Baynote’s technology tracks the behavior of Internet users and harnesses the wisdom of the crowds to guide future visitors to the right information. For Baynote’s CEO Jack Jia, contextual targeting beats behavorial targeting any d

ay: it is more efficient with none of the privacy concerns attached to BT.   What is the premise behind Baynote? As CTO of Interwoven, I helped people publish content on the Web. But the more content they published, the less value they felt it was generating and the more people got lost. There is a statistic out there that says 95% of people abandon a site within three clicks, whether it is an Intranet, an e-commerce site and a support site. I felt there had to be a better way. In the real world, we don’t find information by looking at a list. We ask our peers and family. But on the Internet, we don’t see another soul. If you are lucky and find something, that success is forgotten. There is one solution which is to bring communities together to harness the wisdom of the crowds. The problem is that only 1% of users are willing to participate and they are not representative. They suffer from survey bias.   You talk instead about the wisdom of the invisible crowds. What is that? Because actions speak louder than words, the idea is to watch what people do. Individual ants can’t necessarily find food. But when one of them is lucky, it goes back to the group. A few more ants test the food. If it is good, a chemical trail tells the whole group where to go. Baynote is that chemical trail for Internet users. With a technology that sits on top of the site, we track 24 behaviors for every single user. That information is processed by our engine which mimics human intelligence. After a few days of learning, the engine can make recommendations to new visitors in real time.   Can you walk us through a customer case? We launched in March 2006 and we now have over 60 clients. One of them is US Appliance, a large online appliance company. When a user starts searching for side-by-side fridges, it narrows down the choices and shows recommendations based on what other users have looked for. Using Baynote, they were able to increase their net revenue by 34%. Other clients have seen lifts between 18% and 107%.   How are you different than behavioral targeting? BT tracks your last few visits and it thinks that past behavior can predict future behavior. But that doesn’t work. If I bought a toy on Amazon for a friend, it won’t work to offer me diapers. We saw in 2005 that personalization failed. It is not a matter of getting more data. Academic research shows that it doesn’t help to have your profile in terms of demographics, revenue bracket, etc. It is only right 18% of the time. Contextual targeting is right 70% of the time because we understand your true intent. If you type “broken pipes” on Home Depot, we understand we are looking for glue, connectors or books on the topic.   What about privacy concerns? BT has privacy issues. Even though it is anonymous, it is a particular person they are tracking. But we don’t care who you are, we don’t need to profile you. It is all blended in. There is no problem with privacy.    Are you adding any new functionality? We have added what we call UseRank. Page search is not good. Within a page, PageRank does not work because there are no humanely-added links. But we have the users! We work on top of the search engine to rank the results based on user activity. For US Appliance, UseRank has increased conversion rate by 1,000%. We also have a new function to reshuffle ads according to their effectiveness. Isabelle Boucq FEEDBACK For comments on this article, email us at