News & Announcements

MORE ON MARKET SEGMENTATION AS SEEN IN MARKETING RESEARCH

An End to Doorstops
Segmentation studies should do more than sit on your floor

By Kevin Clancy, Peter Krieg, and Henry Gamse, Fall 2007

Executive Briefing
These days, market segmentation exercises churn out more dust magnets than actionable results. After investing significant time and resources, marketers can't figure out how to use the information from a study, and the reports become little more than dense and immobile reminders of wasted energy. Segmentation can and should do more than group and sort buyers in a market; it should tell you where the money is.

We knew the times they were a changin' when we heard the Eduardo Castro-Wright, CEO of Wal-Mart—the very embodiment of mass marketing—raise a topic of conversation you never hear mention of, let alone in the same breath as "strategy for growth" or "revitalizing the brand": market segmentation. Wal-Mart "is all things to all people," explains Castro-Wright in the Wall Street Journal. The problem is that "you end up underserving everyone because you don't have an offering that is specific to that customer segment."

From retailers such as Wal-Mart to giant packaged goods companies to banks and credit card firms to industrial manufacturers and defense contractors to not-for-profits, hospitals, and philanthropies, contemporary marketers have started thinking of themselves "as portfolios of customers and not of portfolios of products," says as Larry Selden, professor emeritus at Columbia University put it in his book, Angel Customers and Demon Customers: Discover Which is Which and Turbo-Charge Your Stock. Indeed, a recent Economist Business Intelligence Unit and Marakon Associates study of senior executives of large companies found that the large majority (59%) had conducted a major market segmentation exercise within the past two years. The fanfare and attention-grabbing headlines once reserved for new products, ad campaigns, sponsorships, and distribution deals are now equally showered upon market segmentations.

Yet peel away the onion a bit and you find that beneath the grand proclamations there's quite a bit of disappointment, frustration, and poor performance. Take the case of a popular cable television channel we worked with recently. According to top management, the company was interested in an audience segmentation study to see what it could do to expand its viewership. While having the conversation, we noticed a thick binder—what appeared to be a segmentation report by a well-known consulting firm—on the bookshelf right behind the CEO's head. At an appropriate point we said, "Have you ever done one of these before? There appears to be a strategy study right behind you."

"Yes, we did, but it wasn't very helpful," the CMO reported.

We asked, "Why wasn't it helpful?"

To that the CEO and CMO chimed in unison, "We didn't know what to do with the results. The study wasn't actionable." And then we heard the all-too-familiar refrain: "Yeah, and it's become a doorstop."
That's marketingspeak for a work of little value. Its only purpose is as a shelf decoration or, literally, a doorstop. In fact, many companies state say that market segmentation reports make excellent doorstops because they are thick (often a couple hundred pages or more), heavy, and durable because they are housed in a three-ringed binder.

Using a 287-item battery of segmentation and targeting best practices, on which each practice is scored from 50 (an abysmal practice) to 100 (an excellent practice), our firm Copernicus Marketing Consulting and Research routinely finds major companies—some of them supposedly the best marketers in America—earning "D" grades (scores in the 60s) on the most basic of marketing strategy decisions: To whom do we want to talk? It's no wonder the aforementioned Economist/Marakon study also found that just 14% of the senior executives who'd done a segmentation study said they derived any value from it.

In a seminal article in the Harvard Business Review in 1964, Daniel Yankelovich introduces the idea of going beyond demographics as the basis for dividing buyers into groups. He argues that "once you discover the most useful ways of segmenting a market, you have produced the beginnings of sound marketing strategy." Forty years later, he and co-author David Meer follow up on his original thoughts: "The idea was to broaden the use of segmentation so that it could inform not just advertising but also product innovation, pricing, choice of distribution channels, and the like. Yet today's segmentations do very little of this."

Innovation guru Clayton M. Christensen, Intuit's cofounder and chairman Scott Cook, and the Advertising Research Foundation's CSO Taddy Hall also take issue with conventional approaches to market segmentation. As they write in their Harvard Business Review piece, "the prevailing methods of segmentation that budding managers learn in business schools then practice in the marketing departments of good companies are actually a key reason that new product innovation has become a gamble in which the odds of winning are horrifyingly low." They cite scenarios—based on product type and price point (e.g., this group wants a big expensive drill and this group wants a small expensive drill while that group wants a small cheap drill) and type of customer (e.g., small/medium/large businesses, heavy/medium/light users, baby boomers/Gen "Xers"/Gen "Yers"/echo boomers)—as examples of "broken paradigms of market segmentation."

