Our Data Methodology: Where Every Number on VinItel Comes From

Every number on this site comes from somewhere specific. This page documents where, how often it refreshes, what assumptions we made when integrating the sources, and — equally important — the things we deliberately don't claim because the underlying data can't support them.

VinItel is not a data provider. We aggregate from authoritative sources (NMVTIS, NHTSA, auction houses, OEMs) and present them in a single report. Where sources disagree, we surface the disagreement rather than pick a side.

What's in the database

As of the date this page was last updated:

• 16.2M unique vehicles indexed across 16 tables.
• 2.2M NHTSA complaints covering 1995-present model years.
• 212K NHTSA recalls with affected-VIN coverage where the manufacturer reported it.
• ~5M auction listings across Copart, IAA, Manheim, and ADESA, with photo coverage for most listings.
• ~50 U.S. states contributing title transactions through NMVTIS.

VIN decoding

The 17-character VIN is decoded against a combination of NHTSA's vPIC (Vehicle Product Information Catalog) reference data and our own historical pattern library for older or low-volume vehicles vPIC doesn't fully cover.

• Primary decoder: NHTSA vPIC API for current and near-current vehicles (1981+).
• Fallback decoder: internal pattern library for pre-1981 vehicles, specialty trims, and import vehicles vPIC doesn't return data for.
• Cross-check: where the report has access to a dealer-build sheet for the specific VIN, that takes precedence over decoded values for trim and option codes.

Title history (NMVTIS)

Title history comes from the National Motor Vehicle Title Information System (NMVTIS), the federal database every U.S. state DMV is required to report into under the Anti Car Theft Act of 1992. NMVTIS is administered by the American Association of Motor Vehicle Administrators (AAMVA).

What's covered:

• Every title transaction the state DMV processed.
• Brand history (salvage, rebuilt, flood, lemon-law, junk, etc.).
• Odometer disclosure at each title transfer.
• Insurance total-loss reports (insurer participation in NMVTIS is required).
• Junk-yard and salvage-pool reports (recycler participation required).

What NMVTIS doesn't cover:

• The current state's title-of-record before the brand was applied. Some states are slower to report into NMVTIS, so a freshly-issued title may not be in the database yet.
• Brands applied by states that briefly fell out of NMVTIS compliance (a small number, all currently compliant).
• Title transactions outside the U.S. (Canada, Mexico, anything imported on a bond title).

Accident and damage records

Accident records come from a combination of:

• Insurance claims reported to NMVTIS via insurer participation.
• Police accident reports in jurisdictions that share with our data providers.
• Auction damage codes when the vehicle later passed through a salvage auction (these are often the most detailed source).
• Body shop and collision repair reporting where the shop is part of an OEM certification network that reports back to the manufacturer.

Coverage is high but not exhaustive. DOT studies estimate roughly 60% of minor collisions never produce a police report or insurance claim, and those accidents will not appear on any report. We say so explicitly on every report.

Auction listings and photos

Auction data is ingested from public listings on the four major salvage and wholesale auction platforms in the U.S.: Copart, IAA, Manheim, and ADESA. For each listing we retain:

• Listing date, location, lot number, and sale price (where the listing was a completed sale).
• Damage description and damage codes assigned by the auction.
• The full set of listing photos — typically 8-15 per listing, covering exterior, interior, and undercarriage.
• Title brand assigned at the auction (often the most accurate single field for damage severity).

NHTSA recalls and complaints

Recall and complaint data is sourced directly from NHTSA's public complaint and recall APIs. We refresh both every six hours.

Recalls:

• All NHTSA-administered safety recalls since 1966.
• VIN-level affected-vehicle data where the manufacturer reported it (recalls before 2018 often lack VIN-level granularity).
• Repair completion data where the manufacturer reports it back to NHTSA.

Complaints:

• All NHTSA-recorded complaints since 1995, the practical lower bound of the database.
• Component categorization, crash involvement, fire involvement, injury and death counts where reported.
• Free-text complaint narratives (we don't surface raw narrative on public pages but do use them for component categorization and known-issue clustering).

Sales volume and per-10K normalization

Raw complaint counts are misleading. A model that sold five million units will accumulate more complaints than one that sold two hundred thousand, even if it's actually more reliable per vehicle. We use complaints per 10,000 units sold wherever sales data is available, and complaints per 1,000 of our database's observed vehicles where it isn't.

• Primary sales source: published OEM and trade association U.S. sales figures (NADA, Cox Automotive, manufacturer quarterly reports).
• Per-10K normalization applied on every page where both sales and complaint data are present. Pages that can't normalize (older models, low-volume models, models without sales-data coverage) fall back to per-1K of observed vehicles with that fact disclosed in the data tables.
• Per-10K is never mixed with per-1K on the same chart — comparing the two would be apples-to-oranges.

Refresh cadence

• NHTSA complaints and recalls: every 6 hours.
• NMVTIS title and brand data: daily.
• Auction listings: every 4 hours for new listings; completed-sale prices refresh weekly.
• Sales-volume reference data: monthly (data providers publish on monthly close).
• Statistical rollups (reliability rankings, complaint rates per 10K): recomputed every 6 hours alongside the underlying NHTSA refresh.

Sample-size minimums

Small samples produce noisy statistics. To prevent low-volume models from dominating "most reliable" or "worst" lists by accident, we apply minimum thresholds before a model can appear in rollups:

• Make-level pages: requires at least 100 vehicles in the database.
• Model-level pages: requires at least 50 vehicles in the database.
• NHTSA reliability pages: requires at least 100 complaints across the model's lifetime, or 50 per model-year for year-specific pages.
• "Most reliable" rankings: additionally require at least 50,000 U.S. units sold over the included model years.

What we deliberately don't claim

• We are not Carfax or AutoCheck. Some data sources those services have direct contracts with (specifically, some OEM dealer-service records and a subset of insurance carrier feeds) we don't have. We disclose this on the home/comparison pages.
• A clean report is not a clean vehicle. It means nothing bad has been recorded. We say this on every report.
• Market value is a range, not a price. The model can be off by 5-15% on any specific vehicle. We don't publish a single point estimate as if it were precise.
• Reliability scores reflect reported complaints, not measured defect rates. They correlate well with long-term ownership outcomes in our backtests but they are not a guarantee.
• We don't predict the future. Past complaint rates are descriptive. They are a good baseline but a new design can deviate.

Corrections and disputes

If you believe a record on your vehicle's report is wrong:

1. For title or brand errors, contact your state DMV. NMVTIS reflects what your state reports; the state DMV is the source of record.
2. For NHTSA complaint or recall errors, file directly with NHTSA at safercar.gov. Their data flows into our refresh on the next 6-hour cycle.
3. For auction listing or photo errors, contact the listing auction. Once the auction corrects, our next refresh picks up the change.
4. For anything that appears specific to VinItel's presentation (typos, miscategorization, formatting issues), reach us through the contact page.

We aggregate; we don't author. Source-of-record corrections propagate through our refresh cycle within hours, but we don't override authoritative data with private claims.