Carvana scraper for vehicle listings and price data

Carvana data we collect for every vehicle

Carvana addresses each car by a numeric vehicle id in the URL and carries a second identifier, the stock number, together with the VIN. All three go into the row, so a car can be matched between runs and against other sources. The fields come from the search tile and from the vehicle page:

• Identity - vehicle id, stock number, VIN, vehicle URL, and whether the car is offered as used or brand new.
• Vehicle - year, make, model, trim, body type, exterior and interior color, seating capacity, fuel type, drivetrain, transmission, cylinders, MPG, and electric range on EVs.
• Mileage - the search tile abbreviates it; the vehicle page gives the exact odometer reading, and that is the figure we keep.
• Price - the current price, the original price shown as "was" after a reduction, and the dollar value of factory upgrades the tile says the price includes.
• Tags - Recent, Price Drop, No Reported Accidents and Great Deal, the last assigned by Carvana against the Kelley Blue Book Typical Listing Price. Each tag is its own column.
• Features - the feature list in Carvana's own groups, from Safety and Security to Entertainment and Technology, plus factory packages and options by name.
• Condition - imperfections, each with a title, its place on the car and a photo; the inspection report with its date and the result of every point; prior use such as lease or fleet; open recalls; number of keys.
• Delivery - the free shipping flag or the shipping charge, and the "Get it" date, both as quoted for the ZIP code of the run.
• Media - photo count and image addresses.

Field documentation comes with the dataset, and a field Carvana leaves empty on a car stays empty in the row.

Carvana price tracker: price drops, new listings and sold cars

One pass over the inventory is a snapshot. A Carvana price tracker earns its name by repeating it.

• Price steps. The site shows the current price and, after a cut, one earlier figure. A daily run keeps every step in between, so each VIN gets its own price curve, and price trends by model, year and mileage band can be drawn from the curves.
• New listings. The Recent tag marks a car made purchasable for the first time within the last 72 hours. Together with the first date we saw the vehicle id, it gives a dated feed of new listings.
• Sold listings. A sold car drops out of Carvana's search, so sold listings exist only in a dataset that was already running. A car held while a buyer completes the purchase is flagged as pending, and a car missing from the next run is closed as sold or withdrawn with the date it was last seen.
• Slices. The same stock is reachable by make and model, by body style, by city pages such as used cars in Atlanta, and through the filters: payment and price, year and mileage, fuel type, features, colors, drivetrain. We split large result sets along them and merge on vehicle id.
• ZIP code panels. Running the same cars for several ZIP codes shows where Carvana ships free, where it charges, and how soon it promises the car.