Methodology

What is counted

Complaints are deduplicated by ODI number for each vehicle. All-model reports deduplicate the ODI number across included model years. Recall campaigns are deduplicated by NHTSA campaign number. Component counts can exceed complaint totals because a complaint can name multiple components.

Vehicle analysis explorer

Component heatmaps compare complaint counts and within-year shares across model years. Mileage ranges use linearly interpolated 25th, 50th and 75th percentiles, with at least five usable reports required for a range. Co-occurrence counts unique complaints naming both components.

Monthly charts use filing dates and retain ten years of history, including zero-count months. Spike screening considers the latest six complete months: at least 10 reports, at least twice the preceding 12-month average, and at least three times the square root of that average (with a floor of one) above the average. This is a descriptive rule, not a significance test or defect finding.

Equal-age curves start at calendar age zero (the model year) and compare the first three or five fully elapsed calendar years. Valid incident dates are preferred if no later than filing; filing dates are the fallback. Impossible or missing date ages are excluded. Vehicles with incomplete observation windows are excluded. This does not correct for sales, mileage exposure or reporting delays.

Narrative pattern cards publish their case-insensitive keyword rules and source examples. Terms can match negations or comparisons. Coverage badges count usable fields within unique complaints. Examples are redacted, limited selections and may be truncated; they are not representative samples.

Model years and dates

Recall model-year charts include campaigns announced at any time in the snapshot. Complaint timelines use filing dates. Source snapshots are dated and are not a live stream. The snapshot calendar year is incomplete.

Mileage

Charts use the structured failure-mileage field. Missing, zero, non-numeric and implausible values of one million miles or more are excluded. Bins are under 25,000; 25,000 to 49,999; 50,000 to 74,999; 75,000 to 99,999; 100,000 to 149,999; and 150,000 and above. The median averages the two middle values when the count is even. These are complaint frequencies, not risk per mile.

Topic matching rules

Case-insensitive keyword rules match complaint text. A report can match more than one topic. Negations and comparisons can match, so a transmission mention is not proof of transmission failure.

• Transmission and shifting: \b(transmission|gearbox|cvt|shifting|shift(?:s|ed)? (?:into|out of|gear)|torque converter)\b
• Engine operation: \b(engine (?:stall|fail|stop|seiz|misfir|overheat)|stall(?:s|ed|ing)?|oil dilution|gas(?:oline)? (?:in|mix(?:ed|ing)? with) (?:the )?oil|loss of power)\b
• Braking behavior: \b(brak(?:e|es|ing)|abs)\b
• Electrical and battery issues: \b(battery|electrical|wiring|short circuit|parasitic drain)\b
• Steering behavior: \b(steering|power steer)\b
• Fuel system concerns: \b(fuel pump|fuel leak|gas(?:oline)? smell|gas tank|fuel tank|injector)\b

AI summaries

AI summaries are labeled and linked to the ODI evidence they use. They summarize a specified set of narratives and do not establish causation. Summaries must be reviewed before publication. Where no AI summary is published, the site presents a calculated pattern summary instead.

Vehicle record profile rankings

Rankings compare each eligible vehicle only with other vehicles from the same model year. A model year needs at least 25 recorded complaints. The record concern index is 75% complaint-count percentile and 25% distinct-recall-campaign percentile. Lower values indicate a lower recorded concern profile. Overall lists use the median yearly index from 2015 through 2025 and require five eligible years. Counts are not adjusted for sales, registrations, vehicle age, miles driven or reporting behavior, so the index is not a reliability, quality or safety rating.

Search quality

Individual vehicle pages need at least 10 complaints or 2 recalls to be indexed. Problem pages need at least 10 complaints. All-model problem pages need 30 complaints and two recorded model years. Component pages need 10 keyword matches and at least three source examples. Model recall pages need two campaigns. Comparisons are curated, require overlapping years, and never assign unsupported reliability rankings. Sparse reports remain available but are excluded from sitemaps and marked noindex.

Limitations

Records cover U.S. reports with exact make and model names. Variant classification can be inconsistent. Missing reports do not prove the absence of problems. Without vehicle population, age and distance exposure, counts cannot establish failure probability. Verify safety and repair questions with the manufacturer or NHTSA.