Picking Winners & Losers

IT can help you find the hidden pearls in your customer database, but there are risks.

The customer wasn't happy. He had tried several times to get in touch with someone at the local office of Cypress Semiconductor Corp., but no one had returned his phone calls. So he filled out a customer satisfaction questionnaire at Cypress's Web site, and his poor rating of the company automatically triggered an e-mail to Bill Verdi, Cypress vice president of headquarters sales. Within 24 hours, Verdi was on the phone with the customer and the local sales office.

"The customer was totally amazed, and pleased," Verdi says. "We booked business with him within three days."

San Jose-based Cypress has made satisfaction surveys the cornerstone of its customer segmentation efforts and its entire customer relationship management (CRM) program. In the chip maker's eyes, nothing so distinguishes one customer from another as each customer's feelings toward the company. Cypress's real-time customer satisfaction monitoring system from Satmetrix Systems Inc. in Mountain View, Calif., triggers the e-mails to Verdi and will soon produce the data that determines the bonuses awarded to employees, based on responsiveness to customers.

Companies often segment their customers for different treatment. Frequent fliers get the best seats (but not necessarily the best prices). Buyers of gardening books get pitches for more gardening books. Big-bucks investors never get put on hold.

Data warehouses, data marts, data mining tools, statistical and analytical software, and CRM systems are enabling ever more sophisticated customer segmentation. But pitfalls abound, including alienating customers by making inappropriate pitches and ignoring customers with low current returns but high potential.

Cypress has the right idea about surveys, says Fred Reichheld, director emeritus at management consultancy Bain & Co. in Boston. He says companies often spend millions of dollars on surveys but then don't use the results at an individual customer level. "So I could have said in a survey that I am unbelievably dissatisfied, and the next time I talk to a service rep, they have no idea I ever said that," he says.

Another mistake companies make is segmenting current or prospective customers on the basis of demographics such as age, income, sex or education because that information is relatively easy to get. "But the best companies will segment based on fundamental values," Reichheld says.

For example, he says, The New York Times Co. tried without much success to attract new readers based on demographics. Then the publisher discovered that readers of The New York Times shared certain values, such as an interest in lifelong learning. Those kinds of values can be gleaned from surveys and from mailing lists obtained from sources such as The History Channel, Reichheld notes. He now rates the Times "outstanding" in its subscriber campaigns.

FleetBoston Financial Corp. has increased the number of people in its database marketing area from three to more than 30 in five years. The company says the targeted marketing that its customer segmentation allows has boosted the returns from its sales campaigns tenfold.

Using time-and-motion studies and activity-based costing, Fleet has computed the cost of every kind of transaction and customer interaction. Fleet can use that information to compute and track the profitability of every customer, and it can target its marketing efforts to individual customers and households based on their current and past contributions to the bottom line.

To help predict future contributions, Fleet buys data from external sources such as credit bureaus. "We figure out what's the customer's total wallet, then we can see what's our share of the wallet," says Brian Wolf, senior vice president for corporate marketing. The possibility of getting even a small share of a big wallet makes such customers a juicy marketing target.

Fleet also uses a neural network system to watch transactions in real time. It can spot patterns, such as decreasing transaction rates or balances for a high-value customer, that indicate that a customer may soon leave. "We call them," Wolf says. "We found we could cut our attrition rate in half. In most cases, we had products they didn't know we offered."

'If You're Rich, Press 3'

At many financial institutions, including Fleet, caller identification and routing systems linked to a database of customer histories and characteristics are used to ensure that the most valuable customers get preferential treatment. Investors with million-dollar portfolios get to bypass those endless automated voice prompts and are routed to the most experienced service representatives.

The rigor of Fleet's customer analysis isn't for everyone, says David Harding, a principal at consultancy McKinsey & Co. in Minneapolis. "Some companies spend years and years and millions of dollars building these databases, but when it comes to making a calculation around customer value, they can't pull it off," he says.

Often, the problem lies in trying to determine the cost of serving each customer, which is difficult and unnecessary, Harding says. It's usually sufficient to just apply average costs to customers by group.

Some companies fail to send the information mined by analysts to the marketing, sales and front-line customer service people who could actually use it. He recommends having integrated teams of IT, statistical analysis and marketing people.

Customer segmentation sometimes drives dynamic pricing, a model in which prices for airline seats, hotel rooms, rental cars and other items change as supply and demand change, and vary depending on who is buying. Such revenue management practices can add millions to profits, but there are risks.

In September 2000, Seattle-based Amazon.com Inc. found itself in hot water when some customers found that they were being quoted higher prices than other customers for DVDs. Amazon quickly abandoned the practice, calling it a pricing "experiment."

Amazon botched the implementation of a perfectly valid concept - that of offering better deals to those customers deemed to have greater long-term value, says Deepak Sirdeshmukh, a marketing professor at Case Western Reserve University in Cleveland.

"What companies do in a smarter way is stealth differentiation, in which the actual price customers see is the same for all segments, but then in the background you mail or e-mail discount coupons of different value to different customers," he says.

Sirdeshmukh says IT-based initiatives often backfire because it's so easy to spit out promotional e-mails or make dinnertime telemarketing calls to some favored customer segment. "One of the biggest pitfalls of customer databases is that the best customers are bothered endlessly - surveys, new offers, cross-selling - sometimes by multiple people within the company," he says. "People are getting CRMed."

The solution, Sirdeshmukh says: "You need smart thinking on top of the database."

The online division of Zeeland, Mich.-based furniture retailer Herman Miller Inc. is determined to avoid that kind of mistake. The 16-month-old division, called Herman Miller RED, is collecting data on the company's customers and will one day, for example, notify frequent buyers of chairs when the company gets a new model in stock.

But the process isn't ready yet. "Our technology is ahead of our business, and there's a great danger in doing something because you can," says Matt Johnson, the company's Web site manager. "The personalization won't pay off until we have a critical mass of user experiences and order histories."

Special Report

kc_teasers_crm.gif

Sober CRM

Stories in this report:

Copyright © 2002 IDG Communications, Inc.

It’s time to break the ChatGPT habit
Shop Tech Products at Amazon