Relevancy Trumps Creepiness, and Some Thoughts About Behavioral Targeting

By Eric Goldman

On Monday I spoke on a panel at OMMA Behavioral. See the MediaPost recaps (1, 2, 3, 4). The crowd was buzzing about Dave Morgan’s earlier remarks (which I didn’t hear) that behavioral targeting is “creepy,” and throughout our panel discussion, any enthusiasm expressed about behavioral targeting was tempered by creepiness concerns.

I can understand this reaction, as least a little. When I was younger and first learned about the many tricks of marketer targeting, I was initially aghast by the seeming intrusion. They can’t do that, I thought.

As regular readers know, I’ve outgrown those sentiments. Now, I really don’t care what the machines know about me. And if the machines can figure out how to better cater to my interests and reduce the spam in my life, then I’m all for it.

At the same time, I think this latter observation suggests my real problem with behavioral targeting. There will always be some privacy diehards who will object to machine monitoring of their behavior on principle, but most people will be receptive (even after they get through the initial shock about behavioral tracking) if the targeting improves the user or consumer experience. Demonstrate to consumers that behavioral targeting gives them better results, and it’s an easy sale. Relevancy trumps creepiness.

But I haven’t seen any evidence that behavioral targeting has produced these payoffs (or, for that matter, any meaningful payoffs) for consumers yet. Current behavioral targeting practices might give marketers a little conversion lift compared to other targeting solutions (or not), but they have done little to change the overall fact that ads remain poorly targeted and crummy, and consumers still have plenty of incentives to treat ads as the pain to avoid through ad blindness or technology.

At this point, I’m still wondering if and when behavioral targeting will deliver on its theoretical promise. Sure, we can find excuses for the crummy user experiences today–the technology is still being developed, it’s hard to get useful datasets (more on that in a moment)–but those excuses only go so far, and they will wear thin quickly. For behavioral targeting to really be a game-changer, it needs to deliver dramatically improved ad relevancy for consumers, and we’re far from that ideal point.

I’ve argued before that for behavioral targeting to work, the marketer needs a comprehensive dataset about the consumer. Accordingly, a marketer–even an ad network–that relies solely on data collected from a consumer’s interaction with web servers simply can’t see enough data about the consumer to achieve a sufficient level of relevancy for the consumer. My paradigmatic example: no matter how much Amazon knows about my purchases from it and my browsing habits on its site, they still don’t know if I bought a book from someone else unless I tell them (and I have no reason to tell Amazon what books I buy elsewhere).

This is why I’m so intrigued by the Internet access provider-level targeting exemplified by Phorm and NebuAd. In theory, they get access to much better datasets than web server-level targeters. If I browsed for a book on Amazon but I bought the book at, the Internet access provider can know this while neither Amazon or B&N will know about my interactions with the other vendor.

For this reason, I’ve been quietly bemused by the legal fracas over Phorm and NebuAd’s practices. Don’t get me wrong–although the analysis is intensely fact-specific and I don’t have all the facts, I have serious concerns about the legality of their practices. But from my perspective, the battles over the legality of Phorm and NebuAd are a smokescreen for the real issue, which is that marketers who have only server-level data don’t want to compete against someone who has a better dataset than them. So expect plenty of continued fireworks over Phorm and NebuAd, but don’t kid yourself that it’s only the privacy advocates beating up on them.