Helpful or creepy? Overpersonalized Web sites may spook shoppers

Michael Redding describes the get-to-know-you game between man and machine as a version of "Name That Tune."

Take, for example, someone who lands on an online retailer's home page and enters a search term. The Web site can identify the person's rough location and a bit about what the shopper wants to buy. With those two notes, the site can start figuring out who the visitor is.

"Even with just a few pieces of data, you can do a lot," said Redding, who is global director of development at Accenture Technology Labs, the consulting giant's research and development arm.

What if a Web site gleaned a hint of someone's identity before that person arrived at the home page and could greet the visitor with personalized content? Accenture has technology that aims to do just that, and its efforts are part of a larger wave of emerging Web technology that seeks to be smarter about predicting consumer preferences.

People may be sharing more personal information than ever via online social networks, but there remain many subtleties to consumer behavior that technologists are working to demystify through a combination of computer science and anthropology. In some ways, the viral and ephemeral nature of social networks has made following the spread of online information harder than before.

"People assume everything is trackable on the Internet," said Samir Balwani, a strategist at interactive marketing agency Morpheus Media. "But the free form of social marketing -- how things are passed along and not being sure where visitors are coming from -- makes it difficult."

One aspect of Accenture's technology platform is an "optimization engine" that examines the search term a person uses to arrive at a Web site. In a recent demonstration, Redding showed how someone who types "malts" into a search engine may land on a Scottish whisky-maker's site.

The site assigns a sophistication level to the user based on the search term and age, which visitors are required to input before entering. A broad search such as "malts" would peg that person as a novice, and the site would respond by presenting entry-level content such as an invitation to attend a local tasting event.

If another user lands on the same whiskey site by searching "Speyside," a region of Scotland, the site tags the visitor as an expert and displays information about food pairings or pricier brands.

"There's an art and science to personalization," Redding said.

Procter & Gamble signed on as Accenture's anchor client in September, meaning the consumer products giant will be relying on Accenture for its Web infrastructure and may be experimenting with some of these new features.

Science may power the technology, but it takes a human sense of subtlety to know when an online ad or recommendation crosses the line from helpful to creepy.

This rediscovery of the human element springs both from privacy concerns and an acknowledgment of the computer's shortcomings in fully understanding consumer behavior. A person searching for "jaguar" could be referring to an animal or a car. An online clothing store that relies on a customer's shopping history to generate recommendations might be flummoxed if that person undergoes a lifestyle change, such as a new career, that requires a different wardrobe than what previously was purchased.

"In some cases, people are recognizing limitations to the technology, and that's spawned an interest in, instead of algorithmic search, social search," which incorporates suggestions from a community of users, said Jeff Grau, a senior analyst at research firm eMarketer.

Wine.com faced the daunting task of designing a site for a customer base with a wide variation in wine tastes and experience. The site already has an online community for users to share reviews.

In August, it introduced new recommendation tools and a way to sort visitors into three categories: casual, connoisseur and gifter. Shoppers also can search for a wine they like and find similar-tasting products, even if they lack the vocabulary to describe their wine preferences.

Amy Kennedy, vice president of marketing for Wine.com, said the site tries to replicate the personalized service that customers at brick-and-mortar wine shops or restaurants receive from knowledgeable staff.

"What we really wanted to do is serve up recommendations based on your purchase history, your browsing past ... and mesh that with the wisdom of crowds, people who shop like you," Kennedy said.

Wine.com uses technology from RichRelevance, a company that builds recommendation tools for online retailers.

RichRelevance CEO David Selinger, who formerly ran Amazon.com's research and development team on personalization, said one of the industry's biggest breakthroughs was giving online shoppers multiple options. A computer doesn't know, for instance, whether a person who puts a shirt in a shopping cart is looking for more shirts, matching pants or something else.

That's why, instead of showing the customer more shirts, online retailers offer recommendations based on what others who also looked at that shirt have browsed or purchased.

"The (early) promise of computers and machine learning was that we would find the one product you're going to buy next," Selinger said. "The advancement we're making now is peeling that assumption back. There isn't one thing I might do. There are four or five distinct shopping paths."

Multiple choices and transparency about how recommendations are created also help sites avoid "the creepy factor," as Selinger puts it. One challenge in building this technology, however, is that comfort levels differ widely among consumers.

Balwani described it as partly a "semantic thing. No one wants to be targeted; everyone wants to be recommended something."

Growth in social networking and location-based mobile applications, which can send discounts for local businesses to a cell phone as a person strolls by, are bound to stir even more debate about privacy and personal boundaries.

"It's a Big Brother feeling when I search something, and all of a sudden it's my next ad," Balwani said. "If the ads are what I'm looking for, then, technically, I should be happy about that. But it's the whole idea of somebody watching what I'm doing."

Redding said that if done correctly, new advances such as linking a Facebook account to an online retailer could make personalized shopping a positive experience. Many sites are experimenting with this technology, which opens new possibilities of friend-based recommendations and reviews.

"Sites are just starting to tap the potential of this connection," Redding said. "It's a deep experience where every site you go to will feel like home."
___

(c) 2009, Chicago Tribune.
Visit the Chicago Tribune on the Internet at www.chicagotribune.com/
Distributed by McClatchy-Tribune Information Services.