age and you have the reading a car would carry if it had done exactly the national average every year of its life.

Five years gets you 53,935 miles. Ten years gets you 107,870. Those are not thresholds and they are not predictions; they are what the multiplication says, and the multiplication is the whole of it. But they are the first honest reference point most owners have ever had for the sentence “my car has high miles for its age”, and they land a long way from where people expect.

Run the folklore through the same ladder and the gap becomes concrete. A driver who assumes twelve thousand a year expects 60,000 on a five-year-old car, which is 6,065 miles more than the measured average produces — a difference of 11.2 per cent. A driver who assumes fifteen thousand expects 75,000, which is 21,065 miles clear of the average, or 39.1 per cent. Carry both out to ten years and the twelve-thousand assumption is 12,130 miles high and the fifteen-thousand assumption 42,130 miles high, at 120,000 and 150,000 respectively.

A gap of 42,130 miles is not a rounding error. It is the distance between two very different-looking cars on a listing page, and it is entirely an artefact of which unsourced rule of thumb somebody happened to inherit.

Two things follow for anyone reading an estimate.

The first is that the mileage penalty you are bracing for may not exist. If your reading sits near the ladder for the car’s age, you are not selling a high-mileage car, whatever the phrase means to you, and you should stop negotiating against yourself before anybody else does.

The second is the reverse and is worth more. Where the reading sits well below the ladder, that is a genuine, checkable, above-average fact about the vehicle, and it is one of very few facts in this whole process that works in the owner’s favour and can be evidenced on the spot. It is also the fact most likely to be quietly discounted if you cannot show the odometer history — a low reading with no supporting record is a low reading a careful buyer has to treat as a claim.

Which is the point at which mileage stops being arithmetic and becomes evidence. A reading is only worth something if the record behind it agrees, and readings enter the federal record at titling and inspection events rather than continuously, so a clean mileage history is a weaker statement than most people assume. That distinction has a page of its own on this site and it belongs in your preparation, not in your negotiation.

The estimate assumes the cheaper car

The second input worth interrogating is the one that looks like a dropdown and is actually a fork in the road.

An estimator asks for trim, and it asks in the way it does because the base figure it is about to look up is keyed to year, model and trim before anything else happens. Choose the wrong entry and every adjustment afterwards is being applied to the wrong starting point. There is no recovery from that further down the form.

Two failure modes account for most of it. The first is straightforward under-selection: an owner picks the trim name they remember rather than the one on the build, or picks the base entry because the list is ambiguous and the base entry is first. The second is subtler — a car whose descriptor resolves to a group rather than a single trim, which is common enough that a federal decode will sometimes return several trim names for one vehicle. Where the decode returns a group, an estimator has to pick something, and the safe assumption from its point of view is the cheaper one.

Factory equipment sits on top of that and behaves differently again: some of it survives depreciation and some of it was money spent once and never seen again. That is a subject in its own right and we have treated it separately in our guide to how trim level and options affect trade-in value, which sets out which categories of equipment hold up and which quietly do not. What matters here is narrower: the estimate is only as good as the specification you fed it, and the specification you fed it is the one thing on the form you can prove with a document.

Condition tiers are a reconditioning bill in fancy dress

Then the buttons. Excellent, very good, good, fair. Sometimes with a paragraph of definition behind each; sometimes with nothing at all.

It is worth being blunt about what those tiers are. They are not an aesthetic judgement and they are not a compliment. Each step down the scale corresponds to work somebody would have to do before the car could stand on a forecourt with a price on it: tyres, brakes, a windscreen, paint correction, a wheel refurbishment, an interior that needs more than a valet, a warning light that has to be diagnosed rather than cleared. The grade is a proxy for that spend. Nothing else.

Which explains the thing owners find so infuriating about the gap between an online figure and a real offer. The estimate was quoted at a grade you chose. The offer is quoted at a grade somebody else assigned after walking round the car with a torch, and the difference between those two grades is not an opinion to argue with — it is an invoice they can itemise.

The self-report problem is well known and we have covered how it distorts guidebook figures in our page on what a Kelley Blue Book value actually is. The estimator-specific version is worth stating separately, though, because it has a fix.

The fix is to grade the car as the reconditioning list rather than as a school mark. Walk it with a notebook and write down every item a dealer would have to spend money on before retailing it, in the order a stranger would find them. Tread depth on all four. Whether the brakes are near their next service. Every panel with a chip, scratch or dent, and every wheel with kerb damage. The state of the driver’s seat bolster, the headlining, the boot floor. Whether both keys exist. Whether any warning lamp is lit, and whether it has been lit long enough that you stopped seeing it.

