Behavioral targeting is a rapidly growing part of the online advertising mix. Given this, I wondered, would it be practical to build algorithms focused solely on the effectiveness of behavioral data?
In a world of CPA, certainly. In a world of CPC – the world of smaller advertisers – not so much. The key problem is a lack of effectiveness data. Optimizing inventory is easy. You own the ad space where the ad gets shown, you know if someone clicked or not. Behaviorally, you have to acquire the behavioral data and then also acquire the ad space to find out if someone clicked. The inefficiency created in generating learnings seems to me too large.
Sorry I don’t have time for a more thought-provoking version of this post.