How Our Reports Work

The records behind a report

We import recalls, owner complaints, investigations, manufacturer communications and crash-test ratings from NHTSA. EPA/DOE supplies fuel economy ratings through FuelEconomy.gov; EIA supplies weekly fuel prices. The data sources page explains the coverage of each.

Imported source records are retained alongside the structured data used by the site. We match records to makes, models and years, account for differences in naming, and group complaints by component. Counts and comparisons are calculated from those records. Guides may use a dated data snapshot, identified in the article.

Read the counts with these limits in mind

• Recalls count campaigns. One campaign may cover several model years. It counts once in a model's total and once under each affected year. The sum of the year rows may therefore exceed the model total. A campaign can apply to only some vehicles in a model year; check the individual VIN.
• Complaints count reports filed with NHTSA. They are not verified diagnoses or a count of all failures. We do not have the sales, mileage and reporting data needed to turn them into a failure rate.
• Age affects the opportunity to report. An older model year has had more time to accumulate complaints and recalls. A lower total for a newer year does not establish better reliability.
• Service bulletins describe manufacturer guidance. A bulletin may help a technician identify a known symptom or repair procedure. It does not, by itself, establish recall coverage or a free repair.
• Investigations are inquiries into possible defects. An open case is not a final finding. Read its scope and status; a closed case does not by itself prove that a concern was resolved.

How we compare model years

Our two-year comparisons identify differences in EPA-listed engines, transmissions, drivetrains and fuel economy, along with NHTSA crash-test ratings and recall campaigns. They describe differences in the records, not every trim or equipment change a manufacturer made. Complaint totals are shown with an age caveat and are not used to pick a winner.

Some year-overview pages also divide complaints by the number of years since the model year. This is a simple annual average, not a correction for all age effects: reports do not arrive at a steady rate, and sales and mileage still differ. It is not a reliability score. We publish NHTSA's crash-test stars as rated, separately for each tested configuration; we do not create our own star ratings.

What our analysis adds

Research pages examine specific questions across the records, such as a recall that followed an earlier repair or a complaint pattern also described in service bulletins. Each study explains its selection rules, counts and limitations. A pattern can justify a closer check without proving a cause or predicting what will happen to your car.

Text review and data updates

AI-assisted summaries are drafted from a structured data package, checked automatically, and held for human approval. Those checks cannot establish that an interpretation is correct. The reviewer must compare the wording with the evidence. Our AI policy explains the process.

Data tables refresh through scheduled imports; editorial text has a separate review process. A table can therefore receive new records before an article's dated analysis is revised. Check the dates and scope shown on the page, and use NHTSA's VIN lookup or the manufacturer for a vehicle-specific recall check.

Found a mismatch or an unclear claim? Send us the page and the source record.