Data Sources & Methodology
Every reliability score on this site is calculated from publicly available data. Here's exactly where that data comes from and how we use it.
- Vehicle-Years Scored
- 801
- Recalls Ingested
- 5,071
- Complaints Processed
- 119,682
Data last updated
NHTSA Recall Database
The National Highway Traffic Safety Administration maintains a public database of every safety recall issued in the United States. We pull recall campaign numbers, affected components, consequence descriptions, and remedy details for every vehicle in our database. Recalls flagged as "Park It" or "Park Outside" receive higher severity weighting in our scoring formula.
Data Points Used
- Campaign number and date
- Affected component and system
- Consequence description (crash, fire, injury risk)
- Remedy status and instructions
- Park It / Park Outside safety flags
NHTSA Complaints Database
Vehicle owners file complaints directly with NHTSA when they experience safety-related problems. Each complaint includes the affected component, a narrative description, and whether the issue led to a crash, fire, injury, or fatality. We ingest every complaint for each vehicle-year and categorize them by component type for weighted severity scoring.
Data Points Used
- Component category (powertrain, brakes, electrical, etc.)
- Owner narrative description
- Crash, fire, injury, and death flags
- Mileage at time of failure
- Date of complaint filing
NHTSA Crash Test Ratings
NHTSA's New Car Assessment Program (NCAP) tests vehicles for frontal crash, side crash, and rollover resistance. We display these ratings on individual vehicle-year pages as supplementary safety context. Crash test ratings do not directly affect the reliability score but provide additional safety information for buyers.
Data Points Used
- Overall safety rating (1-5 stars)
- Frontal crash rating
- Side crash rating
- Rollover resistance rating
- Test variant details
Independent Repair Cost Data
We use independent reliability ratings and repair cost estimates from third-party automotive data providers. These ratings reflect real-world repair frequency and cost data collected from repair shops across the United States. This data contributes to the repair cost sub-score (30% of total score) and is applied at the model level.
Data Points Used
- Overall reliability rating (normalized to 0-5 scale)
- Annual repair cost estimates
- Repair frequency indicators
- Comparison to category average
Vehicle Sales Figures
We collect U.S. sales figures from publicly available automotive industry sources. Sales data is critical for normalizing complaint and recall counts — a model that sells 500,000 units per year will naturally accumulate more complaints than one that sells 20,000, even if both are equally reliable. Without sales normalization, high-volume vehicles would be unfairly penalized.
Data Points Used
- Annual U.S. unit sales by model
- Used for complaint-per-vehicle normalization
- Sourced from public automotive industry databases
How These Sources Combine Into a Score
Each vehicle-year receives a score from 0 to 100 based on four weighted sub-scores. When data is unavailable for a sub-score (for example, no repair cost data exists for a particular model), the remaining weights redistribute proportionally.
Owner complaints weighted by component type (powertrain issues weigh more than cosmetic ones) and normalized by sales volume.
Independent reliability rating converted to a 0-100 scale. Applied at the model level across all years.
Recall count weighted by severity. Park It recalls count 5x, Park Outside 3x, and standard recalls 1x.
Number of distinct complaint categories multiplied by the log of total complaints. More categories of problems signal systemic issues.
Data Limitations
- Complaint volume is not a perfect proxy for reliability. Owners of some brands may be more likely to file complaints than others. We mitigate this through sales normalization but cannot fully eliminate reporting bias.
- Recall data reflects manufacturer and regulatory response, not just defect rates. A proactive manufacturer may issue recalls for minor issues, while another may delay. Both behaviors affect scores.
- Repair cost data is model-level, not year-specific. The same repair cost rating applies to all years of a given model. In practice, older years may have higher repair costs than newer ones.
- Recent model years have limited data. Vehicles need at least two years on the road to accumulate enough complaints and recall data for a meaningful score. We exclude very recent years from our rankings for this reason.
See the methodology in action