Robert_Charlton - 11:22 pm on Nov 18, 2012 (gmt 0)

I'm very curious about what they'll be measuring. It's hard to tell from the article whether "viewed" in the sense I use the word actually enters into this. Apparently, click-throughs don't (enter into it). The tool may simply be correlating ad impression data with eventual purchase data.

There's no doubt they need to suggest cause and effect, and some sort of (cookie-based) attribution modelling, for example, would be impressive if they can do that.

This example the story cites... assuming no statistical double-talk... is nevertheless very intriguing...

Online retailer Fab.com, which has tested Facebook's new service, was able to reduce its cost per new customer acquisition by 39 percent when it served ads to consumers deemed most likely to convert, Facebook said. Facebook defines a conversion as anything from a completed sale, to a consumer taking another desired action on a website, such as registering for a newsletter.

It may suggest that Facebook works well as a reminder, perhaps a subliminal reminder.

I'm hard put to remember a single Facebook ad I've viewed, though I do remember lots of stuff friends share with me... but I also know I'm not a typical consumer, and Facebook may think it knows 'where I've been', but it really doesn't have a clue.