Personalization reality does not meet promise

Web personalization has been in trouble almost since its beginnings. Except for a handful of well-known retailers like, few companies have used the technology to truly interact with customers on an individual level.

"There was a bunch of euphoria a year or two ago, largely fueled by marketing hoopla," said Guy Creese, a research director at Aberdeen Group, Boston. "People are finding out that it's harder than they thought, and personalization has a bit of a bad name."

The promise was huge. With this technology, the rap went, a company could customize content, sales offers and loyalty programs each time a customer came to the site. Customers could be put into "buckets" based on past behavior as well as predictions of future behavior. Each of these "buckets" could be treated differently, based on needs and behavior. Delighted by all this individual attention and intrigued by offers that matched their real interests, customers would spend more money with the company, or just be a more loyal and happy camper. The company would learn even more about its customers, allowing it to further personalize its marketing. The result? Soaring revenue.

It was a powerful scenario. But the reality did not quite unfold as described. "To be very candid, even our best commerce customers" are using only the most basic type of this technology, said Bruce Gras, director of advanced personalization at BroadVision Inc., a software vendor in Redwood City, Calif.

Gras attributes customers' cold feet to the "very high" labor component needed to create and maintain a world-class system. It takes time and good IT people to put together a "personalization regimen," as Gras refers to it. And then, after all that effort, the system may well have to be revised quickly as market conditions change.

In general, most firms are "still trying to get a handle on who their customers are," said Paul Hagen, senior analyst at Forrester Research, Cambridge, Mass. "Unless you know your customers, you can't really customize the experience for them."

Most Web sites today are still "one size fits all," Hagen said, adding "there's very little concept that I've got different types of customers" with different needs.

Even retailers—which, along with financial services firms, have been the traditional big users of customization technology—are not all necessarily jumping on the bandwagon. According to a recent Forrester survey, said Hagen, more than half the retailers polled said Customer Resource Management (CRM) and personalization were "the same thing." About one-quarter of the respondents stated they would implement CRM before they would do anything with personalization.

Each business unit tends to have its data in its own systems, stored in its own way. "That's not just technically difficult," Hagen said, "it's politically and organizationally difficult, too."

Personalization has been a jerky journey for many IT shops for several reasons.

The top-of-the-line software itself is big and complex, almost on par with a CRM system with its many pieces and parts. Customization, done correctly, also needs to draw from many different data sources. But in many companies, customer data is scattered about in various systems like so many jigsaw puzzle pieces left in different rooms of a house.

Then there is the problem of personalization becoming a catchall term for many different definitions. When some vendors use the word personalization, they are referring to a data mining approach. In this case, a company decides what its customers have been doing, how much they are spending and what they are spending it on. The company knows which services customers have paid for, individually or in "buckets" of many customers. The firm can then match Web promotions or e-mail campaigns to particular customers based on past behavior.

Other vendors, however, talk about personalization in terms of rules-based engines that predict future behavior based on what the customer (or those like him) has done in the past.

So one camp talks about what customers have already done, and another talks about what they will do. Both aspects are important for a truly robust system.

Yet another issue comes into play with some of the rules-based systems on the market, the Aberdeen Group's Creese said. "If you have 20,000 customers, you don't want 20,000 rules, you want maybe 100. But does one rule negate the other? Will the real rule please stand up?"

Beginning again
Despite its rocky start, there are some bright spots on the personalization horizon. Vendors are taking various steps to make the software easier to deal with from different angles. Customers, too, are learning more ways of taking fuller advantage of the technology. (See "Premier's path to customer care.")

Premier's path to customer care
Premier Inc., San Diego, is not a typical user of Web personalization software. It is not a giant retailer, nor is it a financial institution. The company is an alliance of 2,000 hospitals across the United States that acts as a group purchasing organization, negotiating contracts with suppliers of hospital gear, from beds to pharmaceuticals. It also provides maintenance and support for the equipment.

Another major business is IT. Premier develops apps that its member hospitals use to track how they are doing against other hospitals, follow their own spending and run reports. Premier hosts the apps on its own servers, and members come in via the Web.

This is where personalization comes in, according to Todd Wilkes, director of development and quality assurance for Premier. When a customer logs in, "we know which applications they've purchased and what reports they're allowed to run," he said. "We know which hospital and health-care system they're affiliated with, and which other hospitals they can benchmark against." Depending on all these factors, a user's screen options change accordingly, he added.

To do this, Premier implemented MicroStrategy's personalization tools, which have their own software development kit and an API to allow users to create a custom interface. This allowed Premier to connect the MicroStrategy software with its existing Red Brick data warehouse, which holds all the relevant data. No manual coding was required to get the two worlds to connect, Wilkes said.

When a user logs on, the MicroStrategy software paints the user's screen with different options based on rules that Premier staffers have set.

Along with that, Premier tracks the numbers and types of reports and selection criteria that customers choose on the site. "We look at who is doing what, and when, and then we make suggestions," Wilkes said. "We tell them things like they've paid for 10 user IDs, but [they] are only using five." Or Premier can suggest the customer try different types of analysis based on what others at similar institutions are doing.

An improvement still in the concept stage is to have the various reports waiting for customers when they arrive on the site. "They can tell us what they want us to have waiting for them," Wilkes explained. "Right now we're being reactive."

