Don’t Get Buried in Customer Data—Use It

With the advent of customer relationship management (CRM) in the late 1990s, companies came to believe that by using technology to tailor their offerings to individual consumers’ needs, customer loyalty—and company profits—would skyrocket.

But in today’s crowded marketplace, customer loyalty is more elusive than ever. A recent McKinsey study reveals that the annual churn in the wireless industry increased from 17 percent in 1995 to 32 percent in 2000. This trend holds true even in industries less susceptible to turnover. In core retail categories such as department stores, for instance, the top players’ market share declined more than 10 percent.

Not surprisingly, many executives’ faith in CRM has waned. In a 2001 Bain & Co. survey of the 25 most popular management tools, CRM was ranked near the bottom. In a follow-up study, 20 percent of the 451 senior executives polled said that their companies’ CRM initiatives had failed to deliver profitable growth and had damaged long-term customer relationships.

Tempting as it may be to point the finger at your CRM technology, that won’t help you reverse these worrisome trends. It’s quite possible that the problem isn’t with your CRM technology at all but with the way you are collecting and using your data, experts say. Although getting your CRM program in order is an essential component of achieving customer loyalty, there’s much more that you need to do.

“Marketers need a good, thoughtful architecture to base their decisions on,” says Harvard Business School marketing professor Gerald Zaltman. A more strategic approach to data mining can provide the foundation for that decision-making architecture. Below, advice on how to use information about the individual customer and the average customer in concert, and how to probe beneath customer preferences and behaviors to uncover the attitudes that provide a more solid understanding of customer loyalty.

Why you need both individual and aggregated data

One-to-one marketing, a term coined by Don Peppers and Martha Rogers in their influential 1993 book, The One to One Future (Currency/Doubleday), focuses on share of customer: Using the insights about what makes your most loyal customers different to maximize the value of those relationships. By the end of the decade, many marketers had come to believe that the combination of mass customization techniques, sophisticated database software, and the Internet would enable them to actually deliver on the promise of customized offerings to each individual customer.

But that hasn’t happened to the extent it should have, says Cleveland-based consultant James H. Gilmore, coauthor with B. Joseph Pine II of The Experience Economy (Harvard Business School Press, 1999), because “most practitioners have taken the concept of one-to-one marketing and bastardized it into CRM. They’re using CRM tools to design better processes for a nonexistent ‘average’ customer, instead of customizing for individual customers.”

He cites the example of a major hotel chain that asks guests to complete a multiple-question satisfaction survey via their room’s TV set during their stay. When one guest answered “extremely dissatisfied” to all the questions, he was not treated any differently when he checked out. Why? Because his answers went straight to a central repository where they were aggregated with other customers’ responses and used to measure overall market—not customer—satisfaction. A more effective approach would be to feed his answers directly to someone at the front desk who could respond immediately to his needs and create a better experience for him.

“A company’s goal should be to learn more about what each customer needs so that it can close the customer sacrifice gap, which is the difference between what individual customers settle for and what each wants exactly,” says Gilmore. Steve Cunningham, director of customer listening at Cisco, agrees that it’s vital to listen and respond to individual customer needs and preferences. But he believes you must also pay attention to the aggregate data—customer averages based on individual surveys.

“Let’s say that based on the customer survey averages, you realize that your hotel is taking too long to check guests out,” he says. “So you launch initiatives designed to reduce checkout time and prime your personnel to be sensitive to that issue. Despite these efforts, something goes wrong, and one morning the front desk manager sees a long line of guests queued up to check out. Because the survey averages have helped sensitize him to the importance of this issue, he knows he has to do something—for example, pull staff members off other jobs so they can help check people out, or offer free coffee to everyone who’s standing in line.”

Familiarity with the aggregated survey data, in other words, helps the manager tailor his response to individual customers.

Cisco relies on three layers of customer data to inform its efforts to improve customer satisfaction: The overall satisfaction survey that customers fill out annually; interviews with targeted customer segments, follow-on surveys, and sessions with corporate advisory boards that seek to identify an initiative that will address a problem hinted at in the overall relationship survey (“this is the ‘digging and understanding’ layer,” says Cunningham); and, at the most granular level, records of each individual transaction that the company’s technical support group has with a customer.

