Every competitor's price, every listing, every day. What used to be a market feeling becomes a sortable column.
Pricing a car against the market means knowing what the market is asking, and the market changes daily. A dealer inventory scraper reads the live listings on the sites your buyers actually shop — AutoTrader, Cars.com, CarGurus, CarMax, and their equivalents outside the US — and hands you the whole comparable set as data rather than as twenty browser tabs.
The uses stack up quickly once the data exists: pricing against genuine local comps instead of a gut number, spotting which competitor just cut $1,200 across their sedans, tracking how long specific trims sit before they move, and finding sourcing opportunities where a market is thin.
US and international listing platforms, marketplaces and individual dealer websites. Individual dealer sites matter more than people expect — a competitor's own inventory page is often the fastest signal that something has been repriced.
One row per listing: VIN, year, make, model, trim, mileage, asking price, previous prices seen, dealer name and location, days listed, photo URLs, and the full option list where the site publishes it. Re-run daily and you get price history for free, because the system remembers what it saw yesterday.
Delivery is whatever fits your workflow — an Excel file, a Google Sheet that refreshes overnight, a Postgres table, or an API endpoint your own tooling can query. Most dealers take the Google Sheet because their whole desk can open it without installing anything.
Most large listing platforms run bot detection: fingerprinting, rate limiting, JavaScript challenges, and layouts that change without notice. This is routine rather than exotic — the build uses a real browser engine, sensible request pacing, rotating residential exits where the site demands it, and structural selectors that survive a cosmetic redesign.
The pacing part matters and is not just politeness. A scraper that hammers a site gets blocked, and a blocked scraper delivers nothing. Running at a rate the site can absorb is what makes a pipeline last months instead of days.
AutoTrader, Cars.com, CarGurus, CarMax, Carvana, AutoScout24, Mobile.de, Carsales, Facebook Marketplace
Python, Playwright, Selenium, Pandas, Proxy rotation, PostgreSQL, Google Sheets API
Collecting publicly visible listing data is generally lawful in the United States, and US appellate courts have repeatedly held that scraping public pages is not unauthorised access. What matters is what you collect and what you do with it — I stay off personal data behind logins, keep request rates civil, and will say plainly if a specific request looks like a problem.
As fresh as you want to pay for. Daily overnight refresh is the common choice and suits pricing work. Hourly or near-real-time is possible for a narrow watch list, and costs more because it means more requests and more infrastructure.
Yes, and it is often the most useful feed. Individual dealer sites usually have lighter defences than the big portals and update the moment a car is repriced, which makes them a faster signal than the aggregators.
Same work, different markup. AutoScout24, Mobile.de and Carsales are all builds I have done. Non-English sites are fine — the field names get normalised into whatever schema you are already using.
A single site with daily delivery usually lands between $300 and $900. Multi-site pipelines with price history and roll-ups run higher. You get a fixed quote within 24 hours of describing the sites and fields, not an hourly meter.