Hacker News new | ask | show | jobs
by kapgoals 5245 days ago
The point is that there is no way of knowing these things. We're just trying to replicate what a human would be able to decide if he/she were looking at the listings. In this example, http://autoglance.com/img/blog/hidngcars/2.png , wouldn't you contact the seller of Car A before you would contact the seller of Car B? I don't think that many people's first impression would to be skepticism towards car A and hence think that they should look at Car B first.
1 comments

From a search perspective I might click on car A's add first, but then I would click on B's ad before contacting A. It would only take one little thing to make the difference. Nicer car color? Built in navteq navigation with a good screen? V8 v.s. v6? Smoke free? I could easily pay 1K for those differences.

Hiding similar, but slightly more expensive cars also means that you will be showing more entry level cars. I wouldn't wish the entry level chrysler 300 on anyone it is a stupid car that shouldn't have been made, but it will show up first on this search interface because of everything they stripped out and downgraded to make it cheap. If I was searching for a car with something more I would have a better time not using this interface.

If you wanted a particular color, you could already do that with the color filter we have set up. As for things like premium model or navigation, you certainly have a point; we're currently working on a free text search which would allow you to remove any listing that doesn't contain your search string in it's description. Hopefully this would help significantly in the situation you are describing.
The problem is, even if you add more filters, used cars are still different enough from each other that your users likely want to see the hidden results--unlike airplane seats, where some results are more expensive tickets with an additional connection and therefore 99.9% of searchers will agree that result is worse.

What if there are two cars, one of which is $250 more, but which has a few (fairly minor) additional features and is 75 miles closer to the user? Figuring out which car the user wants to see is a hard problem.

That obviously doesn't mean it's not worth working on, but my guess is your current solution is sub-optimal enough you'd be better off displaying all of the results and simply ordering them based on these criteria. Or at least making it very obvious to the user there are other relevant results, since most people are used to all relevant results being displayed in one long list (that's my guess; I'm not an expert in this at all).