Friday, April 10, 2009

A Renaissance? Not Exactly.

Gary Orenstein posted an article on GigaOM this morning titled "The Data Mining Renaissance" that elicited a comment from me about the proper characterization of the evolution of the data mining market.  While I agree that inexpensive data mining platforms based Google-like technologies are proliferating, the renaissance has been in the economics of data mining rather than the capabilities of such systems.

This is an important distinction because much of the interest that has been driving the uptake of large-scale data mining technology is in many of the same kinds of applications that are poorly served by the existing technology. It is in many ways a self-limiting resurgence. Much of the interest in data mining technologies is contextual targeting and preference modeling, which to do well on a large scale require overcoming the Achille's Heels of current data mining technology, poor real-time analytic performance and inadequate support for high-dimensionality data models. 

In other words, the "renaissance" is that expensive data mining technologies are becoming less expensive over time, which is not exactly a remarkable observation in the technology industry. Unfortunately, what is really needed is data mining technologies that are better. I fully expect we will get these better technologies sooner rather than later, after all that is my business, but when that happens it will look more like a revolution than a renaissance as it will enable capabilities that are currently unavailable at any price.

No comments:

Post a Comment