The typical prescription for an ailing segmentation, however, is to just do another segmentation—only this time use a different approach to dividing up the market. If a demographic segmentation didn't give us much to go on, then how about a lifestyle segmentation? If a needs-based approach didn't work, then what if we try a jobs-based approach? The assumption here is that it was the approach itself that rendered the results unusable to the firm. But if a company in that situation spent a little more time concentrating on the underlying reasons that the segmentation studies (currently keeping its doors propped open) didn't work, then it'd discover the universal problem that plagues performance. And that problem is much bigger than which set of preselected variables it used.

Heavy on Segmentation and Light on Targeting
In his book Us and Them: Understanding Your Tribal Mind (Hutchinson, 2006), science writer David Berreby talks about the insatiable urge all human beings have to classify people. It's natural and it's understandable that we want to do that: "Categories of all sorts help explain what's happened and predict what will come next. Human kinds help predict what people will do, and there too they draw from a general capacity to find causes and patterns." So perhaps that apparent primordial need to group and sort explains why marketers become so singularly focused on the actual act of segmenting a particular population of buyers—rather than on what the results of the process are supposed to tell them. Hint: It's who to target.

Consider the reasons marketing executives give for letting the last segmentation study collect dust on their shelves. Most complain that they don't know how to find the people identified in the segments or that the people in the segments weren't different in terms of anything other than the variables that created the segmentation itself. In other words, if you do a demographic segmentation, the segments differ demographically but are the same in attitudes and behavior. If you do a behavioral segmentation, they are different behaviorally—but the attitudes, values, and media profiles are all the same. Essentially, the people identified are not different enough that marketers can find them in (1) databases such as the U.S. Bureau of the Census or (2) syndicated databases such as Prism, MRI, and Simmons.

Recently, for example, the CMO of a marketing powerhouse—the No. 1 in its industry—asked us to help explain the sputtering performance of one of its flagship brands. The executive explained, "We were doing well until last year when we decided to restructure all of our marketing efforts (advertising, product design, promotion, channel choice, pricing, and so on) based on the results of a new psychographic segmentation."

Curious, we asked about how the study was designed and executed. The CMO explained that the company gave a psychographic battery of 46 questions (e.g., "I always like to be surrounded by a lot of people" versus "I consider myself to be a loner"; "I am an extroverted, garrulous person" versus "I tend to be a reticient person with not much to say"; "In any group activity, I always find myself in a leadership position" versus "I'm much more of a follower than a leader") to 1,000 people over the Internet, analyzed the data, and clustered respondents into groups based on their answers. To the CMO's knowledge, the items used to delineate and define the groups were of unknown reliability and validity at predicting (among other things) brand-positive behavior or purchase intent in the category—or any other for that matter. Yet standing orders were to develop marketing programs against the five resulting buyer segments.

No serious thinking or analysis was undertaken to determine whether psychographics might be a useful, practical, or helpful way of segmenting that market. No attempt was made to evaluate each respondent or cluster in terms of potential profitability. No attempt was made to assess whether any of the groups could be found in existing databases. In fact, we eventually discovered that few of the psychographic items had anything to do with behavior (never mind profitability) in the category, and the five segments were perfectly flat in terms of product motivations, problems, demographics, media exposure patterns, and anything else you might care to look at. And when we did try to find the groups in databases, it was impossible; they were all the same.

We shouldn't pick on psychographics too much here because, really, most of the commonly employed approaches are heavy on the segmentation but light on the targeting. See Exhibit 1 for what we mean.

