Methodology

What the data is

We use three public datasets: safety-related owner complaints filed with the National Highway Traffic Safety Administration (NHTSA), NHTSA safety recall campaigns, and fuel economy figures published by the U.S. Environmental Protection Agency (EPA). Where available we also show NHTSA’s 5-star (NCAP) overall crash-test rating. Details and links are on the data sources page.

Coverage: model years 2005–2025, 32 makes sold in the United States. A model is included only if it has at least 50 complaints across all model years, which filters out rare vehicles and data-entry noise. The complaint data currently on the site runs through 2026-09-24.

Complaint Index

NHTSA does not publish sales figures, so complaint counts cannot fairly be compared between different models. Instead we compare model years of the same model with each other, which is exactly what a shopper deciding between a 2016 and a 2018 of the same car needs.

For each model year we compute a raw severity score:

raw = complaints × 1 + crashes × 3 + fires × 3 + injuries × 5 + deaths × 10 + “do not drive” recalls × 10

Newer model years have had less time on the road to accumulate complaints, so the raw score is divided by an exposure factor: the share of a model year’s lifetime complaints that has typically been filed by its current age. We measured this curve on 2005–2010 model years (about 355,000 complaints): roughly 10% of complaints arrive in the first year, 40% by year five, 85% by year ten and 98% by year fifteen. A one-year-old model year is therefore compared at about a fifth of a mature year’s exposure (we floor the factor at 0.2 so that brand-new years with a handful of complaints are not over-amplified). We then divide by a model-year factor: the industry-wide average of that exposure-adjusted score for the same model year across all makes we track. This removes era effects that have nothing to do with a specific model — recession-year sales dips (2009), the surge of airbag complaints filed against 2005–2008 vehicles during the Takata recalls, and changes in how many complaints NHTSA receives per year. The result is ranked among all model years of the same model and expressed as a percentile from 0 (fewest issues) to 100 (most issues). That percentile is the Complaint Index.

Verdicts

• Avoid — Complaint Index in the top quarter of the model’s years (75–100).
• Caution — index 50–74.
• Good — index 25–49.
• Best — bottom quarter (0–24).

A verdict is shown only when a model has data for at least 4 model years and the year has at least 5 complaints. Otherwise we show the numbers without a badge.

Two caps apply on top of the quartiles. Model years 2005–2009 can be rated Good at best, never Best: NHTSA’s complaint records for those years are less complete (fewer owners filed online, and complaints older than the files we load are not included), and the model-year factor deflates them, so a “quietest year” signal there is weaker evidence than for newer years. Any model year with eight or more recall campaigns that require a dealer visit (over-the-air software updates are not counted), or with any “Do Not Drive” or “Park Outside” recall, is rated Caution at best, whatever its complaint volume — a low complaint count does not offset serious or numerous safety recalls.

This is not a reliability rating. The index is based on NHTSA owner complaints and recalls. Complaints are self-reported and unverified, so they measure how often owners reported safety-related problems, not the probability that a specific car will fail. A year rated “Avoid” for a model with a low complaint volume overall can still be a dependable vehicle.

What “complaints” means on every page

A complaint is one report filed with NHTSA, identified by its ODI number. The complaint count shown for a model year — in the scorecard, the timeline, the FAQ, the meta description, the model hub and comparisons — is always the number of unique complaints for that make, model and model year. Hybrid and plug-in variants that NHTSA lists as separate models (for example “RAV4 Hybrid”) are counted on their own pages, not added to the base model. A single complaint can name several components; component breakdowns therefore add up to more than the complaint total, and every breakdown says so.

Component groups

NHTSA tags each complaint with one or more component codes (several hundred of them). We map these to 15 groups such as engine, transmission, electrical or airbags so problem patterns are readable. A complaint that names two components is counted once in each group but once in the total.

Recalls

Recall counts are distinct NHTSA campaign numbers that list the model year. Campaigns flagged by NHTSA with a “Do Not Drive” or “Park Outside” advisory are highlighted and weighted in the index.

Fuel economy

MPG figures are EPA combined/city/highway ratings. When a model year has several configurations (drivetrain, engine, hybrid), we show the median. Annual fuel cost is EPA’s estimate at 15,000 miles per year and the fuel price current when EPA published the data.

Generated text

Summaries, year notes and FAQ answers are drafted by a language model that is given only the numbers and complaint records for that page, then checked by an automated validator that rejects any figure not present in the source data. See the editorial policy.