Digital Bloodhounds: Web Users Follow the 'Information Scent'

Computerworld - Palo Alto Research Center Inc. (Parc), formerly Xerox Parc, is conducting research that it claims can help Web designers and content providers avoid much of the expensive and time-consuming trial-and-error process that marks many e-commerce projects. Using equations borrowed from biological models, Parc is studying "information scent" (download PDF). The notion is that users navigate from page to page based on cues -- snippets of text or graphics -- that they see as they go along. They tend to click on the link that emits the strongest information scent, which is based on the value and proximity of the information thought to be obtainable from that link.
Ed Chi, a Parc research scientist, says the probability that a user will follow a given path increases exponentially with the strength of the scent from that path. If users follow low-scent paths, "that would represent choices in which they are essentially confused," he says.
Based on these concepts, Parc has developed two tools. One of them analyzes what users are currently doing on a Web site, revealing their goals and how well they're being met. The other simulates user activity and can be used to predict the usability of alternate designs. "We envision that a company would use the analyzer first to see what users are currently doing, and once they discover that, use the simulator to help them understand whether the site is actually satisfying user demands," Chi says.
He adds that better tools for Web activity analysis and prediction will be needed as online retailing sites transition from an information-push, or page-based, model to an activity-based model. Traditionally, retailing Web sites have simply provided access to specific products requested by the user. Newer ones are more likely to suggest products based on a knowledge of customer attributes, customer transaction history or insights gained by prompting the user for information.
"The implication for Web site designers is pretty big," Chi says. "Instead of pages with static content, these pages are served up with dynamic content depending on the user's goals. So it's harder to figure out collectively what all your users are trying to do on your Web site."