Very cool. What's the software stack behind this? How are you performing searches across all this very diverse data? For example, you must be doing some sort of text-analysis to retrieve the mileage property out of the text, no?
We're also experimenting with some structured parsing for the universal dealership parser we alluded to - this means trying to figure out which part of the page has a listing so we dont accidentally report the mileage from a warranty or similar listings lower down the page... turns out to be a fairly tricky problem but its also a ton of fun to work on! Will keep you posted if we start having a ton of success automagically crawling the interweb. Thanks!