Monday, June 09, 2008

Adapting Websites to Users

Websites that change to fit the cognitive style of the user could be more effective at selling products online.

By Erica Naone

It's hard to build a website that will please everyone. Some people respond best when they see basic facts on a clean page, others when they have a lot of charts and graphs at their fingertips. Now researchers at MIT's Sloan School of Management hope to make websites better at selling products by making them adapt automatically to each visitor, presenting information in a way that complements that person's style of thinking.

The researchers' initial studies show that morphing a website to suit different types of visitors could increase the site's sales by about 20 percent. While quite a few sites, such as Amazon.com, offer personalized features, many of those sites adapt by drawing information from user profiles, stored cookies, or long questionnaires. The Sloan system, however, adapts to unknown users within the first few clicks on the website by analyzing each user's pattern of clicks.

John Hauser, a professor of marketing at the Sloan School and the lead author of a paper on the research that is slated to appear in Marketing Science, explains that a website running the system would detect a user's cognitive style. It would watch for traits, such as whether or not the user is detail oriented, and morph to complement that style. The changes would be subtle. "Suddenly, you're finding the website is easy to navigate, more comfortable, and it gives you the information you need," Hauser says. The user, he says, shouldn't even realize that the website is personalized.

Morphing to fit the user: Based on a user's pattern of clicks, the prototype website for British Telecom, shown above, changes to fit the user by making a guess at the user's cognitive style. For example, after determining whether he or she thinks in holistic or analytic terms, the website adapts to a form that it deems most likely to sell products to that user.
Credit: Marketing Science/MIT’s Sloan School of Management

The researchers built a prototype website for British Telecom, set up to sell broadband plans. The website is designed so that the first few clicks that visitors make are likely to reveal aspects of cognitive style. For example, the initial page that a user sees lets her choose, among other things, to compare plans using a chart or to interact with a broadband advisor. "You can see that someone who's very analytic is probably more likely to go to 'compare plans' than to the direct advisor," says Hauser. Within about 10 clicks, the system makes a guess at the user's cognitive style and morphs to fit. "If we determine that you like lots of graphs, you're going to start seeing lots of graphs," he says. "If we determine that you like to get advice from peers, you're going to see lots of advice from peers."

In addition to guessing at each user's cognitive style by analyzing that person's pattern of clicks, the system would track data over time to see which versions of the website work most effectively for which cognitive styles.

Peter Brusilovsky, director of the personalized adaptive Web systems lab at the school of information sciences at the University of Pittsburgh, says that, while there's been much work on using cognitive styles to adapt Web pages to users, in most cases it's been for education, not for e-commerce. This Sloan School's approach, he feels, is an interesting one. Although personalization is valuable whether it's done automatically or by the user, Brusilovsky says that "what is possible to do automatically is just impossible to do through the user, since users typically have little time to invest, and may not really be sophisticated [about how to adapt a page the way they want]." He particularly sees potential for these techniques for mobile Web pages, which, due to limited bandwidth and awkward interfaces, are hard for users to personalize.

Two views into the same site: The prototype website that MIT researchers built for British Telecom takes different forms, depending on its guess as to the user's cognitive style. Above left is an example of what an analytic, visually oriented user might see: a Web page full of charts and graphs. Above right is an example of what a more holistic, impulsive user might see. This version of the website minimizes the information, specifically recommending three plans.
Credit: Marketing Science/MIT’s Sloan School of Management

Bamshad Mobasher, an associate professor in the school of computer science at DePaul University, in Chicago, who has done work on adapting websites based on data collected from users' patterns of clicks, says that a lot of other work, including his own, relies simply on matching a user to others who have clicked through a website in a similar way, without trying to discover what those patterns mean. Adding the psychological dimension, he says, makes the task more challenging for a website's designers, and he says that he'll be interested to see whether that approach turns out to be better than measuring users' past behavior alone.

Glen Urban, a professor of marketing at the Sloan School who was involved in the research, says that the team plans to build a full version of its system for the Japanese company Suruga Bank. For the Suruga project, which is being done through the MIT Center for Digital Business, the researchers plan to watch website users for cultural attitudes as well as for cognitive style, evaluating whether visitors have a hierarchical or egalitarian view of society, or whether they think in terms of what is good for the individual or what is good for the collective. Someone with a hierarchical view of society might receive loan advice from someone in a position of authority, while someone with an egalitarian view might receive advice from a peer. Similarly, a person's tendency to think individually or collectively might influence which features of a product are most emphasized. If that experiment goes well, Urban says, he envisions global companies one day using website morphing techniques to build single websites that can adapt to users based on their cultural background, as well as on their cognitive style. The researchers are also working on using their morphing techniques to make banner ads more effective.