Data you can trace. Data you can trust.
Every data point in CarVector is reconciled against multiple authoritative sources. Every correction is logged. Every claim is traceable.
The problem with vehicle data
Most vehicle databases trace back to a single commercial dataset — assembled years ago, then resold, repackaged, and silently degraded ever since.
A copy of a copy
One aging commercial dataset, relicensed and repackaged for years.
No provenance
Nobody can tell you where a horsepower figure actually came from.
Never re-verified
Or when — if ever — a value was last checked against anything.
Silent drift
A 2019 entry might just be the 2018 entry with the year bumped.
We decided that was unacceptable.
The pipeline
Built like a research lab, not a database.
Vehicle data isn't a file you load once. We treat it as a continuously-verified dataset under active reconciliation: a team of specialized software workers runs around the clock, each verifying a different facet of the data against authoritative records, applying confidence thresholds, and logging every correction with what changed, what was trusted, and why. We don't ship a copy of someone else's data — we ship an evidence trail. The newest sources — owner complaints, manufacturer service bulletins, and federal investigations — are mapped to vehicles by year, make, and model; records that don't map cleanly are quarantined rather than guessed, so a bad match never reaches the API.
Continuous reconciliation
- Specification reconciliation just now
- Authoritative dataset cross-check moments ago
- Diagnostic code verification a minute ago
- Recall data refresh minutes ago
- Catalog enrichment cycle earlier
Coverage figures are real catalog counts; the panel illustrates the pipeline's continuous, always-on reconciliation.
How we think about data quality
Multi-source reconciliation
We don’t trust a single source. Specs are cross-referenced against federal regulatory datasets, manufacturer publications, and structured knowledge bases. When sources disagree — more often than you’d expect — we log a correction.
Correction-grade data
Every correction captures what changed, what we trusted, and why. Thousands of corrections and growing — each one a data point that would be wrong in a single-source database.
Complete recall coverage
Federal recall campaigns mapped to year/make/model and refreshed automatically — the safety history most vehicle APIs don’t ship at all.
Provenance on demand
Business and Enterprise customers access the full correction and source trail through the API. If you need to demonstrate where your data came from — for compliance, audit, or training-data requirements — we’re built for that.
Why this exists, in plain English.
All of this — the reconciliation, the correction trail, the recall coverage — exists because the vehicle data layer underneath the independent repair ecosystem is broken, locked away, or priced to exclude. Here's the whole story.
Why this exists