How about a cowl neck sweater to go with that sharkskin flannel skirt?
J.Crew this season added suggestive selling features to its retail Web site, jcrew.com. As visitors view items, J.Crew displays complementary products, hoping buyers will like its ideas and buy more stuff. The suggestions are based on shopping trends gleaned from store, catalog and Web sales.
Behind the scenes, information is crunched in a data warehouse J.Crew built with software vendor digiMine. The data warehouse – which runs on a Microsoft SQL Server database – took six months to develop. It was at 500G bytes last month and continues to grow, says Jayson Kim, senior director of Web marketing applications for the retailer.
The data warehouse pools not only Web site traffic data but also transactional information from J.Crew’s retail stores and catalog sales operations.
For a multichannel retailer like J.Crew, Web site metrics and click-stream analyses don’t capture a full view of its customers, Kim says. Web reports, such as the number of page hits or geographic distribution of visitors, “are geared toward the IT and operational side,” he says, “but they are not very useful from a marketing standpoint.”
“DigiMine is good at analyzing click-stream data and integrating it and overlaying it with other customer data,” he adds.
DigiMine’s Enterprise Analytics suite handles the data mining and analytic processing for J.Crew. The software helps the retailer determine product affinities, such as which pants and shoes customers tend to purchase together.
When a customer selects an item, J.Crew’s Web site suggests complementary items based on predetermined trends. The computations occur offline. Then, the digiMine software feeds the data to the Web site applications, which present product recommendations based on a visitor’s browsing behavior.
It’s as near to real-time as is practical, Kim says. True real-time analysis is computationally intense and would slow site performance, he says.
In the future, J.Crew will use the data warehouse and analytic software to analyze sales trends and create targeted e-mail marketing campaigns.
With transaction data from multiple legacy apps to tie to its Web data, J.Crew knew customization was unavoidable. The retailer took the time with digiMine to develop proprietary schema; if instead they had tried to use out-of-the-box data mining software, J.Crew would have had to take shortcuts that in the end would have limited the project’s effectiveness, Kim says.
“You can’t build an enterprise-wide, scalable data warehouse on a terabyte of data by taking a shortcut,” Kim says. “Starting from scratch is better than taking a third-party product and overlaying it on your data and hoping it works.”
So far the cross-selling features seem to be working well, but J.Crew won’t have hard metrics until after the holiday season, Kim says.
By: Ann Bednarz