Number-crunching has never been easier thanks to powerful PCs, software and servers.
But what data gets crunched and how has never been more difficult to decide.
Business intelligence goes beyond spreadsheets to look at what the company has done – in other words, it looks backwards. Analytics looks forwards – what could the company do.
That’s what Mississauga, Ont.,-based Maritz Loyalty Marketing wanted over a year ago when it began looking for a data mining solution. The company, a division of St. Louis-based Maritz Holdings Inc., specializes in designing and implementing loyalty solutions for North American organizations — for example, point or gift programs for retailers or credit card companies.
Not only does Maritz create the programs, it also helps customers make the best use of them. As a result, it often has large amounts of customer data it has to wade through.
While Maritz already has SAS Enterprise Guide for analyzing data, said Maria Pallante, the division’s director of customer research, it wanted a data mining tool for greater depth.
“We’re always looking for enhancing our offering to clients,” she said.
Data is the cornerstone of a loyalty program, she explained. What staff needed in particular was the predictive modeling capabilities that data mining tools include, plus the ability to handle large data sets.
“For a lot of people data mining is the on-ramp for predictive analytics,” says Evan Quinn, a senior principal analyst at Enterprise Strategy Group who specializes in data management and analytics. Data mining lets organizations can look for patterns or clusters of data to the help forecast certain things.
For example, Pallante said its corporate customers want to be able to identify which of their customers are more likely to respond to a marketing initiative, or which customers are at risk of turning to other products.
Data mining tools increasingly are included as part of suites of products offered by companies ranging from SAS, IBM Corp., Business Objects and Hyperion, Quinn said. In a recent study he found 61 business intelligence suppliers, and about a third of them offer data mining capabilities.
In Martiz’s case, after evaluating a number of solutions on a scorecard it chose SAS Enterprise Miner. A year and a half later Pallante is satisfied with the choice in part because of the software’s strong modeling tools. Data mining helps her company advise corporate customers on the right types of solutions based on previous transactions of their customers.
Quinn says the trend in new versions of data mining solutions organizations is to include visualization tools that help present data in graphs or charts to non-technical staff. Increasingly data mining solutions are becoming “more business analyst-friendly,” he says, that don’t require a data scientist to manipulate.
They also make more use of open source-related technologies such as the R statistical programming and language environment, which is taught in universities. Data analysts are more comfortable with it than a proprietary coding language some software vendors rely on. Also, look for software that leverages predictive model markup language (PMML) for defining models. PMML lets users share models across a number of applications.
Don’t look for merely a data mining solution, Quinn adds; instead find an analytics platform that offers a number of data-related capabilities. Data has to be cleansed and loaded before being mined, he points out, then analyzed and then presented in a way that business managers can use.
“Any reasonably-sized company on the planet should be doing things like data mining,” says Quinn. “Their competitors probably are”
