Predictive analysis scrutinizes data to predict, say, the likelihood that a customer will respond to a sales offer. Text mining searches through reams of unstructured text, such as electronic documents and E-mails, for key concepts. SPSS has offered predictive analysis through its flagship Clementine software for several years, and last year it acquired text-mining software when it bought LexiQuest Inc.
Unstructured data is believed to account for as much as 80% of all data used within businesses. But most data analysis today is conducted on structured data stored in tables in relational databases.
Customer-churn analysis and other customer-relationship-management tasks such as cross-selling and up-selling are expected to be the primary applications for the new Predictive Text Analytics product, SPSS says. Marketers, for example, can use the software to analyze written notes taken by call-center workers during conversations with customers to pinpoint reasons customers are defecting to competitors.
With the software, marketers can judge the overall satisfaction level, price sensitivity, and service frustrations of a company's customers. That analysis can be combined with structured data, such as phone billing records, to identify customers who might accept a new service offer or who might be about to switch providers.
While text-mining tools are generally designed to search through large volumes of data, SPSS has tuned the new product to also work with short strips of text, such as notes taken by call-center employees, customer E-mails, and written responses to Web surveys. That's more difficult to do because less text means less linguistic context for the software to work with, says Peter Caron, SPSS's customer-analysis product-management director.
IDC analyst Susan Feldman is unaware of any other software that combines predictive-analysis and text-mining capabilities. She says the SPSS software can help call-center workers adjust their scripts to respond to specific customers. "That, in real time, is very valuable because losing a customer is a very expensive proposition," she says.
The Abramson Cancer Center of the University of Pennsylvania is using the software to search large volumes of medical literature to identify new symptoms associated with different types of breast cancer, says Michael Liebman, director of computational biology and biomedical informatics at the center. The software's predictive-analytical capabilities are used to analyze medical data for patterns and judge how effectively the information can be used to diagnose the disease. "It's been showing some pretty interesting results and is taking us in new directions," Liebman says.
Predictive Text Analytics is available now, starting at $100,000.