I was looking at an address in Seattle on MapQuest today (yes, I do still use it from time to time), and I was shocked by the ads that I was offered. They all come from Google’s adsense: out of 10 textual ads (yes, quite amazing how much they can cram on that page), 4 were for DUI attorneys, 2 for life insurance, 2 for job search, 1 for a DJ and the last one for a florist. Getting ads for DUI stuff when you are looking around for driving directions gives quite a negative feeling about driving. Probably not what MapQuest would have chosen if they were given a choice.
That’s the inherent problem around any algorithm-based advertising (if not almost any computer-generated prediction). It’s very hard to encode and take into account psychological reactions to things. It’s much easier to just consider the financial correlation. If (I’m inflating the numbers here a couple of orders of magnitude) you put an ad (A) that will make 10% of the people click on it (out of confusion and disgust) and 50% never come back to the website and another ad (B) that only 8% click on it (which let’s say is the average click-through rate on ads – again, this number is a couple of orders of magnitude off), it’s easy to build an algorithm that will tell that A is better than B, but very hard to tell that B is better than A. It takes some time until you realize that there is a traffic reduction going on due to showing ad A.
Anyway, I’m probably going to be back to MapQuest. I keep rotating around the map provider I use not to miss new features and to make sure that I can appreciate all of them better.