VIN Digits Explained: What Each of the 17 Characters Is Actually Worth

Type “VIN digits explained” into a search box and you will get the same diagram twenty times over: a seventeen-character string with brackets under it, labelled world manufacturer identifier, vehicle descriptor, check digit, model year, plant, serial. The diagram is correct. It is also, on its own, close to useless, because it tells you what each position is called without telling you what any of it is worth.

Those are different questions, and the second one is the one a person with a VIN in front of them actually has. If a character is hard to read on a weathered door label, does it matter which position it sits in? If a listing quotes you a number with one character wrong, will the decoder notice, or will it hand back a confident description of a car that does not exist? Which characters are load-bearing and which are just there?

You can answer that empirically, and nobody seems to. So we did. We took four vehicles with different manufacturers, different body types and different countries of assembly, and we changed their VINs one character at a time — every position, two different replacement characters each, recomputing the check digit after every change so the number stayed internally valid. Then we ran all 128 mutated numbers through the free federal decoder and counted, for each one, how many decoded values came back different from the original.

What follows is what that returned. Some of it confirms the diagram. Some of it contradicts the thing almost every VIN explainer on the internet says about the last six characters.

The seventeen characters, as the rule sets them out

Start with the law, because it is short and unusually readable. The VIN’s content is fixed by 49 CFR 565, and the rule does not leave much to interpretation.

Section 565.13 sets the frame. Every VIN is exactly 17 characters. A check digit is part of every VIN and sits in position 9. Every character must come from a permitted set: the letters A to Z minus I, O and Q, which is 23 letters, plus the 10 digits — 33 characters in total. Those three letters are excluded for the obvious reason, that I and O and Q are too easily confused with 1 and 0, and their absence is the cheapest fraud check available: a number containing one of them was misread or invented.

Section 565.13 also carries a line that most explainers skip, and it is the reason the number is as long as it is: no two vehicles built within a 60-year period beginning with the 1980 model year may carry identical VINs. Not thirty years, not the life of the car. Sixty. Everything about the tail of the number follows from that requirement.

Section 565.15 then divides the seventeen into four sections. The first is positions 1 to 3, which identify the manufacturer and the vehicle type. The second is positions 4 to 8, five characters that describe the vehicle’s attributes — and for passenger cars and for lighter multipurpose vehicles and trucks, the rule specifically requires that position 7 be a letter. The third section is position 9 alone, the check digit. The fourth is positions 10 to 17: model year, then plant of manufacture, then the rest.

That is the grammar, and our guide to reading a VIN off the car position by position walks through it as a reading exercise. This page is about what happens when you push on each position and see which ones push back.

One character at a time

The method matters, so here it is plainly. Our four subjects were a 2018 Toyota Camry, a 2013 Ford F-150, a 2016 BMW 328i assembled in South Africa, and a 2015 Land Rover Range Rover built in England. Four manufacturers, four descriptor schemes, four countries between them, and a spread across passenger car, light truck and multipurpose passenger vehicle.

For each vehicle we walked positions 1 through 17, skipping position 9 because it is not free to vary — it is computed from the other sixteen. At each position we substituted two different permitted characters, recomputed the check digit so the resulting number remained arithmetically valid, and decoded it. Then we compared the returned values against the unmutated decode, field by field, and counted the differences.

Two details keep the exercise honest. First, we verified our check-digit implementation against the worked example printed in the rule itself: the sample calculation in 565.15(c) sums to 411 and yields a check digit of 4, and our code returns 4. Anything downstream of a wrong check-digit routine would have been noise. Second, we compared decoded values rather than counting fields that merely looked populated, because the decoder returns a great many fields whose contents are the phrase “Not Applicable” and those are not information.

One preliminary finding fell out before the experiment proper. All four of our starting numbers — sample VINs of exactly the kind that circulate in documentation and test suites — failed their own check digits. Every one of them. Each stated a digit in position 9 that the federal formula does not produce for the other sixteen characters. That is worth sitting with for a moment: the numbers people use to demonstrate VINs are frequently not valid VINs.

