The second of a two-part series. Read part one here .
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
There are a number of obstacles standing in the way of broader adoption of predictive analytics.
Building the predictive data models that underlie predictive analytics is a complex process that requires significant expertise. It's a job mainly for highly trained statisticians, who are in short supply, according to Kurt Schlegel, an analyst with Stamford, Conn.-based Gartner.
Also holding back the adoption of predictive analytics is the fact that it requires moving data sets from BI and data warehouses to a separate application. This process can take significant time and manpower and can sometimes lead to dropped data.
Vendors are addressing the latter problem by beginning to integrate predictive analytics technology into BI and data warehouse platforms. That way, users can perform the analysis in the same environment where they already access BI reports and perform queries.
The benefits include removing the need to transfer data from one app to another. Because of that need, many companies using predictive analytics would run the technology only against a small but, hopefully, representative sample of their data. Moving large data sets to a separate predictive analytics application was impractical.
Now, with the predictive analytics technology in the BI or data warehouse platform, it can access more data and, theoretically, provide more accurate results.
Information Builders' WebFOCUS RStat 1.2, released in early February, is probably the most notable recent example. Netezza is also laying the groundwork for embedding predictive analytics in the data warehouse, announcing on Monday that users now have the ability to build predictive models to run against large data sets in the Netezza TwinFin appliance.
IBM, meanwhile, is continuing its efforts to integrate SPSS technology with the Cognos BI suite.
SAS may be in the best position, however. The company has slowly but surely built itself a reputation for having a top-flight BI suite, with advanced analytics built in natively to the platform. The vendor counts a number of high-profile companies among its BI customers, including CIGNA HealthCare and Credit Suisse, and it has been placed in the leaders' quadrant of Gartner's BI Magic Quadrant reports for the last several years.
Referring to IBM's ongoing efforts to integrate SPSS advanced analytics with the Cognos BI suite, Ken Hausman, product manager at SAS, said: "We have problems, like any company, but that isn't one of them."
Boris Evelson, an analyst with Cambridge, Mass.-based Forrester Research, agreed, noting that SPSS has its own user interfaces and meta data schema that IBM will have to reconcile with its own. "It's a separate product with a full footprint," he said.
Better visualizations needed
But in order for predictive analytics technology to truly break through to the masses of information workers, BI and data warehouse vendors are going to have to simplify the way predictive models are built and visualized, according to Jim Kobielus, also a Forrester analyst.
Specifically, vendors are going to have to develop drag-and-drop features and wizard-based model building capabilities, Kobielus said, so that users don't have to write the complicated algorithms long needed to perform predictive analytics.
"When the dialogue gets to math, that's when people's eyes glaze over," he said.
Easier-to-understand visualizations are also needed to let non-statisticians make sense of the resulting analysis. While vendors are taking steps in both directions, analysts and customers agree that more work needs to be done.
"I'd love to see more visualization," said Chris Brady, who is CIO at Dealer Services, a financial services firm for the auto industry that uses WebFOCUS RStat to predict the lending needs of its customers.
Until then, predictive analytics could remain largely relegated to a few highly trained statisticians at the few companies that use the technology.
"The user-friendly side of it, that's really the next frontier," Kobielus said. "All power to the vendor that can put the power of predictive tools into the hands of information workers."