Hacker News Reader

Sure, but datawarehouse techniques have only just now reached a point in their evolution where they are able to react to pregnancy-related changes in buying behavior, which is still probably going to return false positives. I can't think of anything more significant in a human's life in affecting buying habits than pregnancy, yet that's the cutting edge.

Getting data is easy, it's doing something profitable with it that remains hard.

View on HN

It's the same as any recommendation engine: Because you've bought x, based on similar patterns from other customers you're probably likely to buy y. The fact that it's pregnancy-related makes for a good (and slightly controversial) news article.

Throwing in the additional information that exists outside their system can lead to even better and timely recommendations.

My point is that deep-dive market research, let alone recommendation engines, really aren't very good (yet). Recommendation engines themselves, as a shallow version of what I'm talking about, haven't really progressed at all in the 10 or so years since Amazon et al figured out how to display the top three things that other people bought along the thing you're looking at, which isn't rocket science.

Reply on news.ycombinator.com