Positions 1 to 3: all or nothing

The first three characters are the world manufacturer identifier, issued to the legal entity that certifies the vehicle. They are the most consequential characters in the number and they behave unlike any others.

Across our four vehicles we made 24 substitutions in this block. Not one of them produced a small change. 20 of the 24 emptied the decode entirely — no make, no model, no model year, no engine, nothing at all, returned with an error saying the manufacturer is not registered with the agency for sale or importation into the United States. The remaining four landed on a prefix issued to somebody else and came back as a different company altogether — a trailer builder, a truck-body shop — or as the same manufacturer stripped of its model.

On the BMW and the Range Rover, every single substitution across all three positions wiped the decode. That is the behaviour to internalise: the manufacturer identifier is not a description that degrades gracefully when you get a character wrong. It is a lookup key. It either hits a registered entity or it hits nothing, and there is no partial credit.

Practically, this is the best error-detection in the whole number, and it costs nothing. If you type a VIN into the federal decoder and get back an empty result with a “not registered” message, the overwhelmingly likely explanation is that you misread one of the first three characters — not that you have discovered a grey-market import.

It is also why the first character is a weaker geography clue than people assume. The register is full of codes whose issuing region and actual assembly country diverge, which our model-year and country chart takes apart properly. Our BMW is a case in point: a German prefix on a car the register records as built in South Africa.

Positions 4 to 8: the five characters that carry the car

This is the descriptor block, and it is where the decode earns its keep. The rule tells manufacturers what these five characters must make decipherable — for a passenger car, the make, line, series, body type, engine type and every restraint device and its location — but it explicitly leaves the encoding to the manufacturer. Each maker designs its own scheme inside the five characters and files the key with the agency.

Because the schemes are private, the yield per position varies enormously between manufacturers, and our substitutions show it. Every one of the 40 changes we made in this block moved something; none was inert. But the size of the move ranged from a single altered value to 62.

The pattern is legible once you stop expecting consistency. On the Camry, one character in position 5 rewrote 48 values and a character in position 8 rewrote 61, because Toyota packs a great deal into those slots. On the same car, one substitution in position 7 changed exactly one value. On the BMW, positions 4 through 8 behaved almost uniformly — 23 values changed, over and over, whichever position we touched — which is what a densely coupled scheme looks like from outside: pull any thread and the same block of attributes comes loose.

The lesson for a reader holding a number is not that position 5 means engine and position 7 means restraints. It is that no such universal table exists, that any site printing one is describing one manufacturer’s scheme and implying it is the standard, and that the only reliable way to turn these five characters into a specification is to decode the actual number against the manufacturer’s filed key. That is exactly what the free federal decoder does, across 144 published variables.

It also explains a trap covered on our page about how many VIN characters you actually need: because these five carry so much and vary so much, a number that is short or wrong in this block can still come back looking authoritative.

Position 9: the character that carries no information at all

The ninth character is the strangest thing in the number, and the experiment makes its strangeness concrete.

Position 9 is a check digit. It says nothing about the car. It is arithmetic performed on the other sixteen characters: each character is converted to a value by a table in the rule, each value is multiplied by a fixed position weight, the products are summed and divided by 11, and the remainder is the check digit — written as X when the remainder is 10. The weights are fixed by the rule and sum to 89. Position 9 itself carries a weight of 0, which is what allows the digit to be computed without reference to itself.

Since all four of our sample numbers arrived with wrong check digits, we had a clean natural experiment available. We decoded each number as given, then decoded it again with position 9 corrected, and compared.

The result was the same on all four vehicles: the error code changed from a complaint that the check digit does not calculate properly to a clean pass, and not one decoded value changed. Not the make, not the model year, not the engine, not the plant. The count of populated values was identical before and after — 76 on the Camry, 36 on the F-150, 40 on the BMW, 37 on the Range Rover — and the contents were identical too.

So the check digit is genuinely orthogonal to the decode. It is a transcription test bolted onto the side of an identifier, and its whole value is that you can run it with no database at all, which is what our check-digit validator does in the browser. Change one character anywhere in the number and the arithmetic almost always breaks, which catches precisely the error people actually make.