In the future, he said, customers will be able to ask for reports to be run automatically anytime their hospital—or anyone they compare themselves to—refreshes their data. When that happens, the site will be even more "sticky" than it already is. "They're already getting value, but this will make it easier and will keep them coming back," Wilkes explained.

Overall, Wilkes said, "we couldn't do what we're doing" without personalization technology. "We have to ensure that users of one hospital can't see anything but aggregated data about their competitors—and that the competitors can't see their data." Without personalization, "being able to benchmark would be impossible."

—Johanna Ambrosio

Creese suggests IT-ers take a deep breath and begin to apply personalization with a dose of common sense. Instead of segmenting customers into 50 buckets, go with a maximum of six or eight, he said. And for customers in the highest-value bucket—the ones with whom you really want a personal relationship—dazzle 'em with the technology.

"It's the old business trick of applying proportionate resources," he noted.

Also, Creese suggests starting as low end as possible, keeping in mind that whatever you spend on the software you'll likely have to spend on implementation services. One logical place to start is with your existing content management vendor. Creese believes personalization will increasingly become a part of content management offerings, much like workflow became part of content management.

"There's a realization that you can get a lot of benefit from personalization that's not incredibly sophisticated," he said. "How much personalization firepower do you need?"

IBM, for example, has a personalization package available as part of its WebSphere product line. WebSphere Personalization 4.0 adds a simpler GUI to allow business managers to change personalization parameters. It also tracks visitors as they click through the site.

"Personalization is becoming increasingly important to our customers," said Stefan van Overtveldt, program director for WebSphere technical marketing at IBM. "It helps them differentiate from competitors."

At Art Technology Group (ATG), a personalization vendor in Cambridge, Mass., the focus is on use cases. This is essentially customer segmentation, but for both internal and external use.

In this setup, customers would continue to be segmented as personalization products traditionally have done. But company employees in the telemarketing center or elsewhere would have their own use cases for how to help different customer segments, explained Bill Morrison, product marketing manager at ATG. This would result in a lower-level customer service representative having different options to present to customers than a higher-level rep, he explained.

The problem with customer profiles in general, Morrison added, is "you can't predict every situation. And that's where a lot of companies get stuck with personalization—they try to figure out what all customers want in every scenario. Just start someplace."

Decisions, decisions
Key questions to ask when figuring out which personalization path to take include:

  • What do we want to do with this? What are our key goals: selling more products, selling more expensive products, rewarding loyal customers?
  • How much personalization do we want or need, and can we get it from an existing content management vendor or seller of other Web tools? Do we need a high-end specialized package?
  • Who are the key stakeholders in decisions about personalization: marketing, sales, customer service or all of these?
  • Do we have an understanding of who our customers are, what they buy and what they want? (Be honest here—analysts say many companies think they have a handle on this when they don't.)
  • If we don't know much about our customers, is it worth building a data warehouse or some other means of finding out?
  • Do we have a means of tagging dynamic Web pages through XML or some other way?
  • If not, can we afford to implement dynamic page tagging? (Without this, there is no way to really customize content.)
  • If we are going to use a rules-based system to determine likely behavior of customers, or to advise them on potential purchases or services, who makes the rules? Who changes them?
  • How many "buckets" of customers do we have? How many do we need? (The fewer, the better. Personalization gets very complex, very quickly.)

—Johanna Ambrosio

One huge advantage of Web personalization, Morrison said, is that it can reduce customer service costs from elsewhere in the business. EMC saved $10 million per month by shifting customer service online from its more expensive call center. It retained the idea of personalized customer interactions, but is handling them electronically now.

The notion of a multichannel customer approach is gathering steam. Consider, for example, customers booking hotel rooms. If reservations made on the Web can only be cancelled on the Web, it can be a great inconvenience to customers who are on the road without Web access.

With multichannel access, customers can get service from anywhere—PDAs, kiosks and cell phones, as well as the usual Web, phone and e-mail. The idea is that the entire organization knows what is going on with the customer; for example, that she bought something and has called three times to ask about the procedure for returning it. This way, the customer does not need to repeat her story.

One of the easiest ways to implement personalization is to just ask, said H.A. Schade, director of product marketing at KANA Inc., a Menlo Park, Calif., CRM vendor that provides some personalization technology. "We believe very strongly in permission-based personalization," said Schade. "That builds trust because it gives customers a sense of control."

New wares on the horizon
KANA's big push right now is in the area of real-time analytics. As customers come to the site, or calls come into a call center, the analytics "look in real time at the analytics generated by the interactions themselves," explained Schade. "We can assess that I have a greater preponderance of e-mails or calls associated with a particular product, which may indicate that a problem exists and that you need to alert someone." Or even, for instance, tell a supplier that a product might be defective.

BroadVision is working closely with Teradata to connect the former's personalization tools with the latter's data warehouse. The objective, BroadVision's Gras explained, is to "automate" the connection of the two environments, and thus the personalization development and maintenance effort, as much as possible.

The step after that will be to bring the system "full-cycle," Gras added, for the type of real-time analytics KANA Inc.'s Schade referred to. After the customer has a personalized experience, the data will feed back into the customization system so future marketing campaigns and Web experiences can be further personalized.

These will be "critical" advances over the next year, predicts Gras. "While people had a lot of fear of getting deeply involved with personalization, automation will help them get over those fears."

At the end of the day, though, Forrester's Hagen said he does not "see any 'aha!' advances on the horizon. Companies are getting smarter about how they interact with customers. But this is slow and difficult work."