• by Steve Smith , Staff Writer @popeyesm, May 30, 2008

In pulling together panelists and topics for the upcoming OMMA Behavioral event in San Francisco (July 21),
this third iteration of the series broadens the view of "behavioral" to include more of the technologies and techniques that leverage user data in different ways both on-site and offline. We have several panels this time focusing on data-driven optimization at retail sites and next-gen applications of behavioral tracking to do more predictive modeling. Coincidentally, a new piece of research from MIT addresses some of these concerns head-on and points the way forward to a next stage of Web site architecture that actually targets different users with variable versions of the same site. An upcoming article from several MIT researchers suggests that retail sites can move beyond personalized sites and recommendation engines into "Website Morphing."

Professors John R. Hauser, Glen L. Urban, Gulherme Liberali, and Michael Braun conducted this research and test at MIT and with British Telecom. According to the report, "We propose another approach: ‘morphing' the website automatically by matching website characteristics to the customers' cognitive styles. Our practical goal is to morph the website's basic structure (site backbone) and other functional characteristics in real time... . Website morphing is an example of targeting optimal marketing communications to customer segments."

Well, easier said than done. As the researchers lay out in dense, obtuse, equation-laden detail (full paper is available for download here, each step is fraught with variables and technical hurdles. First, you must define a cognitive style, which generally describes the different ways in which people gather and process information. In this test, the group settled on four constructs that rendered eight styles: leader vs. follower, analytic/visual vs. holistic/verbal, impulsive vs. deliberative, and (active) reader vs. (passive) listener. Ideally, a site would determine an individual user's cognitive style by analyzing the clickstream. Using the right algorithms and probabilities, every click a user makes is essentially a decision that helps define their process for gathering information and making judgments.

Once a site segments incoming audiences into a range of cognitive styles, it can determine which sets of features map well against each segment. For instance, more a graphically oriented cognitive style  responds better to graphs ands images, while the verbal style looks for text and audio features. In the test for British Telecom, the researchers used hands-on surveys of the 800 or so subjects to do some of this mapping, but they found that features like more general information about the company appealed to "holistic" audiences, while tools that let users interact and manipulate information worked better with "analytical" users. The "followers" were looking for validation from other users and sought more postings from others, while the "deliberative" thinkers liked to post comments themselves.

The researchers found that "respondents who are holistic/verbal or readers prefer focused content... . Impulsive respondents prefer small information loads. The tendency to go first to plan comparisons and virtual advisors while avoiding general information appears to [be] a trait that distinguishes analytic/visual from holistic/verbal respondents."

In the experiment for BT, which sells broadband plans in the U.K. market, the test audience experienced the site differently according to their cognitive styles. The site "morph" involved more and fewer features, graphs, tools, links and virtual advisors. The end result was a substantial increase in intent to buy. The researchers argue that site morphing against cognitive styles can improve sales by 20%. For BT that would mean a potential $80 million in additional revenue just by behaviorally targeting an entire site design.

If retailers can better understand the ways in which specific customers really shop and decide, then the very voice and architecture of the store can target their cognitive process, not just their immediate needs or shopping goals. Recommendation engines are fine for pushing goods to the end user they are likely to want, but it doesn't do anything to help them make the decision to buy. Personalization of sites generally require some user self-selection and proactive involvement in the process. The morphing the researchers tested here is entirely passive.

Where this is headed is anyone's guess, but dangling a 20% increase in sales in front of marketers and site owners is sure to get noticed.

And by the way, as we get closer to nailing down the final roster of panelists for OMMA Behavioral, there is always room for solid case studies of campaigns and brand strategies that leveraged a range of behavioral approaches. Contact me directly at popeyesmith@comcast.net . As we get closer to the show, we will again go to our readers for input on the topics you would like to see raised on these panels and in our popular Grill the Vendors closing panel.

advertisement

advertisement