To illustrate how Cisco uses these three layers, Cunningham cites a hypothetical example. Assume that for a given year, the average score for product reliability has slipped a bit. Drilling down to the bottom two layers of data, Cisco discovers a problem with the power supply for its routers. It launches an initiative to solve this problem and identifies the number of spare power supply parts it sends out weekly as the measure it will use to track the progress. The transactional measure—the number of spare parts shipped weekly—may start to come down fairly soon after the initiative has been launched, but it may take a while before the change shows up on the annual relationship survey.

“You need both the aggregate and the transactional information,” says Cunningham. “The survey data tells you about the overall health of your relationships with customers; it tells you which way the wind is blowing. It also helps prevent you from running after individual problems that may not be significant in the aggregate. The transactional data gives the detail behind the relationship.” It helps you pinpoint specific issues that need to be addressed to boost overall customer satisfaction.

Digging deeper

To boost customer satisfaction and, ultimately, customer loyalty, you have to do more than listen simultaneously to customer averages and to individual customers. You also have to look for what lies beneath the externals of customers’ behavior (what they buy, how they buy, and when they buy). “Without capturing what is going on inside customers’ minds and hearts, and integrating that information with the factual external experiences, the picture is incomplete,” says Doug Grisaffe, chief research methodologist for Indianapolis-based Walker Information.

“CRM tools enable you to collect a lot of rich data about a customer’s frequency and time of purchase, the size of her orders, and what she thinks of your company,” says Harvard’s Zaltman. That’s necessary but not sufficient data: It doesn’t tell you anything about “why customers do what they do, think what they think, and why they like or don’t like your products. Getting that level of insight requires more intensive interactions with customers than CRM tools permit.” It requires that you develop a “poetic insight into customers—a deep knowledge that enables you to intuit their answers to questions you haven’t even asked them.”

In one-on-one interviews with customers, Zaltman uses a process he describes as metaphor elicitation to get at the beliefs, emotions, intentions, and often unconscious attitudes that people have about a product or brand. As he explains in his recent book, How Customers Think: Essential Insights into the Mind of the Market (Harvard Business School Press, 2003), the information gleaned from these interviews as well as from surveys and observation is used to create a consensus map—an illustration of the particular bundles of constructs that customers have developed based on their experience and emotional connection with a product or brand.

A consensus map that Zaltman developed for General Motors reveals the richness of the metaphor elicitation approach. As expected, customers associated GM products with quality and competitive price. But there was more: Customers also linked GM with patriotic feelings. By buying GM cars, they saw themselves as not simply helping Americans keep their jobs, but as fulfilling a larger obligation that they felt toward their country.

Once you understand these often surprising bundles of associations, you can reinforce and sometimes alter them with the messages your company sends to consumers.

Based on the consensus map Zaltman produced, GM’s domestic managers redesigned the customer experience at dealerships and added subtle cues in their advertising to make the idea of patriotism more salient. For GM’s overseas managers, the task was more difficult but no less valuable for that. Realizing that GM products also produced patriotic associations among foreign purchasers, the overseas managers “found cues that underscored patriotic associations with the local country without pressing the American button,” says Zaltman.

Reams of customer data are no guarantee that you’ll be able to increase your most profitable customers’ loyalty—you have to be sure that you’re collecting the most relevant information. Listening for the attitudes that inform customers’ behaviors and preferences, Zaltman maintains, gives you “a more solid basis on which to craft and implement strategies that will improve customer loyalty.”

Find it here: http://hbswk.hbs.edu/item/3596.html

Where is Consumer Generated Marketing Taking Us?

Consumer generated marketing is a fact of life to which all of us will have to adapt. Adaptation means learning how to use CGM to provide one form of input in fashioning product and marketing decisions. Those are the messages from respondents to this month’s column, who seemed to treat what some might think is the most revolutionary concept in marketing to come along in a long time in a very matter-of-fact way.