What it emphatically does not do is establish that a vehicle exists or that the number belongs to the car it is attached to. A VIN copied off a real vehicle and stamped onto a stolen one passes the check digit perfectly, because it is a real VIN. Passing means the seventeen characters are internally consistent. It means nothing else.

Positions 10 and 11: the year, and the quietest character in the number

Position 10 is the model year, drawn from a fixed code table in the rule. It is a real information carrier and the substitutions confirm it: every change we made there moved something, in one case 58 values at once, because moving the model year moves the entire specification that hangs off it.

The code table repeats, which is the well-known trap, and the rule provides the tie-breaker in a place most people never look: for passenger cars and lighter multipurpose vehicles and trucks, if position 7 is a digit the model year belongs to the earlier half of the cycle, and if position 7 is a letter it belongs to the later half. That is why the rule insists position 7 be alphabetic on modern light vehicles — it is carrying a second job. Our model-year chart lays out both halves of the cycle.

Position 11 is the assembly plant, and it turned out to be the quietest informative character in the number. Across our four vehicles, substitutions there produced the smallest changes of any position outside the inert tail — frequently one or two altered values, typically just the plant city, state, country and operating company. That is the honest shape of it: the plant character tells you where the car was screwed together and, on most vehicles, nothing more.

The Ford was the exception, where a plant substitution moved 17 values. That is not the plant character secretly encoding the drivetrain; it is an artefact of how the manufacturer’s filed patterns are keyed, where a different plant code selects a different set of records. Worth knowing, because it is the kind of thing that makes a naive reader over-interpret a single character.

Positions 12 to 17: not a serial number, for most of the register

Here is the finding that overturns the standard diagram.

First, the uncontroversial half. Across our four vehicles we made 48 substitutions in positions 12 through 17. All 48 returned an identical decode. Not a single value moved at any of those six positions on any of those four vehicles. The federal decoder simply does not read them — which is consistent with the way it reports its own work, reducing every VIN it decodes to a shortened descriptor that stops at position 11.

That is the basis for the familiar claim: the last six characters are a production counter, unique to the car, invisible to the decode. And for a high-volume manufacturer it is true.

Now the other half. The rule provides a second arrangement for low-volume manufacturers, and it is not a footnote. Under 565.15(a), a manufacturer identifier may be six characters rather than three, occupying positions 1, 2 and 3 plus positions 12, 13 and 14. The rule gives a flag so you can always tell which arrangement you are looking at: a 9 in the third position always indicates a six-character identifier. When that arrangement applies, only positions 15, 16 and 17 are the sequential production number.

How common is that? We pulled the federal register of manufacturer identifiers across all nine vehicle types it recognises. It holds 12,997 distinct identifiers. 9,827 of them are six characters long. 3,170 are three. Every single one of the 9,827 six-character codes carries a 9 in the third position, and not one of the 3,170 three-character codes does — a rule with no exceptions in nearly thirteen thousand records.

So the six-character arrangement is not the special case. It is the majority case, by roughly three to one. It dominates exactly where you would expect small builders to cluster: of 9,622 trailer identifiers, 8,033 are six-character; of 1,578 motorcycle identifiers, 1,101 are; of 481 low-speed vehicle identifiers, 343 are. Even among passenger cars, 93 of 348 identifiers are six-character, and among trucks 104 of 314.

We tested the behaviour rather than trusting the rule. We built a valid number on a six-character identifier belonging to a small trailer manufacturer, confirmed it decoded to that company, and then ran the same substitution experiment on its tail. Positions 12, 13 and 14 changed the answer every time — each substitution returned a different company altogether, a run of unrelated small manufacturers, or nothing at all. Positions 15, 16 and 17 changed nothing, in every trial, exactly as positions 12 to 17 had on the four mass-produced vehicles.

