Cool article on behavioral targeting - How do you think they solved it?

Question

I just read an interesting article on how Target used purchase behavior as a predictor to whether they were pregnant or not. Here's the link.

Could someone explain how Target build the predictive model, or their best guess as to the process?

Update

The original article is in New York Times.

Answer 1

From what I've understood from these articles, Target managed to identify patterns when the pregnant customers, which know that they are pregnant, start buying pregnancy related items. This is a much easier task than to develop pregnancy test based solely on shopping habits.

Target has data with shoppers buying history and indicator whether the shopper was pregnant or not. Hence this is a classification problem, you observe a feature vector and you need to assign class to it. There are a lot of methods for doing that.