This is a hot buzzphrase and has been for 5 or 6 years in online advertising.
“Behavioral targeting allows marketers to respond to identified consumer behaviors and gives them the ability to deliver “timely” messaging in an environment outside of where the behavior was captured.
What this means to advertisers is that they now have the ability to deliver a message in a non-automotive research environment to a consumer who has been identified as “in-market” for a particular vehicle segment based on a previous behavior they had captured on a 3rd party automotive research site.
This level of laser-precision targeting was previously unheard of. Think about it. Imagine for a moment the total U.S. population. Then assume for the purpose of illustration that 2% are in-market at any given time. Of that 2%, the audience becomes even more fragmented across every possible vehicle category from SUVs to Luxury Vehicles to Sedans, etc. As you can see, tracking down in-market car shoppers can be very difficult, and looking for specific category intenders is like trying to find a needle in a haystack.
Outside of search marketing, contextual advertising and behavioral targeting, every other form of media is likely to have considerable waste, or, in other words, deliver messaging to people who would not consider the respective vehicle in the first place.”
Pretty marketing heavy and pretty light on “behavior” other than identifying content, keywords and the sites users look at. Here they post info that goes a little deeper. That’s better.
So what’s missing? the same thing that’s usually missing from campaigns and most online analytics, about 99% of behavior. Really. It’s good to track the clicks, the content affinities, personnas, and some “heat tracking”, but really no one is going to do much better than keyword targeting until they really analyze people’s behavior (schedules of reinforcement, consequences, histories, values). Not all of this can be done online. It must be observed. It must be experimental.
I’m not suggesting these advertising systems don’t do better than just display ads with no targeting. I am saying that behavioral targeting isn’t nearly as effective as people claim nor is it all that behavioral, beyond a very limited set of web page tracked behaviors. I am also not suggesting the industry can’t do better. it can.
I would not, as an advertiser, yet throw tons of money at BT as it rarely outperforms a well designed search keyword program. Why?
The cost! The amount of work it takes to run a BT program (asset creation, algos to retarget, spend management, etc. etc.) adds up quickly. Search marketing has far fewer of these complexities for the end advertiser.
Consider that BT companies are aggregating data across an infinitely varied set of contexts (lots of sites) vs. just Google, Yahoo, and MSN for Search marketing. That alone, increases the complexity by many orders of magnitudes. Keyword searches also are wonderfully simple behaviors and the results and resulting clicks are easily tracked feedback mechanisms. BT doesn’t have any of this, and certainly in a simple form.
So how will BT ever get there?
- Standardization – publishers and advertisers are going to have to agree on creative types, behavior metrics, etc.
- Transparency – users, publishers and advertisers are going to have to be far more open with data. only by very open access to data can everyone do their part in reporting and responding to behavior
- Offline data tie ins – BT companies need offline data or at least, off site, data to tie back all these ad implementations. Without knowing all the schedules and consequences inbetween, you can’t really be sure you’ve targeted anymore effectively than just running a demographically targeted ad
In a follow up I will post my outline for my vision for a Behavioral Targeting Tracking and Advertising Company.
I agree with your observation of a large part of behavior being neglected. I think the issue is that “behavioral targeting” as it’s currently understood mostly involves simple parameters such as time, section of site visited, browser type etc. It is not trying to make a comprehensive picture of the person’s interests, just finding correlations between simple parameters that reflect the interests on some vague level.
You talk about BT companies aggregating a lot of data, but I think the basic issue is that they don’t really have the most relevant data. They have taken easy-to-find parameters which you can calculate correlations with instead of actually trying to extract the data that is most relevant for an end-users behavior.
At Leiki we’ve developed behavior targeting starting from very detailed content analysis (done with an ontology with 200k concepts). The idea is that there’s a very detailed list of topics for every end user with an associated interest level for each topic. I think this is a large part of the “99% of the behavior that is missing” that you mention.
Petrus,
Yeah, we are in aggreement. What is collected isn’t behavior nor context.
After reviewing your materials (not a super detailed review…), you clearly have some interesting things.
How might I see it in action? The ontological context is important, but it’s not 99% of the behavior. Fortunately, 99% of the behavior isn’t needed, which might be your point.
Herein lies a different problem. there’s no way to actually TEST BT companies/software like you can test Google AdWords. How does one sample without spending a bunch?