Bob Nemens commented, “Traditional marketers may be quick to dismiss Internet chatter as coming from the ‘uninformed.’ … If Thomas Edison had been given the option at the time, I bet he would have spent significant time on the online underground.” Fernando Polo dismissed the potential pitfalls of bias caused by listening only to outspoken users of the Internet by saying, “Excuses such as ‘listening to the wrong complaints’ are just that: excuses. Text-mining technologies can now help companies listen to their customers better than ever… My advice: ‘Don’t let your competitors listen to your clients; do it yourself first.’” Andrea Learned pointed out, “Eventually, those slightly later adopters and less-active types will join in to make blogs a more representative discussion vehicle… The same will likely happen in the consumer generated marketing realm when slightly later-to-adopt consumers realize how much they can influence manufacturers….”

The proper role of CGM was a source of some comment. In particular, the findings of a study cited in the column that associated companies utilizing mechanisms for paying attention to “emerging customers” with the fostering of “disruptive technologies” raised some eyebrows. As Christophe Meili put it, “I’m surprised that consumer generated marketing would foster disruptive innovation. I was under the impression that … true disruption would be generated less by popular demand than by hard radical thinking.” Flavius Chircu suggested that if we regard consumer generated marketing as “something akin to the other party in a dialogue … (it) becomes a source of incremental improvements whereas anything revolutionary could only come during the ‘breaks’ in the dialogue.” And Caleb DeGrenier commented that “companies still need to surprise the market with innovative products that no single customer would have thought of.”

This brings us full cycle to some of our original questions about consumer generated marketing. Is it really something “new under the sun”? Is it, for marketers, a disruptive technology in its own right, something offering decision makers more for less (or, more accurately reflecting the definition of a disruptive technology, less for a lot less)? Or is it something to be regarded as providing just one of several important inputs to future product development and marketing decisions involving primarily non-disruptive technologies? Does chatter matter? And how much? What do you think?

Find it here: http://hbswk.hbs.edu/item/4781.html

A Balanced Scorecard Approach

Happy customers are good, but profitable customers are much better. In this article, professor and Balanced Scorecard guru Robert S. Kaplan introduces BSC Customer Profitability Metrics. From Balanced Scorecard Report.

Find it here: http://hbswk.hbs.edu/item/4938.html

The Box Office Power of Stars

Just how much do movie stars contribute to box office success? HBS professor Anita Elberse researched the notion of “star power” to better understand how A-list players contribute to Hollywood’s bottom line

Find it here: http://hbswk.hbs.edu/item/5025.html

A Survey-Based Procedure for Measuring Uncertainty or Heterogeneous Preferences in Markets

People who buy retail prescription drugs, invest funds, or participate in auctions rarely have complete information about the product they are buying. Often the only auction information participants have is the number of bidders, observed bids, and product characteristics. If data from an auction, for instance, is a function of bidder behavior, then external survey data may help in testing hypotheses about bidding behavior. Researchers often avoid using surveys because they consume time and effort, but Yin presents a survey design technique and econometric tool to deal with a general population of survey respondents. Her application tested eBay online auctions selling personal computers. Key concepts include:

• Survey data may be a good complement for market data, especially for auctions, as a measure of uncertainty or different preferences.
• Survey data may be more valuable than other methods of evaluation because it exploits the human ability to assess complex sets of information.
• A survey may be implemented more quickly with a larger number of respondents, even if they are inexperienced, than with a smaller number of experienced respondents, by correcting for survey bias.

Find it here: http://hbswk.hbs.edu/item/5410.html

Will the “Long Tail” Work for Hollywood?

The “long-tail phenomenon” is well documented: Amazon.com makes significant profits selling many low-volume books. But can the long tail work for video sales as well? A new working paper by professors Anita Elberse and Felix Oberholzer-Gee suggests that it may not bring the same benefits to Hollywood. Key concepts include:

• For video sales, the long-tail phenomenon is not as pronounced at it is for books. There is evidence of a shift in sales to the tail for video, but an increasing number of titles do not sell at all.
• Hollywood strategists have no easy answers for pumping up revenue, given a decline in the number of blockbuster hits. This new research suggests that the long-tail phenomenon might not be a panacea for video sales.
• The music industry may be more of a long-tail beneficiary than the movie industry.