That list is the grade. If it is short, the top tiers are honest. If it runs to a dozen lines, no button you press will make the offer match the estimate, and the useful move is to fix the items that cost you less than they cost a dealer and leave the ones that cost you more — which is nearly always the split between presentation and mechanical work.

One more caution about the tiers themselves. The definitions are the operator’s, they are not standardised between operators, and they are not the same scale an auction lane uses. Two estimators can return different figures for the same car, honestly, because “very good” does not mean the same thing on both. Comparing two estimates therefore only works if you graded identically on both, which almost nobody does.

Which market the number came from

The largest single reason an estimate and an offer disagree is that they are quoted in different markets, and the form rarely says which one you asked for.

There are broadly three answers to “what is this car worth” and they are genuinely different amounts. What a dealer will allow you against another purchase. What a private buyer would pay you. What the same car will be advertised at on a forecourt after it has been reconditioned, photographed and warranted. The third is the one owners see when they browse listings, and it is the one that has the least to do with the money they will receive, because between the second and the third sits the reconditioning, the weeks of standing, the cost of selling and the margin.

Underneath all three is a wholesale benchmark built from dealer-to-dealer auction transactions, which the trade can see and you cannot. The offer is derived down from that benchmark rather than up from anything, and our guide to the Manheim Market Report sets out, from the operator’s own documentation, how the figure is computed and what gets subtracted between it and you.

The practical instruction is short. Before you compare two numbers, establish that they are the same kind of number. An online instant offer is a bid. A guidebook trade-in figure is an estimate of what a dealer might allow. A guidebook private-party figure is an estimate of what a stranger might pay. A forecourt price is an ask, and an ask is not evidence of anything except what somebody hopes. Most arguments about valuations are two people holding two of these four and each assuming the other means theirs.

The part of the trade that never appears on the offer sheet

There is one component of a trade-in’s value that no estimator computes, no offer sheet prints, and no negotiation can grant or withhold, and it is frequently larger than the amount people spend an afternoon arguing over.

In most states, when a car is traded in against a purchase from a dealer, sales tax is charged on the difference between the price of the car being bought and the allowance given for the car being handed over, rather than on the full price. Where that rule applies, the trade does two jobs: it pays down the purchase, and it shrinks the amount that gets taxed. The saving is fixed by the state’s own tax code. It exists whether or not anyone at the desk mentions it.

It is not universal, and the exceptions are not obscure states. Some grant no credit at all, so that a trade-in there reduces what you owe and nothing else. At least one caps the credit at a dollar limit its legislature revises. Others allow it only against a new vehicle, or only on passenger cars. The state-by-state list, taken from a state treasury’s own published form and dated to the calendar year that form was issued for, sits on our page explaining how a car trade-in works, alongside the odometer disclosure you will be asked to sign. Tax rules change by legislative session, so read any such list as current to its stated date and check your own state’s published schedule before you rely on it.

What matters for reading an estimate is the arithmetic. Where the credit applies, the effective value of trading in is the allowance plus your state’s rate applied to that allowance — and that total, not the allowance, is the figure a private sale has to beat before selling privately leaves you better off. Working it out on your own numbers takes seconds and changes the answer surprisingly often.

What that panel is showing is the break-even: the point at or below which the convenience of trading in is free, and above which you are being paid something for the advertising, the strangers, the test drives and the title transfer you would have to handle yourself. Where a loan payoff exceeds the offer, it also names the gap for what it is rather than folding it politely into the next contract.

What you can check before anybody appraises anything

The reason to take an estimate apart input by input is that most of the inputs can be confirmed in advance, free, from sources that have no stake in the answer. Two of them cannot be confirmed at all. Knowing which is which is most of the value of the exercise.

Read down the third column and two rows have no answer in them. The condition grade cannot be confirmed because you are the one supplying it, and the regional adjustment cannot be confirmed because no operator publishes it. Those are the two inputs that move an offer away from an estimate without warning, and neither is knowable from a form.

Read down the fourth column instead and one entry recurs in different words: damage that was repaired and never reported, wear you have stopped seeing, work done outside a campaign. The record is a record of things somebody filed. It is very good evidence about what was filed and no evidence at all about what was not, which is the reason an inspection remains the last word on any car and the reason no estimate can substitute for one.

How to read an estimate you have been handed

Suppose you have the number. Here is the order in which to interrogate it.

Ask which market it is quoted in. Trade-in allowance, private-party, retail ask, or a firm bid. If the page does not say, it is usually the one that flatters the reader, and you should assume the least useful interpretation until you can tell.

