Quick, who indexes higher on purchasing behavior on Amazon, men or women? That's a question former Amazon R&D executive David Selinger asked on a panel on recommendation engines, the more
transparent, less creepy side of behavioral targeting. The answer: neither. Selinger, now head of RichRelevance, used that example to illustrate that assumptions about behavior--that women overindex
on online buying--turn out to be wrong 99% of the time. "You need to see all dimension of consumer behavior to create a full profile," he said. Without getting a 360 view of user behavior, you won't
be able to derive any meaningful data. That sounds like a recipe for a lot of tracking.
When it comes to recommendations, Michael Strickman, CTO of ChoiceStream, which works with companies
like Zappos and TicketMaster, you have to go by what people do, not what they say. In other words, someone might tell a movie site they want to rent "Citizen Kane" or "Seven Samurai" but end up
actually watching a trashy movie. What's the last movie you watched on Netflix?