Find it here: http://hbswk.hbs.edu/item/5520.html

Customer lifetime value

Donald Lehmann discusses Managing Customers as Investments, a new book in which he and coauthor Sunil Gupta explain how to calculate and apply customer lifetime value.

You write that teaching the core marketing course to Columbia MBAs was one of your main inspirations for the book.

During the dot-com bubble, I noticed there were a lot of companies with a lot of valuation that I knew was just not sustainable. Basically, the students were less than enthusiastic when greeted with that news, so I decided to prove to them why. The basic proof was this: a business runs on revenue, revenue comes from customers, so if I can project customer revenue, I can project the value of a business. And in point of fact, several of the dot-com companies were grotesquely overvalued, based on any reasonable projection of revenues that would come from customers.

We had a model that basically married together one of the fundamental approaches in marketing, which is to predict life cycles of product diffusion curves — which we use to predict the number of customers — and then straight old extrapolation to project the margin per customer. We put the two together. So our theory, if you want to call it a theory, was that customer growth would not continue forever, that at some point it would level off.

How does your approach differ from previous methods of valuing customers?

Most people have not valued customers in the past. Those that have are people in direct marketing and direct mail, in particular, who have a model called RFM: recency, frequency and monetary value. So, you measure how recently a customer has been with you, how frequently they buy stuff from you and how much money they spend on it, and you use that to get some sense of how valuable a customer is.

We took the discounted cash flow paradigm and used that to estimate the value of an individual customer. So we simply project into the future the revenues and costs at a customer level and then figure out what that customer is worth. To me, it was very logical and quite obvious, and while there are a number of other customer books out there that also talk about that notion, nobody had pushed it as far as we did. In particular, we show that if you sum the projected values of the customers, that number would give you some indication of what the firm ought to be worth. That was the unique link.

The closest to this has been what the direct marketing people have done, because they have terrific data sets on individual customers. We were looking at aggregate data and trying to get a sense of what an average customer is worth. We used basically public information. To say, “We have a great method to measure the value of a company if you can get access to internal company data” is not overly useful.

You used many examples and case studies in the book. Are any especially representative or illustrative in supporting the methods you advocate?

What got us interested in the telecom cases was that the price per customer was jumping dramatically in a series of mergers and acquisitions, which suggests the market wasn’t efficient. So we wanted to project how many customers would a Verizon or a similar company end up with and how much could they expect to make per month, and therefore per year, off them. Project it out, discount it back in present-value terms, add it up across all the customers and see what that company would be worth; or, at the individual customer level, look at what we projected a customer would be worth and compare it to the implicit price from the merger and acquisition. They overpaid, in my opinion.

We also had, for some of the dot-coms, an interesting difference in that some of them were pretty close to our estimates and some of them were grossly overvalued according to our method. One of them is eBay. And you can argue eBay’s got a different business model, because they’ve got both buyers and sellers going through there.

Can you talk more about your method and how firms can use it?

You can use it at the individual customer level to figure out what an individual customer or segment of customers is worth on a per head basis and then relate that to how much effort you spend to acquire and retain them. If a customer is going to generate discounted cash flow profits of $400 over their lifetime, you don’t want to spend $500 acquiring them. And yet, I think an awful lot of companies, because they have not done that relatively simple calculation, have overpaid for customers and either over- or underpaid on retention. If a customer’s worth $20,000 to you and you’re spending $10 a year to retain them, you’re probably not spending as much as you should.

Does your method offer a formula to determine the right amount to spend in acquiring and retaining a customer?

The method gives an upper bound. You would not spend more for a customer than they are worth. If a customer is worth $400 to you and you spend $399 on them, it’s a good deal — you made a dollar. Is a dollar on $399 a good rate of return? That depends on the firm’s cost of capital, risk preference, etc.

