The New Borders Wine List and Product Recommendation Engines
BackBy Kelly Shermach — October 2008
If there’s one thing Sandy Aldrich has learned, it’s that bookworms enjoy their time in the kitchen, too.
The senior manager of strategic partnerships at Borders Group Inc. said the success of the company’s in-store Cooking Destinations sections — which offer nonbook items, kiosks for book recommendations and cooking demonstrations and discussions led by local food artisans and organic growers — led the company to seek other opportunities leveraging its existing database. In June, Borders partnered with online wine purveyor Vinesse to launch the new Borders Wine List.
“It’s not that much different than, ‘We thought you’d be interested in this book,’” said Toronto-based Emma Warrillow, president of Emma Warrillow & Associates Inc., a data intelligence and strategy consultancy. “The recommended books [function] can expand or stretch to other products and services with slightly different technologies and tools.”
These recommendation engines used by companies such as Borders, Amazon.com or the multitude of e-tailers simplify the statistical research and make data use more immediate, especially for firms holding large data resources.
“We’re seeing more of it as we see companies looking at the asset they have in the database,” Warrillow said. “Certainly there’s a lot of information in those databases.”
A bookseller such as Borders has data on the number of books each registered consumer has purchased, the amount each has spent on Borders’ products and some demographic information asked of the consumer on the program application or since. The latter may prove most valuable in targeting value adds to members of the database.
To generate such a targeted list, database owners may use predictive modeling, more simplistic member profiling or cluster database tests, which are manual member groupings. Recommendation engines typically rely on collaborative filtering software built on predictive modeling.
Collaborative filtering is similar to predictive modeling but does not integrate the comprehensive consumer picture. For example, it neglects the highlighted consumer’s past activities, purchases, etc., that may showcase characteristics of success in deepening the relationship and gathering more relevant and timely data for the database as a whole.
Predictive modeling is the best practice in the case of separate third parties trying to generate a profit. At Borders, the third parties build their own databases from the introduction of the senior counterpart’s members to its new associate.
“Borders is the entry place,” Warrillow said.
Vinesse maintains the Borders Wine List database, meaning customers who sign up become customers of Vinesse, Aldrich said. For a variety of reasons, Borders does not maintain the database itself.
“How much value is Borders going to have from that data? How would it use that data?” Warrillow explained. “Wine consumption is different from book consumption.”
Kelly Shermach is a freelance writer based in Chicago, Ill., who frequently writes about technology and data security. She can be reached at firstname.lastname@example.org.