The tail is not one thing. On a Camry, positions 12 to 17 are a counter. On a custom trailer, positions 12 to 14 are the back half of the maker’s name and only the final three are the counter. The character that tells you which world you are in is position 3, and the decode confirms it: the trailer came back with a manufacturer, a vehicle type and very little else — only a handful of substantive values, the rest of the response being the word “Not Applicable” repeated. A small builder files a small key, and the decode is correspondingly thin.

The tail must be digits, and how many tells you the weight class

The substitution runs kept throwing an error we had not gone looking for: invalid characters present. It only ever appeared when we had put a letter into the far end of the number, and tracking down where exactly it appeared turned into the most practically useful thing on this page.

The rule behind it is 565.15(d). The last five characters of a VIN must be numeric for passenger cars and for multipurpose passenger vehicles and trucks rated at 4536 kg — 10,000 lb — or less. For every other vehicle, the last four must be numeric. That is a real constraint on a real VIN, and it is absent from essentially every popular explainer.

The decoder enforces it, and enforces it precisely. We substituted letters into positions 12 through 17 on five vehicles, adding a 2012 Freightliner Cascadia to the original four to bring a genuinely heavy vehicle into the test.

Position 12 accepted a letter on all five, every time — consistent with the rule, which never requires position 12 to be numeric. Our BMW in fact carries a letter there natively. Position 13 split the group exactly in half: the two passenger cars rejected every letter we tried, and the heavy truck accepted every one. Positions 14 through 17 rejected a letter on all five vehicles without exception. We repeated the position-13 test with four different letters on each vehicle to make sure we had a rule and not a coincidence, and the split held perfectly every time.

Read that backwards and it becomes a diagnostic. Find the earliest tail position that refuses a letter, and the decoder has told you which side of the 10,000 lb line the vehicle sits on. Refuses at 13, it is a light vehicle. Refuses only at 14, it is not.

There is one honest wrinkle, and it belongs in the record rather than in a footnote. The Ford F-150 patterned with the heavy truck, accepting letters at position 13, even though its own decode reports a weight rating of 7,001 to 8,000 lb — comfortably below the threshold the rule keys on. On that vehicle the decoder is more permissive than the regulation. We are not going to invent an explanation for it; what we can say is that the enforcement follows the manufacturer’s filed pattern rather than the decoded weight class, and that a light truck may therefore be checked to the looser standard.

The usable version for a buyer: a letter anywhere in the last four characters of any road vehicle’s VIN is a genuine red flag. You have misread the number, or somebody has made it up. Handwritten VINs on bills of sale get this wrong constantly, because a 0 becomes an O and an S becomes a 5 the moment a number leaves a keyboard.

When the decoder calls a correct check digit wrong

One more result, because it will otherwise cost somebody an afternoon.

When the decoder raises the invalid-characters error, it also raises the check-digit error — and it does so on numbers whose check digit is arithmetically correct. We confirmed this on two vehicles by computing the check digit independently, with the routine we had already validated against the rule’s own worked example. In both cases the character in position 9 was exactly the character the federal formula produces, and the decoder reported it as not calculating properly anyway. It then went on to return the vehicle correctly.

The order of operations appears to be that an illegal character in the tail poisons the check-digit evaluation rather than being reported on its own. The consequence for anyone reading decoder output is worth stating flatly: when you see a check-digit complaint alongside an invalid-characters complaint, do not go hunting for a transcription error in position 9. Look at the last four or five characters for a letter that should be a digit, fix that, and the check-digit complaint will very likely disappear with it.

What no character carries

Everything above is about what the seventeen characters encode. The more important point is the boundary of the whole exercise.

A VIN is assigned at the factory and never changes for the life of the vehicle. That is its entire purpose: it is a constant, so that everything else can be indexed against it. Which means, necessarily, that nothing that happened after the stamping is in the number. Not a collision. Not a flood. Not a salvage or rebuilt or lemon brand on a title. Not an odometer roll-back, not a lien, not the number of owners, not whether it is stolen.

No amount of reading the digits recovers any of that, because it was never written there. Those facts live in records that are keyed to the VIN and held elsewhere — by state title agencies, by insurers, by auction houses, by the federal recall system. The number is the index. The history is the file.