Is your method equally relevant to all industries?

Yes. A friend of mine at Rite Aid got data on their stores. I think this method perfectly allows you to value a retail chain. All you need to know is how many stores they’re going to have and how much they’re going to make per store, and that basically is what the value of the chain is. In essence, what people in finance do is project out revenues and discount them back at the aggregate level of the chain. All we’ve done is bring it down to the individual store level and then aggregate it back up.

So for a going business that’s been around for a while, where it’s pretty obvious how many units there are going to be and how much they’re going to make per store, the methods are the same. But if you’re in a situation where none of the stores are making money and you discount that back, you get a negative number, implying that the chain is worthless. The advantage of our method is that we look forward, so we would get a number that makes more sense in that scenario.

How do companies account for investments in customers on their balance sheets? How would you account for this investment?

The accounting rules are such that you do not account for the value of customers or the value of brands, which together in lots of organizations make up well over 50 percent of the value of the firm. Most companies, consistent with the accounting regulations, expense all marketing. Sometimes they capitalize R&D, but they expense marketing. We would view it as an investment. If I spend $400 to get a customer or if I spend $400 to get a piece of milling equipment or a photocopy machine, they’re all investments. You get some payback over some period of time. There is some risk that the thing will break down, which is the same as the customer defecting. There’s an annual maintenance cost for a photocopy machine or a piece of milling equipment or to retain a customer.

Have you noticed any change in firms’ behaviors as a result of this book? What would you like to see change as a result?

Well, I think firms should be much more rigorous in evaluating strategy through some kind of metric. I think customer lifetime value is a very useful metric. It’s not the only one I would use, but it’s an important one for evaluating decisions, certainly merger and acquisition decisions. Let’s assume Harrah’s has decided they want to get more customers to their casinos. You know how much it costs for the program they plan, whether it’s a TV ad blitz during the Super Bowl or direct mail or whatever. How do you figure out if that makes any sense?

One thing you can do is say, “I know what new customers are worth to me. I have to get so many new customers to break even to justify this expenditure — is that reasonable or not?” Often, when you look at that question, the answer falls into one of two categories: “Of course it’s reasonable, that’s going to be easy,” which means you spend the money in a hurry; or, “There’s no way we can generate that many customers from this campaign,” in which case you can save that money.

Donald Lehmann is the George E. Warren Professor of Business and Sunil Gupta is the Meyer Feldberg Professor of Business at Columbia Business School.

Defaults make a difference

People sometimes sign up for things and then opt out later. Or they opt out now and opt in later. In both cases, they change their decision. Does it matter? In some cases, no, but in other cases it matters a great deal.

Take organ donation: when you renew your driver’s license, you can sign a card that makes you a potential donor. Can hospitals that handle organ donations rely on your pledge, and if so, to what degree? And what about people who don’t sign up — do they actually opt in at a later date? Knowing who will come through and who will not — and how best to sign them up — could help hospitals plan how many needy patients to put in their pipeline.

In organ donation, these are matters of life and death. In other domains — like flight insurance, retirement savings or Internet privacy — the stakes are lower but still high enough to make it worthwhile to look for an answer.

Professor Eric Johnson worked on this problem with Daniel Goldstein, using the Virtual Laboratory of the Center for Excellence in E-Business at Columbia Business School. They ran experiments over the Internet that gave hundreds of people essentially the same choice but in two different forms: (A) agree to be a donor, with an opt-out clause, and (B) decline to be a donor, with an opt-in clause. The researchers expected some difference, but the size of the spread shocked them. A’s outnumbered B’s by a factor of 2 to 1; people were twice as likely to agree to be a donor when they had to opt out as when they had to opt in.

Further research on different countries showed similar results. Rates of donor agreement and actual donation were low in Denmark, Germany, Netherlands and the United Kingdom, where potential donors had to opt in. These rates were much higher in Austria, Belgium, France, Hungary, Poland, Portugal and Sweden, where potential donors had to opt out.

