In this article Augustine Fou argues that behavioral targeting is close to believing in Santa. Everybody does so in their early marketing days, but after a certain amount of experiences you mature and don’t take it that serious any more. It is still nice to attract clients, as most buzzwords are, but the magic disappears.
His main arguments against using behavioral targeting is that he believes it is based too much on the past (and you can’t tell future behavior from past behavior) and that the behavioral data collected is not valid as it cannot be enriched with personal identifiable information from the user (for privacy reasons). Fou also presents a solution: wait for the user to search for your product, then deliver what he needs and where he needs it and everybody is happy. The user pulling for offers is for Fou the perfect substitute for pushing offers to somebody who belongs to an almost randomly formed target group.
The article was online for about an hour when the first comments came in, criticizing the click-through calculations used, the limitation of online marketing to the SEM sector by only focusing on searches, and the problem about how user will know about a product if they have to search for it in order to find it.
I do agree with these critics, but there is something bigger at stage, especially for online marketers. Fou is group chief digital officer in the healthcare sector. He works in online marketing. So he has influence on how different marketing options are perceived and used. Yet somehow, the advantages of behavioral targeting did not reach him, so the industry might well do some homework to spread the word about how behavioral targeting has been evolving. Not only to consumers, but also to marketers. And how many different types of behavioral targeting techniques are there, some more and some less efficient than others. Marketers won’t buy what they don’t understand!
Does past behavior really indicate future behavior? Well it does. It just depends on the way the past is framed. Of course yesterday’s behavior is much more likely to influence tomorrow’s behavior, than what somebody did 3 month ago. So at best behavioral targeting does not rely on generational or even monthly data, but gives more weight to more recent data. Interests shift, evolve, get inspired by information found offline and online, and are enacted within a relatively short timeframe. Before buying a washing machine, the potential customer will look for washing machines. But not only on search engines, but she will draw information from many different sources. That is why it is crucial not only to focus on SEM. Somebody might start the search on google, but from there, gathering information may occur on a very broad array of different sites. And not all of them work with search engines, or will allow you to tap into the search results. So, do use behavioral data, but don’t go back too far. Be present at the beginning of the funnel.
The second important differentiation is the source of the behavioral data. Of course a banner on a website gathers different statistics than for example a toolbar or other web based applications. While website banners have to abstract from the data of many people, toolbars can add up the interests of one person alone. This does not mean that the toolbar has to communicate with an outside server or mix this knowledge with personal identifiable information. But if somebody is surfing a lot on different sites with washing machines, chances are the person is interested in washing machines. Here behavioral targeting trumps contextual targeting, because it is not moment bound. While contextual can only react the minute the person is available (e.g. logging into facebook), the behavioral part can help to understand what the person was looking for just before logging into facebook (e.g. looking for a washing machine) and can act accordingly. Today it is already technically possible to abstract the data to a point that the user cannot be identified in the retro-perspective, while the targeting engine still remembers what the person looked at (not where, but what).
In this approach the past purchases are not that important. The assumption is, if you search for something you might want to buy something related. Of course, if you bought a car, it can be reasoned that you might need car insurance. However, you definetly will be surfing around car insurance sites as well, so I do not need to know that you just bought a car. Your behavior will tell me. And through showing the best available car insurance offer at you while you are looking for information, I might well advance you in the purchase funnel from “what kinds of insurance options are there” to “I’ll get that one”.
Of course, all this implicates that the offer is worth the attention, because even the best targeting won’t help you if the offer is crap.