Sometimes sophisticated data mining can't be done with out-of-the-box software and you have to invent the process yourself. Of course, it helps if you're a multimillion-dollar company with loads of resources and expertise.
Hewlett-Packard Co. has begun an initiative to upgrade its advanced Web analytics capabilities to identify natural affinities among the many products it sells.
"We are trying to develop an analytic framework that provides recommendations during an online purchase and to our call centers that identifies natural up-selling and cross-selling opportunities," said Seth Romanow, HP's director of WW analytics for eBusiness and customer operations. "Or in layman's terms, we want the ability to ask customers if they want fries with their burgers."
HP wants to increase the average online order size for its products. "For HP, up-selling would be adding anything that would enhance value to a PC, such as increased memory, a DVD drive or video card upgrade," Romanow said. "Cross-selling would be adding a monitor, docking station, digital camera, printer or even extra ink to a laptop computer purchase."
The manual pilot began with HP's small and medium business channel in November, mining primarily aggregate shopping cart information. "It showed us what items are most often paired together," he said. "For instance, we found that lightweight laptops were being paired with significantly different products than a desktop replacement laptop."
HP then built banner ads along the purchase path that would suggest the most complementary products and highlight upgrades, all in real time, Romanow said. "We also provide those same up-sell and cross-sell recommendations to our call center, as well as identify those affinity products that had the highest margins while making sure they were in stock," he said.
In the near future, HP wants to bring in customer data such as company profiles and purchase histories to build a more accurate picture of the customer and make better recommendations.
"The more data points we can provide, the more likely we'll be able to increase the dollar value of the sale," Romanow said. "The real value is being able to deliver these recommendations with high predictability and also generate a higher level of customer satisfaction, because having the right configurations or peripherals will enhance their experience with the product."
Results so far have been impressive: Overall product attach rates have increased 14% to 17%, and average revenues have increased from 13% to 63% depending on SKUs, he said.
Collaborative selling will be one possibility in the future. "And we'd be able to share the information with our business partners and resellers so they can boost their sales, too," Romanow said. "It's all about selling the right stuff to the right people at the right time."