Data Sources and Confidence

Carvocate combines public vehicle data, dealer-provided listing claims, user-entered information, and limited calculations. It keeps origins visible because no single source provides a complete vehicle history or purchasing decision.

• Missing information remains missing.
• Estimates and user-entered values are labeled.
• A lack of returned records is not proof that no problem exists.
• Carvocate does not manufacture government or dealer data.
• Manual market values are labeled as user supplied, even when the user names a provider.
• Source availability can change, and failed refreshes stay visible.

After a valid VIN is decoded, Carvocate automatically checks the supported NHTSA recall and complaint APIs, records the documented public-API limitations for investigations and manufacturer communications, attempts an EPA configuration match, and starts the nationwide market sample when vehicle identity is sufficient. Each source keeps its own returned, no-records, confirmation-needed, or unavailable state so one limitation does not erase successful results.

Source categories and authority

Confidence and provenance labels

Labels appear next to values and source states. They use text and icons so meaning is not carried by color alone.

Carvocate Market Estimate

When MarketCheck is configured, Carvocate requests up to 50 active used-vehicle dealer listings for one bounded provider sample. A United States query is called a nationwide active-listing sample; it is not every national listing and is not guaranteed to be geographically balanced. Regional searches use the selected ZIP and radius. The records contain asking prices, not completed transaction prices.

A decoded VIN with year, make, and a provider-supported model is enough to attempt the nationwide baseline. ZIP, seller price, dealer, and repairs are not required. If mileage is missing, Carvocate may produce a preliminary estimate from real comparable mileage distribution, widen the range, limit confidence, and label the result. It does not assign a default mileage. Adding mileage later refines a compatible saved sample locally without another provider call.

Carvocate requests one broad, deduplicated provider sample by make, provider-supported model, and year, then applies a local fallback ladder: exact configuration, strong configuration, same series, same model family, adjacent year, and broad provisional family match. Each tier has its own minimum count, effective weighted sample, strong-match share, weak-match limit, series integrity rule, and price-spread limit. The first tier that passes is used; listing volume alone cannot make a weak tier acceptable.

When VIN Series supplies a verified provider alias, Carvocate records the display model, provider model, market family, duty class, and weight series separately. The live provider taxonomy maps the displayed NHTSA identity Silverado HD with Series 3500 to MarketCheck model Silverado 3500HD. Silverado 1500 listings are never admitted to that heavy-duty family. A 2500-versus-3500 difference remains a disclosed, heavily penalized family fallback and cannot be presented as an exact-series match.

Confirmed trim and mileage are local refinement factors rather than provider filters. Evaluated listings are labeled Excellent, Strong, Broader, or Weak from deterministic weights based on series, year, trim, drivetrain, engine, body or cab, bed, mileage similarity, freshness, and field completeness. Higher and lower trims are not treated as equivalent. Configuration differences never receive an unsupported directional dollar adjustment.

Confidence evaluates effective weighted sample size, exact-series and configuration match rates, missing fields, mileage fit, price spread, outlier burden, freshness, geography, conditional-price flags, and fallback depth. Results are labeled Precise, Best available, Provisional, or Unavailable. Insufficient evidence suppresses the estimate; provisional evidence may support a cautious range without supporting an exact opening offer.

The visible range uses the weighted 25th-to-75th percentile of included asking prices and widens by separately disclosed fallback-tier, observed mismatch-share, and unknown-mileage components. With unknown subject mileage, the mileage allowance is derived from the comparable mileage interquartile spread and bounded from 3% through 8%; no default mileage is assigned. The midpoint is the weighted median. Results are rounded to the nearest $100 and remain estimates, not certified appraisals or guaranteed sale prices.

Data freshness

Source results include checked dates. Market reports distinguish retrieval age from the freshness of the listings inside the sample. When a saved report is opened recently or while offline, Carvocate may show cached local results from IndexedDB and label refresh failures. Cached values can differ from current source data. Dealer listings can change, disappear, or block retrieval after a prior check.

Users can retry live sources from a report, re-run a listing check, or edit details when a source cannot be refreshed. A market refresh can change as inventory enters or leaves the active sample. Carvocate classifies midpoint changes as negligible, small, meaningful, or major using both dollars and percentage, and explains meaningful changes without alarming over rounding noise.

Source status

Carvocate does not hard-code “operational” health claims. It reports the outcome returned by the current workflow and keeps individual source status separate from VIN, market-estimate, and Buyer Decision confidence.