Rental recommendations
If you enjoyed "Chicago," will you like a dance documentary? Maybe.
Robb Hecht rents mostly dramas and science-fiction flicks from Blockbuster.com. So he was surprised when the movie service recommended "Rize," a documentary about an urban dance style -- and even more surprised when he liked it.
That makes Hecht, an adjunct professor of marketing at New York's Baruch College, a success story for Blockbuster Inc., which has revamped the software on its Web site to persuade customers to rent more movies.
Previously, Blockbuster.com's recommendations relied heavily on the obvious: recommending another Jack Nicholson film or horror movie to somebody who had just watched "The Shining," for instance. But with newer software it began using about two years ago, the site now looks at subtler factors -- such as whether previous movies a customer rented were mindless or challenging, or whether they were plot-driven rather than character-driven.
Online purchase recommendations have been around for years, but they've often generated complaints for being simplistic or just plain wrong. These days, some Web sites are relying on increasingly sophisticated techniques to help steer customers to additional products. The system Blockbuster uses, developed by ChoiceStream Inc., works by assigning characteristics to an item and searching for other items that share the same characteristics. By combining that data with older techniques -- like looking at past purchases, or seeing what other people who bought the same item also bought -- it comes up with a suggestion.
"Creating this service that knows you, in this market of superabundance, is probably the single most important way of building loyalty," says ChoiceStream CEO Steve Johnson.
Hecht was steered to "Rize" by Blockbuster.com because he had at least one other musical in his queue ("Chicago") plus at least one other documentary ("March of the Penguins"), according to ChoiceStream. Combining that with the fact that, based on past rentals, Hecht enjoys edgy, urban movies, it recommended "Rize."
Since adding the software, Blockbuster says it has lost fewer customers, in percentage terms, to rival services, and the number of movies in the average customer's "to watch" list has grown by almost 50 percent.
Basing suggestions on what other customers who bought or rent that product did requires massive amounts of data. The technique falls short when a product is new.
Alternatively, basing recommendations purely on past purchases can backfire when the purchase was a gift -- for example, when a person who once bought a fishing rod for someone else is constantly nudged toward tackle boxes.
Amazon.com, some of its customers complain, is one site prone to this sort of faux pas. The retailer says it has worked hard to improve its recommendations by asking customers if their purchase is a gift so it doesn't get factored into the recommendations, for example, and encouraging customers to rate items they have bought so Amazon can improve its suggestions.
Not everyone is crazy about this flood of personalized recommendations, though. Linus Kafka, a lawyer turned historian who lives in Los Angeles, calls them "creepy" and delights in trying to trick the software.
"I don't like being targeted for marketing," he says. "I find it a compromise of my own privacy. I think they should sell me the item, and the relationship should end there."