Used-car price history analysis

Access and scope

Set CRAWLORA_API_KEY to your Crawlora key. Prefer the listed Crawlora MCP tools when connected; otherwise use the bundled scripts/crawlora.sh, which sends x-api-key to https://api.crawlora.net/api/v1. Read reference/endpoints.md for exact methods and parameters. Check application code as well as HTTP status; payloads are inside data. Stop on authentication errors, back off on 429, and retry a transient upstream failure once. Bound requests by the requested scope and credit budget. Preserve source IDs, URLs, source dates, and retrieval/crawl times separately.

Explain observed asking-price changes for a listing or matched vehicle cohort. Establish source, geography, make/model/year/trim, mileage/condition constraints, observation window, and price basis before collecting comparable records.

Resolve listings and their recorded events

1. Discover accepted source/make/model/trim filters through dataset facets, then search with a bounded page budget. Stored sources are carmax, autotrader, and carsdotcom. CarMax inventory is enumerable; the other two are best-effort ZIP sweeps whose geocoded area can approximate the search location rather than the seller's address. Keep those coverage differences visible.
2. Copy returned IDs in <source>:<source_listing_id> format into item/history routes. A VIN is a vehicle identity lead, not the dataset listing ID. Different sellers or relistings of the same VIN remain separate listing histories unless dates and seller evidence establish an explicit linkage.
3. Price history records the first observed price and observed changes, not a row for every crawl/day. Cars.com may include backfilled native price-history events, bounded by enrichment coverage. Keep snapshot_date separate from crawled_at: a backfilled event predates the crawl that obtained it.
4. An empty history means no indexed history was returned; one row means one recorded price, not proof the price never changed before observation began. Preserve event IDs/schema version and anomalous or duplicate dates for review.
5. Refresh selected listings through their source's live detail route when needed, resolving that route's own ID/URL from listing evidence. A failed or missing listing is unavailable evidence, not proof the car sold or a seller accepted a lower price. Do not substitute another vehicle with a similar title.

scripts/crawlora.sh /datasets/vehicle-listings/facets facet=source
scripts/crawlora.sh /datasets/vehicle-listings/search make=Toyota model=Camry source=carmax page=1 page_size=10
# Select an actual returned listing ID before:
# scripts/crawlora.sh "/datasets/vehicle-listings/price-history/$LISTING_ID"

Calculate comparable changes

For comparable positive prices, show change = later − earlier and percent change = change / earlier × 100, with both observed dates. Keep increases, reductions, and missing/invalid price values separate. Do not label a first-observed price as original MSRP or a price reduction as a negotiated saving. Do not interpolate daily prices or time-at-price beyond observed events.

Match year/trim/drivetrain, mileage, seller type, condition and accident/owner history when supplied, region, currency, and fees before comparing vehicles. Unknown attributes remain unknown. Report cohort sample size and distinct listing counts; two histories for one VIN are not automatically two independent vehicles. Return the price-event ledger, matched-comparable table, coverage/freshness notes, current-versus-stored discrepancies, and questions to verify with the seller. A research request does not authorise contacting dealers or purchasing a vehicle.