Saint Paul, MN, cataloger Rivertown Trading Co. offered its customers upsell items specifically targeted to the individual caller this holiday season, as part of a test of software newly-designed for call centers.
The software, produced by Net Perceptions, Minneapolis, is a telephony-based version of technology that has been previously used by online retailers. Just as Web retailers, such as Amazon.com, tell customers who are purchasing a book what other books buyers of their selection have bought, the new program, Net Perceptions for Call Centers, allows call center agents to do the same thing.
“What it does is filter out vast amounts of data rapidly in real-time mode. The software allows vendors to look at consumer relationships to products and assign appropriate preferences,” said Prakash Puram, general manager, consumer and retail marketing for Net Perceptions. “It will look at you and try to compare you to other people similar to you and draw inferences about what your likes and dislikes may be.”
For Rivertown Trading, this represented a shift in upsell strategy.
“Before we had selected products based on the product's sales history rather than on customer preference,” said Donna Avery, Rivertown Trading president.
While complete results of the test, conducted on three of the company's catalogs–Daily Planet, Signals and Video — were not available, preliminary impressions were positive.
“The overall sense we have is that it has been able to give a significant lift to our ability to add dollars to sales through upselling products that are more targeted,” Avery said. “The group lens allows us to highlight products that seem to be a good fit.”
Test runs of the product by prospective clients have resulted in 30 percent to 40 percent increases in the number of upsells, and in one case a doubling of the number of upsells, according to Net Perceptions chief technological officer Paul Bieganski.
The program is purely mathematical in its analysis of products that a customer has purchased. It studies the mix of products ordered, and in the case of a repeat customer, merges that with the history of what the customer has purchased in the past. It then draws from product mixes closest to the mix that the customer has ordered when making a recommendation. As a result, the upsell items may vary significantly from what the customer has ordered, Puram said. So for example, if previous customers have purchased several candles and a pair of socks, the computer might recommend socks to a candle buyer, though that might not be the most logical choice.
“Our software is not concerned about what the products are, its about what the product mix is,” said Puram.
Puram defends the computer's selection of what might on the surface look like illogical choices by noting that people have varied tastes and may not always want related items.
“When I worked at IBM I bought a lot of white shirts because I had to wear them for work, but that is not the only thing I liked to buy. On my days off I wore completely different things,” he said.
While the computer program does not concern itself with the nature of the product, it can be taught certain characteristics of products. For example, it can learn that candles are replaceable items and therefore people who buy them may need to replace them at a future visit.
For Rivertown Trading, these back end computations seemed like not only a way to increase sales but a way to add value to the customer communication.
“We are always interested in providing better service to our customers. Understanding them better and knowing better what their needs are is one way to do that,” Avery said.
At Rivertown Trading, customers are not told that the products they are being offered as upsell have been selected for them by a computer. The company felt it would be simpler to offer the products as regular upsells.
Early indications show that the software will most likely be continued after its initial test.
“Preliminary impressions are that we will continue to with this method in the future,” Avery said.
