IBM’s Watson is now working as a psychologist for ad tech firm Unruly
The UK-based firm is employing the supercomputer to generate large numbers of psychological profiles of potential buyers at scale, cheaply.
Marketers have run psychological profiles on consumers for quite a while, but they take time and money to do well. Now, IBM’s Watson supercomputer is helping ad tech firm Unruly generate psychological profiles at scale, quickly and cheaply, to help target video ads.
This week, the UK-based firm announced its new Watson-assisted Unruly DNA tool. The central goal is to increase sales by trying to replicate “light buyers” of a brand’s products, the idea being that these kinds of customers already know the product line and haven’t reached their limit of purchases. This is the latest effort by the News Corp-owned firm to understand the hidden feelings of customers. In September, for instance, it unveiled a multilayered program to measure biometric responses of viewers to online video or TV ads.
The first step in Unruly DNA is the creation of a panel of 10,000 people, half in the US and half in the UK. Each panel member fills out an online questionnaire about their profile and their buying habits. In the initial beta test of the last few weeks, the DNA tool focused on consumer packaged goods.
The panelists provide information about their Twitter and Facebook accounts, and Watson employs linguistic analysis and machine learning to digest those social posts and generate psychological profiles of each panelist, according to the Big 5/OCEAN standard. Marketing data firm Cambridge Analytica, best known as Donald Trump’s data firm, similarly uses the OCEAN standard to guide its creation of psychological profiles for every adult consumer in the US.
That standard offers five major characteristics — openness to experience, conscientiousness, extraversion, agreeableness and neuroticism (which Unruly labels as “emotional range”) — plus sub-characteristics for each of the main ones. Watson scores each person for each characteristic and sub-characteristic, based on their deviation from a mean that Watson has separately assembled from an analysis of millions of social profiles. Here’s a screen shot of Unruly DNA representing the psychological attributes:
Watson then groups the panelists by their brand behavior, such as those who have purchased a given product at least once in the last six months. It then creates an aggregate psychological profile for that buying segment.
For instance, Unruly said, it has discovered that buyers of Maybelline cosmetics are “highly imaginative,” purchasers of vitamin water Glaceau “like to explore new things,” and those who buy hair care product Tresemme “are particularly agreeable.”
Now, Unruly is ready to find a larger pool of new buyers.
It matches the psychology/demographics of each buying segment of Unruly panelists against the psychology/demographics of large segments of online users in third-party data from data management platform (DMP) Lotame.
Those segments of online users have been generated and supplied to Lotame by data providers like V12 or VisualDNA. Those providers take a seed group of online users who have filled out a demographic questionnaire and psychological profile, and then lookalike-match key attributes with anonymized third-party data of many other users, to get large numbers of users that are segmented by those same attributes.
By matching its panel of users to the larger data set in Lotame by its demographic and psychological attributes, Unruly expects to discover a large number of potential customers for its clients’ products. Those potential customers are then targeted for video ads that appeal to their personality traits, primarily on websites and through cookies.
Unruly Chief Strategy Officer Scott Button told me that Lotame and its data providers don’t have profiles that combine demographics, psychology and buying habits. With Watson’s addition of psychology profiles, Unruly approaches Lotame with a data set that combines all three and tries to assemble a much larger universe of potential buyers.
Button said that this DNA project, having just launched, does not yet have any clients or any data to indicate if it works better than other kinds of targeting.
But, he said, Watson is “faster, cheaper, and more scalable” in generating psychological portraits than, say, having panelists take the tests, using specialized software to analyze them, and then having human experts validate the results. When IBM measured Watson’s psychological scoring against test groups, he said, the error rate was low.
He added that he’s “not aware of anyone else doing what we’re doing” with Watson.
Opinions expressed in this article are those of the guest author and not necessarily MarTech. Staff authors are listed here.
New on MarTech