When James Copell wants to look for evidence of trends, he fires up an application called iQuest, from iQuest Analytics Inc. in Rehoboth, Mass. The application combines social-network analysis techniques with text mining and categorization techniques. Copell, who runs Copell Financial in New York City, a boutique investment research firm, uses this tool to simultaneously search through a variety of data types, including internal corporate documents, e-mail and voice-mail logs, and external data from the Web. It then shows him sociometric diagrams with results such as the most relevant topic—or most important person or event in a topic—and everything it’s related to, over time.
For instance, Copell could choose to follow a product launch and see what kind of response it’s generating around the Internet (where search engines such as Google rank results by relevance, not by most recent). This could aid him in predicting that the product’s sales will help its company exceed expectations (imagine, for example, tracking the spread of the iPod over time). Or he can use the tool to find unusual patterns, such as whether a shortage of cranes in the Middle East means anything for oil production.
“I read something about how they were building so many chemical plants in the Middle East that they had run out of cranes, and I wanted to know why they were building the plants,” Copell says. He also wanted to know if the high number of chemical plants meant that the oil industry didn’t need to build processing plants, perhaps signaling it was at peak capacity, which in turn might affect oil prices over time in a way that would matter to stock pickers. His firm designed a series of searches that it executed over several weeks to explore what was on the Web on the topic. He says iQuest also helps show correlations, such as how crane shortages relate to oil prices, or even Middle Eastern arms sales.
In the past, if you wanted to predict trends based on data, you tended to hire statisticians. Or perhaps you bought a package such as SAS. Says iQuest CEO Joseph Rosenthal: “You’ve always been able to pay a McKinsey a million bucks to get this kind of work done,” but an iQuest license might cost $100,000.
Predictive analytics tools are leaving the domain of the statisticians and coming down to where companies can afford to put them in the hands of departments. Mark Bryant, market analysis manager at Patterson Dental, a dental products distributor in St. Paul, Minn., says these tools already offer useful features that “you don’t need a Ph.D. in mathematics to understand.”
Patterson Dental uses Loyalty Builders, a marketing package from its eponymous vendor in Portsmouth, N.H., which has a feature that examines buying data and predicts what a customer is likely to buy in the next 90 days. Bryant says Loyalty Builders can give useful predictions about customers and sales for what he calls a much more attractive price point than was available in the past. He predicts that within five years, “if you’re not using data and predictive modeling characteristics on your customer databases, you’re going to be behind” your competition.
But Dan Vesset, an IDC analyst, cautions that predictive analytics tools are still not very widespread in general packaged applications. “We’re still waiting for packaged functionality, where various applications come with predictive tools,” he says. IDC says predictive analytics makes up about 20 percent of all BI spending, and should grow slightly faster than the overall BI market for the next few years, with an expected average of about 10.5 percent. —M.F.
