Is there life without third-party cookies? (Part 3 of 3)
With the third-party cookie starting to lose ground in the ad-targeting world, it’s time for a multi-layer approach to audience segmentation that delivers a richer, more satisfying result in the end. I’m talking about getting more meaning out of the content a user is reading right now, rather than “spying” on their entire browsing history with a tracking cookie.
Below are some of the layers of audience segmentation that we can derive from a single piece of reading material. They are all based on probabilistic inferences, which means that in isolated cases, some may be bad guesses. (The person reading a review of Jay-Z’s latest track on iTunes might be a conservative Mozart-loving septuagenarian, but come on! That’s not likely). Just think of these as inductive inferences that, in aggregate, are going to be correct far more often than not, and therefore a segmentation based upon them will be at least as accurate as the cookie-based method ever was, with regard to its grabbing the intended audience. (In part 1, Content vs. cookie, I wrote about the accuracy of targeting based on the content versus the cookie, and how the two methods are on more of an even footing than people realize.)
Physically: Is there a biological or anatomical theme in the material that suggests physical properties of the reader? If they are reading about what hairstyles are best for thinning hair, this raises the odds that the reader is over 40 and female, whereas a high presence of certain punk-culture slang increases the chance that the reader is male and under the age of 25.
Mentally: What is the topic density of the material (i.e., how far does it go into detail on a single topic)? What is the reading level? Is it 14th grade (meaning, second year of college), or 5th grade (which sadly is just slightly below the public average)? These features have implications for the sophistication level, the attention span, and the education level of the reader.
Socially: Does the text have a political leaning, left or right? Are there implications for the user’s affluence level, such as discussion of luxury goods? Is there an entertainment angle that would indicate social behaviours, such as sports-event tailgating? Are there indications of their current projects, such as “helping your kids through their college application process”, which would imply that the reader is not far from being an “empty nester?”
Spiritually: Yes, I said spiritually. Is the reader absorbing optimistic material, or pessimistic? Is there grit, grime and dirt in the content, or poetry, elegance, and grace? Yes, we can capture probabilistic indicators of these with linguistic analysis.
I’m not saying Temnos has captured all of the above, today. Rather, I’ve drawn a map of the territory we are traversing. We’ve covered most of it and are not far from the rest. My point is that it’s time to realize that the cookie-trail was always an indirect, approximate way to get at an audience segment. It is also time that we embraced content analysis as something just as reliable, and more viable, in our increasingly privacy-protected lifestyle.