Autotrader DB Scanner?
Discussion
I've had an idea knocking around for a while now ... a computer program which reads the Autotrader website everyday, and builds up statistics on "actual" car prices, rather than those in Parkers/Glasses etc.
Cars could be automatically categorised by age, mileage, geographic area, colour, etc - though probably not by condition, since this is harder to read automatically.
Once it's built up a reasonable amount of data about each model, it should be possible to spot bargains - e.g. cars which are more than a standard deviation below the average price for their age/mileage.
It could then bring up a series of links to the day's bargains, for the user to read himself. In most cases there would be a reason for the car to be cheap, e.g. damage, which the computer would have trouble reading; but it should also identify some genuine bargains.
The program could also work out which geographical areas are most/least expensive for a certain model... so you could buy a cheap car in a cheap area, and sell at a higher price in an expensive area.
May also be useful if you're just looking for a specific car for your own use, and don't have time/inclination to scan the website several times a day.
Having programmed it up, I could use it myself and/or license it to dealers/enthusiasts/etc.
Anyone want to point out the flaw in my logic..?
Cars could be automatically categorised by age, mileage, geographic area, colour, etc - though probably not by condition, since this is harder to read automatically.
Once it's built up a reasonable amount of data about each model, it should be possible to spot bargains - e.g. cars which are more than a standard deviation below the average price for their age/mileage.
It could then bring up a series of links to the day's bargains, for the user to read himself. In most cases there would be a reason for the car to be cheap, e.g. damage, which the computer would have trouble reading; but it should also identify some genuine bargains.
The program could also work out which geographical areas are most/least expensive for a certain model... so you could buy a cheap car in a cheap area, and sell at a higher price in an expensive area.
May also be useful if you're just looking for a specific car for your own use, and don't have time/inclination to scan the website several times a day.
Having programmed it up, I could use it myself and/or license it to dealers/enthusiasts/etc.
Anyone want to point out the flaw in my logic..?
Major flaw I can see is that you will be scraping Autotraders copyrighted data from their database, so you will run the risk of being sued.
I thought of such ideas for my car classified website www.purpleads.com but soon dismissed them on those grounds.
Autotrader are unlikely to grant you access to their db as if they thought the idea was good enough they would probably do it themselves.
Spidering their db will also use a massive amount of bandwidth which they would (should) notice and close the hole anyway.
You MAY get away with it if you were to use a sample of the data but you would still be using their copyright so are still running the risk of being sued.
I'm no lawyer, just my 2p.
Cheers, John
I thought of such ideas for my car classified website www.purpleads.com but soon dismissed them on those grounds.
Autotrader are unlikely to grant you access to their db as if they thought the idea was good enough they would probably do it themselves.
Spidering their db will also use a massive amount of bandwidth which they would (should) notice and close the hole anyway.
You MAY get away with it if you were to use a sample of the data but you would still be using their copyright so are still running the risk of being sued.
I'm no lawyer, just my 2p.
Cheers, John
Yep, I don't think you'd get direct DB access, but it wouldn't be too difficult to write a perl script which would parse the html pages it returns.
Once you factor in travelling to the car, maybe only buying 1 in 3 you look at, advertising, insurance, etc etc it would only make sense to do it for relatively high value cars (where there's a greater opportunity for arbitrage).
So perhaps you could pick 25 target models, and read those pages say twice a day - i.e. 50 total database accesses, which would be a tiny proportion of their daily traffic. If it was done at random times of day, there wouldn't be any particular pattern to detect or object to.
But... I guess the copyright problem may present itself if the program was sold publically. Although there is a site which (presumably legally) compiles stats of eBay prices, and identifies the most lucrative markets; wish I could remember the URL, but it seems completely legit. Perhaps the same copyright issues don't apply.
Any thoughts much appreciated. Cheers.
>> Edited by blueski on Thursday 9th December 22:58
Once you factor in travelling to the car, maybe only buying 1 in 3 you look at, advertising, insurance, etc etc it would only make sense to do it for relatively high value cars (where there's a greater opportunity for arbitrage).
So perhaps you could pick 25 target models, and read those pages say twice a day - i.e. 50 total database accesses, which would be a tiny proportion of their daily traffic. If it was done at random times of day, there wouldn't be any particular pattern to detect or object to.
But... I guess the copyright problem may present itself if the program was sold publically. Although there is a site which (presumably legally) compiles stats of eBay prices, and identifies the most lucrative markets; wish I could remember the URL, but it seems completely legit. Perhaps the same copyright issues don't apply.
Any thoughts much appreciated. Cheers.
>> Edited by blueski on Thursday 9th December 22:58
PetrolTed said:
I've dreamt up all sorts of ideas for classified ad systems but at the end of the day, people like looking for cars. Take the 'skill' or fun of browsing car ads out and it becomes a boring process.
Yep... agree.
But how about using this system to come up with buying (to sell) suggestions for myself? As a means of spotting bargains across lots of models, anywhere in the country... without spending all day on the Autotrader website.
Trading standards had a bit of software that scanned in local classifieds, Loot and AT etc. It was programmed to identify telephone numbers and alet them to repeated numbers (ie trader hiding as a private sale). I am not a techi but it can't be that difficult to OCR make and model, price etc. and drop it in a database?? Are you then breaching copyright any more than those bloody canvassers are?
Gassing Station | Business | Top of Page | What's New | My Stuff