Thursday, July 3, 2008

Why Behavioral Targeting is not a Publisher-driven Technology

A colleague recently pinged me to get my thoughts on behavioral targeting companies and the new influx of a number of different companies. Basically, I argued that if you are focused solely on trying to sell behavioral targeting to publishers your company will fail. This dovetails a bit with a post Mike Nolet recently did on the Plight of the Ad-Technology startup where he points out three reasons why its hard to make money as an ad company: 1. Integration sucks, 2. IP Ownership, and 3. Pricing is Hard. While all those points are true, my argument basically comes down to the fact that no amount of technology can change the core audience of a publisher (see my value chain analysis of the ad industry for more detail).

The basic problem of behavioral targeting technology is:
  • The sites with good data / good audience don't need you
  • The sites that need you don't have good data or a good audience
This is the fundamental reason why companies like Revenue Science and Tacoda had to create their own ad networks (early on RSI tried to sell BT as a product to publishers). They needed some way to transfer value between their publisher partners. The only way to get your BT technology to work is to buy the data/audience from the sites that have it in order to sell it to the sites that don't. (Hence the recent rise of ad exchanges to help facilitate this further)

Unfortunately, if you are publisher-focused, there is an upper bound to how much money you can make. Here's why:

1. The valuable audience you have identified is fixed (if you can target all WSJ members on Facebook, you're still limited to just targeting those WSJ members)

2. If you decide to expand that audience with inferred behavior (e.g. "People like you also liked") you still can't reach the rest of the population and your CPM values will drop

3. Each individual has an upper limit to how many advertisements they will respond to, so you can only target these valuable segments a limited number of times in a given period before they burn out

4. The more niche the audience segment the harder it is to pull together enough volume to be valuable

5. After the initial obvious verticals are gone (finance, autos, electronics, real estate), new valuable verticals are harder and more costly to identify
If you are strictly focused on trying to sell BT technology to publishers, you are not going to make a heck of a lot of money based on the dynamics above. However, there are two other ways to make money if you have invested in good BT technology:

1. Start an ad network and become a facilitator of buying and trading audience between publishers


2. Align with Marketers and Agencies to improve the effectiveness of a Marketer's ad buy.

The real value in BT is actually not in raising the average CPM of a publisher site (which is a fallacy in itself), but in improving the effectiveness of a Marketer's ad buy by identifying those individuals most likely to respond to the Marketer's offer. The best way to take advantage of that value is by aligning with the Marketers rather than the Publishers.