Comments on Behavioral Targeting Piece

Ari Kaufman gave me a writeup on his blog Friday, submitting a few comments on this piece about Behavioral Targeting on iMedia. He writes:

Conceptually he did an interesting job of presenting the idea, but I am not sure that with the technology he is proposing to leverage, it is necessarily possible. Or maybe it’s possible but not really reliable.

As a follow-up to that, let me say that I think the idea of boosting profits by getting rid of the customers who actually take advantage of loss leaders, discounts, rebate policies, etc. is patently ridiculous. I wouldn't advocate actually making an attempt to use this approach. Count me as part of the camp that thinks Brad Anderson was nucking futs when he advocated identifying "devils" and avoiding marketing to them.

But on to Ari's point...

There is a huge difference between how Behavioral Marketing is talked about today from a conceptual standpoint and how behaviorally-targeted ad programs are actually sold. Proponents of Behavioral Marketing who have something to sell seem to talk quite a bit about custom segments, recency and frequency algorithms, unique behavioral identifiers for an advertiser's target audience and a bunch of other interesting stuff. Yet, when it's time to actually get down to brass tacks and sign an insertion order, the sales reps push back against anything that's remotely custom and try to sell category inventory instead.

I've been talking to my editor at iMedia about this, and there's some interest. Could be a panel at one of the summits, or it could be a column, or possibly both. I'll let you know once we have some direction.

Some of the companies selling behaviorally targeted solutions are able to demonstrate in theory that behavioral segmentation (and even the elimination of various undesirable segments) is possible. I have yet to see it done in practice, though. I'd challenge the behavioral marketing companies of the world to produce a case study with a testimonial from a client that shows custom segmentation and the elimination of undesirable segments actually worked as advertised.

I have no doubt these things are theoretically possible. It's the implementation I'm concerned with.