Use of Collaborative Filtering

The Internet has opened up a range of new opportunities for personalization that may lead many catalogers to reconsider their attitude toward database marketing. Amazon has shown us the way. Amazon uses emails and its Web site to suggest books and movies to customers that they might be interested in, using collaborative filtering to come up with the suggestions.

Collaborative filtering is provided by www.Netperceptions.com, among others. It consists of software that makes use of a client’s database of customers and their previous purchases. The goal of the software is to determine what a person’s lifestyle and preferences are and then to find other customers within the database who have similar tastes. Whatever the cataloger had sold to those similar customers might appeal to you, so the cataloger suggests it to you. One of the most convincing uses of collaborative filtering was an experiment tried with Great Universal Stores (GUS), the largest cataloger in the United Kingdom. Before collaborative filtering, GUS customer service reps were getting a 20 percent cross-selling rate on orders placed from the paper catalog. That means that in one call out of five, the rep was able to sell the customer an item from the catalog that the customer had not intended to buy.

Netperceptions.com introduced collaborative filtering into the GUS call center software, using the huge GUS customer database. The suggested next best product showed up on the customer service rep’s screen as soon as the rep typed in the first purchase requested by the caller. The company’s previous success at cross-selling was due to the customer service rep’s native intelligence. If a woman ordered a dress, the rep might suggest matching shoes, a belt, or a handbag. The suggestions from the collaborative software were described by the GUS reps as “spooky.” The software might suggest selling towels to the woman who ordered the dress. Towels? What was the relationship of towels to the dress? There was none—except that the software had found similar women in the GUS database and had found that many of them also bought towels. The amazing thing about the software was that it worked. After getting used to it for 6 weeks, GUS customer service reps were getting a 40 percent cross-selling rate—double what they had been getting before. Now that is database marketing, folks. Collaborative filtering is used by more than 200 American companies. It requires a large database and super fast software. It is not for the mom-and-pop catalog operation, or for the timid.

In the past 2 years, I have probably spent more than a thousand dollars with Amazon, some of which was the result of Amazon’s collaborative filtering suggestions. One email that I received a year ago has stuck in my mind. It said:

Two years ago you bought Dark Sun by Richard Rhodes. You may be interested to learn that Richard Rhodes has just had a new book published called Why They Kill. To learn more about this book, or to order it, click here.

Well, I clicked, I bought it, and I loved it. This new book had nothing to do with the previous book except the author. This is database marketing.