Features

Gregory PS: At recent data mining conferences there seems to be a split between researchers and industry people, and relatively little communication between the two groups. You have been on both sides of research / industry fence and were able to transfer research ideas to successful products. What can you suggest to researchers to help them better appreciate industry problems? How can industry people keep track of the relevant research ideas?

Ronny Kohavi:
I�ve been fortunate to work at companies that allowed me to do research and publish, while working (Amazon being the exception). The key was that the research was close enough to the job that it was worth doing, and getting peer feedback led to improvements that could not have happened otherwise.

The first thing I�ll say is that most experienced managers will usually support such activities if the employee makes it clear this is something they want to do, and how it can benefit the company and the employee. The KDD Cup was initially debated at Blue Martini because we had to work through legal issues, and then spend hundreds of hours releasing the data, judging it, and writing papers. However, the opportunity to have lots of researchers look at the data and uncover insights, and the possibility of helping the brand were enough for the company to support this. Netflix�s prize is a great offer by Netflix that helps bridge the gap between research and academia by providing a real problem with practical implications. The value and publicity they�re getting is well worth the money.

I have always liked working in industry, where the problems were real and solving them means having some business impact. I believe that researchers should team up with industrial partners more, and that companies should recognize the value of releasing sanitized data to help researchers help them.