When it comes to personalizing the web experience, Bookspan LLP has both an advantage and a disadvantage. Its advantage is that customers can’t come onto the web site unless they register, so Bookspan knows a lot about them before they even do anything on the site. The disadvantage is that information about customers resides in a legacy database so Bookspan can’t always deliver up personalization while the customer is still on the web site. “We know a lot about our customers already-their purchase history and their account history,” says Steven Brita, director of Internet marketing for Bookspan, which operates book clubs. “But we need to make sure that when they are online the personalization we provide is as accurate as possible with the legacy data.”
Bookspan, a partnership between Bertelsmann AG and AOL Time Warner Inc., is in the enviable position of knowing much more about all its customers than most retailers. Its properties include such well known properties as Book of the Month Club, the Literary Guild, the Doubleday Book Club, the Quality Paperback Book Club, as well as specialty units such as the Mystery Guild and the Science Fiction Book Club. As the operator of all the major book clubs, it knows the preferences and buying histories of many consumers.
The Holy Grail
And so it’s in a position to deliver personalization on the web site to a degree that others are not. But Brita says the company has a way to go before achieving the level of personalization it desires. “The Holy Grail is to deliver personalized recommendations while the member is on our site,” Brita says.
Bookspan’s personalization efforts are reflective of how the online retailing industry has struggled with personalization. The industry generally perceives value in personalization, but many retailers can’t deliver truly personalized results and for still others, personalization is a lower priority. “Retailers’ databases still don’t have the breadth and depth they need,” says Ken Cassar, retail analyst with Jupiter Research Inc. “And they still have not integrated offline and online data.”
Today, most online personalization is the result of collaborative filtering, whereby a retailer aggregates data from similar buyers, sees what they’ve bought that other members of the group have not bought, then recommends those products to members of the group who have not already bought them. It’s familiar to most shoppers through Amazon.com’s “people who bought that also bought this” approach.
But even that method, Cassar says, is not widespread. “A lot of people made investments in collaborative filtering in 1998 and 1999, but it’s hard to find active deployments today,” he says. Part of the reason for the delay in widespread use of personalization software is that retailers prefer to rely on merchandisers to make recommendations, rather than software. “Retailers don’t want algorithms to drive their business,” Cassar says. “They want merchandisers who know the products and have instincts that they trust.”
Nonetheless, there are certain segments where software-driven personalization can work, Cassar says, and one of those is books. “Personalization will work in categories where there are a lot of products and preferences are driven by opinions and not by any quantifiable attributes of the product,” Cassar says.
Right info to right person
That’s exactly the direction that Bookspan is heading, Brita says. As a first step, Bookspan recognizes members when they register to access the site. “We need to make sure that we give you the right featured information,” Brita says. “If we know you like fiction, you’ll get a fiction featured selection. The same with non-fiction. We need to render the correct title online just as we do in mailings.”
But Bookspan wants to take the personalization to the next level and be able to offer specialized titles depending on a member’s web site activity, while the activity is taking place. For instance, right now, a person who visits the sale and clearance site today will get an e-mail tomorrow promoting the books the customer prefers that are on sale. Bookspan would like to be able to do that while the customer is online. “We really need to marry our legacy data with the data we collect online so we can match up what we know about them from the mainframe with what we know about them from what they do on the web site,” Brita says.
To achieve that integration, Bookspan recently began working with Art Technology Group Inc. to implement middleware between the legacy system and the web system.
That will allow Bookspan to offer more targeted promotions that would zero in on what the customer is likely to respond to and could improve the company’s profitability. “If I know that you’re on the sales and specials page, then I could offer you promotional books at discount. But if I know that you regularly visit the author interviews and excerpts pages, then I could alert you to new excerpts and not have to offer you the discount,” Brita says.
But registration is not the only way to provide personalization, some vendors say. SPSS Inc., for instance, is providing an analytics component to retailers that tracks what customers look at online, what they buy online and then, working with vendors such as ATG, combines the data with customer data in a mainframe. It also allows retailers to predict which customers will respond to what kinds of marketing. “This gives us a propensity prediction,” says Jay Henderson, product manager of SPSS NetGenesis product. “We can look at a customer’s behavior to determine their propensity to respond to a particular campaign.”
So far, several retailers have contracted to install the NetGenesis product. SPSS says SoftMap, a Japanese technology retailer, experienced a 300% increase in profitability after installing NetGenesis vs. the prior-year period. SPSS says others are still in the early stages of bringing it to their sites.