BI, in good times and bad

Considerable attention is now focused on peer ranking as a means for identifying the weakest performing assets in a business operation, according to LaRow. In that type of analysis, performance of one facility is compared to another. Customers are also compared; in a financial services environment, those who are at the bottom of the profitability spectrum might get bumped. Also, triggers can be used to send an alert if a metric such as a cost ratio exceeds a specified threshold, so that warning signs are noticed.

Customer insight

The hospitality and entertainment industries have also been affected by the economic slowdown, because discretionary spending accounts for the majority of revenue in those sectors. BI can help by identifying promising customers, thereby allowing for more accurate marketing, and making business operations more efficient.

Harrah’s Entertainment, a gaming and hotel enterprise, has a successful rewards program with 10 million active customers (defined as those who have visited a facility within the last 12 months). The company uses software from SAS to analyze and predict customer behavior. It’s a longtime user of business analytics, having begun to segment its customers based on a variety of factors since 1998.

"We use SAS predictive analytics to model data contained in a very large Teradata warehouse, and determine the frequency with which guests are visiting our facilities, how much are they spending and what factors are affecting their behavior," says David Norton, senior VP and chief marketing officer at Harrah’s. After analyzing historical data, Harrah’s develops a model, and then can predict the likely behavior of other similar customers.

"We can offer incentives that have a high probability of success," adds Norton, "and minimize marketing costs as compared to a more general, less well-targeted campaign."

How well do predictive analytics work when circumstances are changing and outcomes may not depend on historical trends? According to Steven Pinchuk, general manager of profit optimization systems at SAS, projections can be modified by incorporating information from outside the data warehouse. For example, an estimate could be made of the degree to which higher gas prices would affect travelers’ decisions, and used to predict a new trend.

"Most forecast models use a single technique, such as a time series or linear projection, and use the same one for all situations," Pinchuk says. "SAS Forecast Server has over two dozen different techniques, and it automatically applies the best model, depending on the available data."

Analytics for law enforcement

Government is equally under pressure to respond to shrinking budgets and make the best possible use of resources. In Richmond, Va., the police department employs Clementine, a predictive analytics and data mining solution from SPSS, in conjunction with a BI solution from IBI. The system uses predictive analytics to optimize the deployment of police officers, identify minor crimes that could become major problems and expedite investigations.

Clementine is run against large data sets to detect patterns in dispatches and identify areas where more expensive tactical units are likely to be needed. In addition, by analyzing data from criminal offender databases, the Crime Analysis Unit (CAU) found a correlation between some types of property crimes and subsequent sexual assaults. In addition, the CAU can analyze large amounts of data to solve cases such as drug-related homicides more quickly.

SPSS’ predictive analytics tools are often used in conjunction with other BI solutions that do not offer predictive analytics to provide additional capability. For example, it has a partnership with MicroStrategy’s products to extend their functionality into the predictive arena. SPSS also partners with Business Objects to OEM its predictive analytics and data mining technology as part of the Business Objects XI platform. In one Business Objects application, data identifying accounts that might go into collections within the next six months is presented on a dashboard.

"We do a lot of fraud detection in both the commercial and public sectors," says Patrick McCue, VP for global alliances at SPSS. Its software is also being used by tax collection agencies worldwide to identify those tax delinquents from whom recovery of overdue taxes is most likely to be possible. "The mortgage industry may have saved millions of dollars if they had used predictive analytics, for example, and followed through on what they had learned," McCue adds.