The free federal decoder returns the build. The free federal recall lookup returns open safety campaigns against the vehicle. Both are genuinely free and both are worth running before you look at anything paid. But the title-brand chain, the reported-damage records and the odometer history are commercial data, and if you are about to hand money to a stranger for a used car, that is the point at which a paid VIN history report from CarCheckerVIN answers the question the digits cannot.

The digits tell you what the car was born as. Nothing in them tells you what it has been through.

Common questions

What does each digit of a VIN mean?

Positions 1 to 3 identify the manufacturer and vehicle type. Positions 4 to 8 describe the vehicle — model line, body, engine family, restraints — in a scheme each manufacturer designs and files with the agency. Position 9 is a check digit and describes nothing. Position 10 is the model year, position 11 the assembly plant. Positions 12 to 17 are the production sequence on a high-volume maker, but on a low-volume one positions 12 to 14 are part of the manufacturer identifier and only 15 to 17 are sequential.

Which VIN digits actually change what a decoder tells you?

Positions 1 through 11, excluding the check digit. We changed every character on four vehicles and measured it: all 80 substitutions across positions 1 to 8, 10 and 11 altered the decoded result, while all 48 substitutions across positions 12 to 17 left it byte-for-byte identical. The federal decoder reads the first eleven characters and stops.

Does the check digit tell me anything about the car?

Nothing. It is arithmetic on the other sixteen characters and carries no vehicle information at all. We proved it by correcting the wrong check digits on four vehicles: the decoder’s error code changed and not one decoded value did. Its value is that it catches typos without any database, which is real but narrow.

Why must the last characters of a VIN be numbers?

Because 565.15(d) says so. The last five characters must be numeric on passenger cars and on lighter multipurpose vehicles and trucks; the last four must be numeric on everything else. The federal decoder enforces this and returns an invalid-characters error when you break it, so a letter in the last four positions of any road vehicle’s VIN means the number has been misread or invented.

Are digits 12 to 17 always the serial number?

No, and this is the most common error in VIN explainers. It holds for high-volume manufacturers only. Where the maker holds a six-character identifier — which is 9,827 of the 12,997 identifiers on the federal register — positions 12, 13 and 14 are the second half of that identifier and only 15, 16 and 17 are sequential. We confirmed it behaviourally: on a six-character identifier, changing position 12, 13 or 14 returns a different manufacturer entirely.

What does a 9 in the third position of a VIN mean?

It means the manufacturer identifier is six characters rather than three, continuing at positions 12 to 14. The rule states it as an absolute, and the register bears that out with no exceptions we could find: all 9,827 six-character identifiers carry a 9 in the third position, and none of the 3,170 three-character identifiers does.

Can two vehicles have the same VIN?

Not legitimately. The rule forbids two vehicles built within a 60-year period beginning with the 1980 model year from carrying identical VINs. If you meet two vehicles with the same number, one of them is wearing a cloned VIN — and because a cloned number is a copy of a genuine one, it passes the check digit and decodes perfectly. That is precisely why the decode cannot be the end of your checking.

Which digit tells me the engine?

There is no single answer, and any site giving you one is describing one manufacturer’s scheme. Engine type is encoded somewhere in positions 4 to 8, but where, and in how many characters, is the manufacturer’s choice. In our tests one character in position 5 rewrote 48 values on a Camry while a character in position 7 on the same car rewrote one. Decode the actual number rather than reading a table.

If I get one character wrong, will the decoder notice?

It depends entirely on where. Get one of the first three wrong and you will almost certainly get nothing back at all — 20 of our 24 substitutions there emptied the decode completely. Get one of positions 4 to 11 wrong and you will get a confident, wrong answer. Get one of the last six wrong on a high-volume vehicle and the decode will not change, though you will no longer be describing that specific car.

Do the digits tell me if the car has been in a crash?

No. Nothing that happened after the factory is in the number, because the number never changes. Crash damage, title brands, flood history and odometer discrepancies live in separate records indexed by the VIN, not inside it. Run the free federal recall check first, then buy a history report if you are about to purchase.