Vehicle Reliability & Defect Scoring Methodology
A rigorous, transparent mathematical framework for normalizing federal NHTSA owner complaints, historical recall campaigns, active ODI investigations, and interquartile failure mileage distributions across US automotive fleets.
1. Data Provenance & Authority Separation
MazicBay maintains strict boundaries between official federal evidence provided by the US Department of Transportation and derived proprietary analytics engineered by MazicBay:
| Evidence Type | Primary Source | Analytical Scope |
|---|---|---|
| Owner Complaints | NHTSA (US DOT) | Federal ODI complaint records filed by vehicle owners. |
| Safety Recalls & Probes | NHTSA Federal Register | Mandated recall campaign IDs, consequences, and free remedies. |
| MazicBay Reliability Index | MazicBay Intelligence Model | Sales-normalized defect risk model (1.0 to 10.0 score). |
| Problem Clustering | MazicBay Semantic Engine | Rule-based automotive taxonomy mapping and confidence ratings. |
| Estimated Repair Costs | MazicBay ASE / OEM Benchmarks | Typical out-of-warranty labor & part replacement benchmarks. |
2. Mathematical Formulation of the Reliability Index
To prevent high-volume vehicle distortion (where a top-selling model like the Toyota Camry with 400,000 annual sales receives more total complaints than a low-volume luxury model with 20,000 sales), each risk dimension is strictly normalized into a bounded $[0.0, 1.0]$ factor:
MazicBay Reliability Index = 10.0 × (1.0 - Risk Score)
Calculated as: min(1.0, (Complaints / Sales × 10,000) / 40.0). If exposure data is unavailable, score is marked UNAVAILABLE.
Weighs critical safety failures: (Crashes × 3.0 + Fires × 4.0 + Injuries × 2.0) / Total Complaints.
Penalizes active federal safety recalls (min(1.0, ActiveRecalls / 4.0)) and open ODI defect probes.
Penalizes defects occurring early in vehicle lifespan (under warranty <36,000 miles) vs high-mileage wear (>100,000 miles).
3. Interquartile Mileage Percentiles (P25 / P50 / P75)
Simple arithmetic averages are heavily skewed by data entry typos and extreme outliers in federal complaint forms (e.g. 1 erroneous entry of 1,000,000 miles). MazicBay calculates robust non-parametric percentiles:
- Median (P50): The exact midpoint of reported failure odometer readings (50% of defects occurred below this mileage, 50% above).
- Interquartile Range (IQR = P25 to P75): Represents the core 50% distribution window where majority failures materialize.
- Sample Size Auditing: The exact count of odometer-verified complaint entries is stated on every model-year card.