VinUnpack
No uploadvPIC 2026-10

Data & sources · Guide

Where VIN Data Comes From: NHTSA vPIC Explained

A behind-the-scenes look at the public dataset behind the decoder: its tables, its coverage, its update cycle and its honest limits.

By KianPublished 2026-11-09Updated 2026-11-096 min read

What vPIC is

vPIC stands for Vehicle Product Information Catalog. It is a public dataset published by the National Highway Traffic Safety Administration. Manufacturers submit information about the vehicles they build, and vPIC organizes it into tables that map manufacturer codes, VIN patterns, model years and vehicle attributes.

vPIC is not a vehicle history database. It does not contain accidents, mileage or ownership. It is a catalog of what manufacturers say they built, expressed through VIN patterns and manufacturer identifiers. That makes it the right source for a decoder and the wrong source for a history report.

The snapshot used by VinUnpack

VinUnpack uses the vPIC 2026-10 snapshot, extracted on 2026-10-21. The decoder ships a compact static file rather than calling an API at runtime, so a VIN is matched locally in the browser. The trade-off is that the data is a snapshot, not a live feed.

TableRowsWhat it contains
country199Country and region names
make12,387Vehicle makes
manufacturer23,098Manufacturer names
wmi13,014World manufacturer identifiers
wmi_make14,191WMI-to-make links
make_model32,070Make-to-model links
element161VIN field definitions
wmi_vinschema41,810WMI-to-schema and year links
vinschema25,299VIN schema definitions
enginemodel346Engine model descriptions
enginemodelpattern2,510Engine pattern links
vindescriptor69VIN descriptor definitions
vPIC 2026-10 base tables used to build the site data

The shipped decoder file is smaller than the source tables. It contains 13,014 WMI entries, 6,834 VIN schemas, 96,668 Tier A/B patterns and 2,211 model lookups. The build script filters the raw tables to the patterns the decoder can actually use.

How a VIN becomes a decoded field

The decoder starts with the manufacturer code, which is the first three characters described in the VIN position breakdown, extended to six when the third character is 9 and a longer manufacturer code is required. It looks up the WMI to get the make, manufacturer, country and coverage tier.

Next it resolves the model year from position 10, using the 30-year cycle and the vehicle-type rule. Then it builds a pattern key from positions 4–8 and 10–17 and checks the schema patterns linked to that WMI and year range. The most specific matching pattern supplies the model, engine, plant city, state and country. This is the same pattern logic explained in why positions 4–8 are manufacturer-specific.

If a field has no matching pattern, the decoder returns Not decoded rather than a guess. If the make is Tier C, model and engine are marked Not supported in v1. The result card shows the tier so the coverage is visible, not hidden.

Coverage is strongest for the US market

vPIC is a US regulatory dataset. It covers vehicles sold in the United States well, including the WMI codes and patterns manufacturers submit. Non-US vehicles may still have a recognizable WMI and model year, but their model and engine patterns can be missing or expressed in a different format.

This is why VinUnpack splits makes into Tier A, B and C. The tier reflects the completeness of the public pattern data for that make, not the quality of the vehicle. A Tier C make can still decode to year, make, plant code and country; it just does not have enough pattern coverage for a reliable model and engine field.

A US-market vehicle can also have a gap when the manufacturer changed a VDS scheme mid-year or when a pattern was submitted after the snapshot date. The decoder reports the gap instead of borrowing a pattern from a different year or trim. The decoder treats those gaps the same way: it shows the fields it can support and labels the rest. It does not substitute a later model year's pattern for an earlier one, because that would produce a confident wrong answer.

How the data is updated

The snapshot is extracted from the published vPIC archive, curated into the tables above, and then built into the decoder's static JSON file. The build is reproducible: the extraction command and the build command are documented in the repository. When a new vPIC snapshot is published, the site can be rebuilt against it without changing the decoder logic.

The site does not blend in live third-party data, scraped listings or dealer feeds. A single, dated source is easier to audit. When the data is wrong or missing, the honest response is to update the snapshot, not to fill the gap with an assumption.

A dated snapshot also makes the result reproducible. If a decode looks wrong, the same VIN can be checked against the same snapshot and the same patterns, and the discrepancy can be traced to a specific table or pattern rather than to a live service that changed overnight. That is useful for debugging and for explaining to a user why a field is missing.

What vPIC cannot do

vPIC cannot tell you whether a vehicle was in an accident, how many miles it has, whether it was stolen or whether it has a lien. It cannot guarantee that a VIN belongs to a specific physical vehicle. It also cannot cover pattern data that manufacturers never submitted.

Within those limits, it is a strong public source: broad WMI coverage, manufacturer-submitted patterns and a transparent update path. VinUnpack's design choice is to show exactly that coverage and to name the gaps. What a VIN decoder can and cannot tell you explains the boundary in more detail.

Decode a VIN with the vPIC 2026-10 snapshot

Sources

Related guides

Have a 17-character VIN? Decode it free in your browser — no account, no upload.