Check the specification it used. Not the one you meant to enter — the one it printed back. Wrong trim is the single most common reason an estimate and an offer are hundreds apart, and it is the only error on this list that a document settles instantly.

Put the odometer reading against the ladder rather than against the number in your head. Decide whether the car is above or below the measured average for its age before you accept anybody’s characterisation of it, including your own.

Re-grade the condition as a reconditioning list. If the list is long, re-run the estimate a tier lower and treat that as the realistic figure. If the list is short, keep the evidence — receipts, records, both keys — because those are what stop a grader assuming the worst.

Pull the free federal checks before anybody else does. Recalls and title brands take minutes, cost nothing, and are the two findings most likely to take an offer down before you have said a word. Nothing in a report should ever be news to the person who owns the car.

Add the tax credit if your state grants one, and only then compare trading against selling. The comparison is not allowance against private price; it is allowance-plus-credit against private price, minus what the private sale will cost you in time and hazard.

Get a written offer. An estimate is a model’s opinion. A dated, written, no-obligation offer to buy is evidence, it can be carried to a second desk, and it converts every subsequent conversation from an argument about value into a comparison of bids. It is the only way to turn an estimate into a fact.

None of that produces a valuation, and it is not meant to. It produces something more durable: a number you understand the construction of, and a short list of the ways it could be wrong. An estimate you can take apart is worth several you cannot.

Common questions

Why will this page not give me a trade-in value?

Because we do not have the data that would make one honest. Producing a valuation requires transaction records for cars like yours, updated continuously, and this site holds federal vehicle and safety data rather than a market feed. A figure invented to fill the gap would look exactly like a real one, which is precisely what makes it dangerous. What we can do is set out what a real estimate is built from, so you can judge the ones you are given.

What mileage is considered high for a trade-in?

Lower than most people think, and there is a measured reference. FHWA’s Table VM-1 puts the average light-duty vehicle at 10,787 miles for report year 2024, retrieved 7 September 2026, and the combined light-duty figure has not reached twelve thousand in any of the twenty-five report years from 2000 to 2024. A car near that pace is average, not high-mileage. Bear in mind it is a fleet-wide figure across vehicles of every age and use, not an expectation for your particular model.

Why is the dealer’s offer so much lower than the online estimate?

Usually two reasons stacked. The estimate was quoted at a condition grade you selected and the offer was quoted at one somebody else assigned after inspecting the car; and the estimate may be a guidebook figure while the offer is derived down from a wholesale auction benchmark after reconditioning, transport, holding cost and margin. Ask what the appraisal found and ask for the reconditioning estimate itemised. An itemised list is a conversation. A lump sum is not.

Does an estimator know about my car’s accident history?

Generally not, unless you tell it. Most consumer estimators price the model and your answers, not the vehicle’s record. The buyer will pull the record before quoting, so the history reaches the number at their end rather than yours — which is the argument for pulling it yourself first, so that nothing in it surprises you at the desk.

How much does the condition grade actually change the figure?

More than any other button on the form, which is why it is the one people get wrong. Each tier corresponds to a bundle of work a dealer would have to pay for before retailing the car, so a step of grade is a step of spend. The definitions are the operator’s own and are not standardised between them, so two estimators can disagree honestly on the same car.

Should I trade in or sell privately?

Do the comparison with the tax credit counted, not without it. Where your state credits a trade-in against the sales tax on the replacement, the effective value of trading is the allowance plus your rate applied to that allowance, and that is the number a private buyer must beat. Then subtract what the private sale will cost you in advertising, viewings, payment risk and paperwork. Where no credit applies, the private price advantage stands in full and the decision is simply whether the extra money is worth the work.

Do free online offers commit me to anything?

A no-obligation written offer commits you to nothing until you accept it, and its real use is as a floor rather than as a sale. A dated written figure can be carried to another buyer, which turns your car from something you have an opinion about into something you have evidence about. Read what the offer says about expiry and about re-inspection, because both are normal and both can change the number on the day.

Does a low odometer reading always help?

It helps when it can be supported. A reading well below the ladder for the car’s age is a genuine advantage, but a low reading with no service record and no odometer history behind it is a claim rather than a fact, and a careful buyer prices it as one. Very low mileage also carries its own maintenance questions on an older car, since seals, fluids and tyres age on the calendar rather than on the clock.

Can a VIN decode tell me what my car is worth?

No. The federal record describes how a vehicle was built and certified; it contains no price of any kind and never has. Anything returning a value from a VIN is running a valuation model over the decoded attributes, which is a legitimate thing to do and a different thing from looking something up. The decode is still worth running, because it settles the specification the valuation model is about to assume.