The Value of Analyzing Customer Data

ANALYZING CUSTOMER data to increase revenues is nothing new. Online companies such as Amazon.com analyze customers’ past purchases in order to decide what merchandise to try to sell them in the future. And brick-and-mortar retailers have long identified sales relationships among products and positioned them near each other to sell more of both.

But many organizations have stayed away from data analysis because it can be very costly. However, application service providers that offer to handle everything from data collection and storage to data analysis are popping up all over, says Michele Rosenshein, a former e-commerce analyst at New York City-based Jupiter Media Metrix. You’ll find Primary Knowledge based in New York City, Coremetrics and WhiteCross Systems in San Francisco, and WebTrends (recently merged with NetIQ) in Portland, Ore.

Coremetrics and WebTrends both provide JavaScript tags that clients place in a webpage’s HTML code. Each data tag has predefined variables, including unique visitor identification, site referral, product or content browsing, shopping cart actions and order processing. The tags relay data directly to the ASP’s servers, where analytic software generates standard reports for the client. Primary Knowledge, on the other hand, has clients store data locally and periodically send that data to Primary Knowledge’s servers, where software then analyzes it. All three companies also have statisticians on staff who can produce customized reports on demand. Other companies, such as WhiteCross Systems, specialize in more complex operations, including neural networking.

Rosenshein predicts that during the next one to two years 65 percent of websites will move to outsourced data analysis.

Still, she warns that these services may not be for everyone, especially sites generating 500,000 or more transactions per day or multichannel retailers with significant internal data warehousing investments. Other companies will find that complex analyses require constant adjustments to mathematical models and that it doesn’t pay to outsource the operation.