How AutoArb works

1. Discover. Enabled sources are scanned on schedule, each read by its real access method: API, feed, browser or manual. Sources that forbid automation are marked manual-only and never circumvented.
2. Normalize. Every listing becomes the same canonical record: year, make, model, price, mileage, location and condition notes. Missing fields stay missing.
3. Economics. The engine adds price, fees, transport, the repair range and title costs into one all-in figure, then prices the outcome three ways: conservative, expected and optimistic.
4. Max acquisition price. Expected market value minus your target profit minus expected costs. This is the ceiling: bid above it and the deal stops working.
5. Confidence. Every figure carries its source, its method and its confidence. Where data is missing, the card says unknown instead of guessing.

• We don't fabricate vehicle history or title status. If a source publishes no VIN, the history stays unverified.
• We don't auto-bid or auto-buy. Capital decisions need a human.
• We don't present a heuristic market value as a fact. It is labelled an estimate, with its confidence, until real comps back it.
• We don't bypass CAPTCHAs, logins, rate limits or platform terms.

Where a claim came from decides how much weight it gets, strongest first:

1. Government or manufacturer records
2. Professional industry data
3. Forums
4. Community anecdotes
5. Heuristic or AI inference

The lower tiers can raise a hypothesis. They never override stronger evidence without a human review.