In many situations, which default you build in makes a big difference. If you want people to do something in the future, ask them to agree to it now, with an opt-out clause. The default option — where they take no further action — is then in your favor. Johnson and Goldstein noted this result even in simple online agreements, where the default is a box already checked and the viewer can opt out by unchecking the box. That approach yields more positive results than asking the viewer to check the box.

The principle behind the results of this research might have even wider implications. When people see an option for the first time, they don’t yet have a preference one way or the other. Instead, they construct both the problem and the solution right there and then. So how you present the question — opt in or opt out — makes them see the whole matter in two very different lights. This is true for minor items like online offers and for major decisions like how much to save for retirement — and even, as we see with organ donation, for questions of life and death.

Eric Johnson is the Norman Eig Professor of Business and director of the Center for Excellence in E-Business (CEBiz) at Columbia Business School.

Emotional accounting

In purely rational economic terms, money is fungible. It shouldn’t matter where the $20 in your wallet came from, whether you earned it at a job or found it on the street. But people act as if it does, and in the 1980s the concept of mental accounting emerged. According to this concept, people categorize money they receive by its source, and deposit it in different mental accounts. Money received from a windfall such as winning the lottery would go into one account, for example, whereas money received as income from a job would go in another.

Mental accounting explains why investment bankers tend to spend their bonuses on plasma TVs and exotic vacations but save money from their salaries to buy homes. Over the past few years, Professor Jonathan Levav, working with A. Peter McGraw of the Leeds School of Business at the University of Colorado at Boulder, pursued this idea further and found that in addition to mental accounting, people tend to categorize money according to the feelings they associate with it, a process he calls emotional accounting.

To illustrate how we engage in emotional accounting, Levav compares two scenarios: receiving $200 as an unexpected gift from an aunt and receiving $200 as an inheritance from an aunt who just died. Although the money is from the same source, in the first case it triggers positive feelings, while in the second it produces what Levav terms a “negative affective tag.”

The way people evaluate money becomes more complicated when mixed emotions are involved. Suppose an investment banker received a bonus but knows that a colleague who didn’t work as hard got more money. In this case, the banker is glad he got a bonus but also angry because he feels he was treated unfairly. Meanwhile, a banker who received a bigger bonus than one of his close friends may feel more guilt than happiness.

To assuage such negative emotions, Levav says, people employ various cleansing and avoidance strategies. “If I have negative feelings about money, then I’ll launder it of its negativity,” he says. “This can mean spending it in ways that are virtuous or utilitarian. So I don’t buy the plasma TV for myself, but I might buy one for an orphanage. It’s not just about the product; it’s really about the use.”

Levav and his coresearcher conducted several studies that tested how university students spent money they received under different circumstances. In one experiment, students who completed a market research survey were given a choice afterward of different $2 coupons. They could spend the coupons either on ice cream in the cafeteria or on books in the university bookstore. Half of the students were told that the grant for the coupons came from the computer firm Dell, which had positive or neutral associations within the student population. The other half were told the grant came from the tobacco company Philip Morris. About 44 percent of the students who were told the coupons were paid for by Philip Morris chose the utilitarian textbook coupons, double the percentage of those who were told the coupons came from Dell.

“In essence, people tell themselves, ‘Let me do something good with the money so I don’t feel bad about it anymore,” Levav says. When people are angry — such as the banker who feels he was cheated on his bonus — they tend to set the money aside and give their anger a chance to dissipate. People who feel guilty are more likely to donate the money to charity. When teachers at an affluent Chapel Hill, N.C., high school received bonuses based on their students’ standardized test scores, they donated the money to a rural school, and said their students’ performance was partly the result of the community’s wealth. Levav and his coresearcher hypothesize that the teachers’ donation was a way to cleanse themselves of the negative emotions they associated with the money.

“When we make the decision to spend money virtuously — paying off a chunk of our college tuition, rather than paying off debt racked up on a weekend in Vegas — we can erase any bad feelings associated with it,” says Levav. “It may not be rational, but it makes us feel a lot better.”

Jonathan Levav is assistant professor of marketing at Columbia Business School.