Beyond CPM: Bidding And Behavioral Data

For many ad networks and publishers, deploying behavioral data has already proven itself as a selling point for justifying higher CPMs. In the discussion below, Alex Hooshmand, director of  product management at Right Media, argues, however, that defining behavioral targeting's value in terms of the CPM model may be selling the true potential of the behavioral revolution short.

BI: Looking at all the different ad networks and publishers who've jumped on the behavioral bandwagon in the past two  or three years, what sorts of gaps do you think remain unfilled?

Hooshmand: A core limitation -- really, I think you could call it the dirty little secret of behavioral targeting right now -- is that if you go out there to look for big success stories, they are few and far between, at least if you define success as involving performance that can scale up repeatedly over time in multiple or long-term campaigns. To be sure, there are plenty of examples of small top mid-size tests performing well versus other kinds of placements. But when large advertisers get ready to ramp up their spending and reach, the performance drops.

BI: Why is that?

Hooshmand: Part of it, I think a big part, is a short-sighted model for deploying it. Once a BT trial 'works' on a small scale, publishers or networks -- and who can blame them? -- based on the premises of CPM, jack the rates up. Consequently ROI keeps falling with every subsequent attempt.

BI: So there's sort of a law of diminishing effects built into the way behavioral targeting is marketed and sold?

Hooshmand: Publishers essentially cherry-pick their best behavioral targets and then raise CPMs unsustainably high based on them. Also, because behavioral targeting's become such a hot-button issue for advertisers looking to make display advertising more efficient. So advertisers naturally are willing to experiment with higher CPMs, but publishers haven't had enough pressure to justify those rates up till now.

BI: So how does your approach differ from the conventional one?

Hooshmand: With auction-exchange pricing, you've got a model where you're only paying for users you think are valuable, at a price that makes sense for your ROI goals. There's a much closer balance pf performance and price. A second difference between what we're doing and the norm is that we can aggregate people interested in a particular area, say car shoppers, across all the publishers on the exchange, all on a single platform.

BI: So there's a combination of wide-scale aggregation and transparency?

Hooshmand: That's right. Because there's an open bidding process, it compels each individual publisher to provide more granularly segmented data to differentiate themselves competitively.

BI: So you've noted how an exchange challenges publishers to enhance behavioral segmentation. How does it change how advertisers deploy BT?

Hooshmand: From an advertiser point of view, the benefit of an auction model for BT lies in the opportunities for information asymmetry it provides. The example that comes to mind is an art auction, where the bidders who know more about the potential value of a painting will gain an edge based on what they know. With an exchange, if advertisers know the specific kind of data points they're looking for, they have a much better chance of finding maximum ROI. Advertisers tend to know what kinds of behavioral data they need most at a given time to add value to a campaign. It behooves publishers who have confidence in their user data to open it up to a bidding process.

BI: Do you expect the availability of new options such as yours to expand adoption of BT?

Hooshmand: I think we're at an inflection point. The transformation involves moving beyond a sell-side approach where publishers say 'we've got some data about users and here's what it costs -- take it or leave it' -- to a situation where advertisers can judge what it's worth to them.

BI: What are some other 'next frontiers' you see for behavioral targeting going forward?

Hooshmand: Search engines and marketers have been sitting on a gold mine of data that's so far been stuck in one channel. But we now have several clients who are using post-search behavior to target banner or display ads. Looking a little further ahead, there's enormous potential in mobile user data, but at this point there remain too many barriers to the kind of data transparency that would really be conducive to an exchange model. I don't see that happening in the very short run -- but eventually this is another area that can become very conducive to an exchange model.