Data mining might be decades old, but the rate at which they collect and analyze data, and the accuracy with which they make predictions has made that old kind of data mining into cave paintings. About 10 years ago (keep that in mind, things have advanced MUCH further) Target got in trouble for how well their predictive algorithm works. They figured out that due to hormonal changes women who have gotten pregnant change their shopping habits, and what that change was, so they started sending maternity related coupons to pregnant customers, problem is their algorithm was so good they were sending these coupons to women (and teenage girls) before the WOMEN knew they were pregnant. This caused some concern, especially among the families of pregnant teens, so they had to ease off the program, they still do it but they hide it better now, they send generic looking circulars that just “happen” to have maternity clothes and lotion and all those other things pregnant women buy.
Even if you fight the cookies a lot is known about your computer and its travels on the internet. And most people don’t fight the cookies. And the cross level data is epic. This quiz asks which WB cartoon character you like most, somewhere out there is a list of stuff people that like Bugs Bunny buy more of than people who like Marvin the Martian, and that selection will change the popup ads you see. That’s the big change in data mining now, they now have access to enough cross referenced data that they know there are no innocuous unrelated choices, everything you do tells them something about what you’ll buy later today, and next week, and next month, and next year.
My post was about this quiz, and so far it doesn’t seem to be an issue for me.