Bibliothèque

NHTSA vehicle recall burden intelligence (agent-curated)

NHTSA vehicle recall burden intelligence (monthly)

Monthly recall-burden ranking of US-market passenger-vehicle makes from the official keyless NHTSA recalls API (universe enumerated via vPIC, model years 2021-2026). Campaign rows are deduped on NHTSA campaign number; affected models and model-year spans are collapsed; the raw Component string is reduced to a top-level taxonomy (STEERING:COLUMN -> STEERING); Consequence/Summary text is scanned for crash/fire/injury/death mentions and folded into a documented severity model (3*crash+5*fire+2*injuries+10*deaths+4*park_advisory, where park_advisory flags NHTSA 'do not drive' / 'park outside' notices); a 0-100 recall_burden_score = 50% min-max(log1p(n_campaigns)) + 50% min-max(log1p(severity_sum)) is ranked with deterministic tie-breaks into b1..b4 tiers. Phone numbers in free text are redacted; Remedy/Notes boilerplate is never stored. The score is a recall burden signal, not a defect rate: counts are not normalized by sales volume. Primary key: campaign_number. Cadence: monthly. Caveats: campaigns exclusively affecting pre-2021 model years are out of scope; recall data ticks as NHTSA publishes, so same-month re-runs may differ slightly.

• transport
• signals
• monthly
• rankings