About

I'm a partner in the advanced analytics group at Bain & Company, the global management consulting firm. My primary focus is on marketing analytics (bio). I've been writing here (views my own) about marketing, technology, e-business, and analytics since 2003 (blog name explained).

Email or follow me:

-->

11 posts categorized "Open Source"

February 05, 2008

Memo to MIHOO: A Modest Proposal for The Next Step

There are three ways an ad network can compete:  better reach, better data, better analytics.  Know more people, know more about them, know what to do with that information.  Jeffrey Rayport noted yesterday in MarketWatch that recent consolidations in the ad network business have made it a game that "only the mighty can play."  Meanwhile, Danny Sullivan wonders in Ad Age whether the Microsoft-Yahoo combination will be enough to catch Google, even if they can pull the integration off.

So here's an idea, borrowing from a time-honored software business strategy most recently exploited by Facebook when it opened up an API to its "social graph" (also inspired by recent fiddling I've done with the IBM Many Eyes project).  MIHOO should provide a way for "third-party algorithm developers" to query its combined data hoard of users and their clickstreams.  That way, they can "unleash the power of open-source algorithm development" by letting the unwashed mathematicians of the world back-test their ideas for getting consumer X to respond to offer Y in context Z.  Of course, this would come with a price: when you go to run your algorithm on the MIHOO network to "monetize" it, MIHOO takes the gross cut to the network and pays you, the algorithm developer a royalty.

Symbiotic way for MIHOO and the little guy to take on big GOOG.  Lots of problems with this to be sure, starting with privacy (even though it already "is", can data be further anonymized to make it tough to infer my personal clickstream?).  And, as a practical matter, can MIHOO expose a data warehouse that could take mass pounding from the world without slowing to a useless crawl?  Further, would it be worth it?  There is a diminishing return to finer and finer targeting ideas, after all.  As for the competitive risk -- that people will get insights on MIHOO's data set that they can deploy on some other network (Google's or otherwise) perhaps there is a way to abstract the data being queried until a promising algorithm can be "optioned" by MIHOO.

Granted, it's a bit out there... but work with me people, or help me find the fatal flaw.