Exhibit 1 Five Segmentation approaches: pros and cons

Need/Benefit Segmentation
Pros
So-called natural segmentations—buyer needs seem so basic
Easy to do
Intellectually interesting
People love to name the groups
Good for new product ideas
Some insights for advertising copy
Cons
Needs (i.e., importance ratings) are not problems
Different techniques for measuring needs yield different outcomes
Common approaches understate the true importance of intangible emotional attributes and benefits
Generally, the segments have similar brand preferences, consumption patterns, demographics, and media exposure patterns
Segments can't be found in databases
Behavioral Segmentation
Pros
Easy to find a small group of consumers who account for the "lion's share" of category volume
Managers love to talk about the 20/80 rule
Very easy to do
Simple and capable of being understood by everyone in the organization
Because it's based on only one or two questions, the segmentation can be found in other databases
Cons
Heavy buyers are often price conscious and psychologically locked into whatever brands are available on sale
Product usage is often not correlated with any variable other than family size
Generally, the segments have similar brand preferences, consumption patterns, demographics, and media exposure patterns
Little understanding of the differential needs of the target. Compared to other targets, they are more similar than different
Psychographic Segmentation
Pros
Like need/benefit segmentation, attitudes and personality characteristics are interesting and fun to work with
People love to name the groups
Simple and capable of being understood by everyone in the organization
Offer some insights for advertising copy
Cons
Attitudes are often weak preductors of buyer behavior and brand choice
As a result, the segments have similar brand preferences, consumption patterns, demographics, and media exposure patterns
Segments can't be found in databases
Little understanding of the different needs of the segments. Compared to other targets, they are more similar than different
Segments can't be found in databases
Demographic Segmentation
Pros
Simple and easily understood by everyone in the organization
Describes people you're familiar with: your spouse, daughter, next door neighbor
Media services and agencies find demographics easy to work with
Because it's based on only 13 standard demographic questions, the segments can be found in other databases
Groups are often differentially reachable with media
Cons
Demographics rarely predict buyer behavior
Little understanding of the differential needs of the segments. Compared to other targets they are more similar than different.
Generally the segments have similar brand preferences, consumption patterns, demographics, and media exposure patters
Not being different, the targets represent an inefficient media buy
Job Segmentation
Pros
Currently popular among top management—recently featured in The Harvard Business Review
So commonsensical, it's a wonder marketers haven't discovered it before
Simple and capable of being understood by everyone in the organization
Appeals to "rational" decision makers
Cons
It's a 30-year-old approach masquerading as something new
Emphasis is on rational drivers of brand choice. Ignores the role of emotional triggers
Overlaps with "occasion"-based and "needs"-based segmentation. Hard to tell where one ends and the other starts
Segments have similar needs, brand preferences, consumption patterns, demographics and media exposure patterns
Segments can't be found in databases

 

It's Not All in the Statistical Methodology
It's important to think about a successful segmentation as the result of a process rather than merely the outcome of using one technique or another. Some researchers love to wax poetic over the latest statistical algorithm for segmenting markets. But it's been our experience over two decades that the contribution of any model is modest to a felicitous segmentation outcome. Usability, actionability, and applicability to different marketing, operational, and business decisions all depend on understanding exactly who in the company plans to use the results and how they plan (or hope) to use them. As Yankelovich and Meer rightly point out, "few marketing chiefs know or have thought about which of their company's strategic decisions would benefit from the guidance of a segmentation." That's why they need to gather representatives—from all branches of the organization—who will (if God is good) use the segmentation to ensure that the results will address as many collective needs as possible. In an up-front meeting, marketers need to tackle the following two critical research issues.

What are the number and nature of the variables to be used in the ultimate segmentation scheme? Will you use numerous variables (20-30) in a complex algorithm or a handful (5-10) to construct a fairly simple scoring scheme? And can it include database variables-customer account data, third-party data, block-level census data, or all three? We generally find that in consumer research, in which large companies will use sophisticated media plans or database modeling, a company can afford to entertain a large quantity of variables. In business-to-business (B-to-B) marketing, however, the ultimate use of the segmentation will be fairly straightforward—by salespeople or for searching business databases.

For instance, Lafarge (one of the largest diversified suppliers of construction materials in North America) wanted to segment its B-to-B base of customers to enable its sales force to quickly identify which package of products and services to offer—to increase the likelihood and profitability of a sale. In that case, a complex algorithm with multiple variables would have gone over about as well as a pregnant pole-vaulter. The segmentation criteria had to be easy and few enough for the sales force to identify, assess, and react to quickly in the field. So limiting solutions to only a handful of variables was realistic.

What does management want that segmentation to predict? How does management want the segmentation to be used? What is the research trying to explain—brand preference, openness to switching accounts, and vulnerability to leaving the customer base? Does management want to understand high versus low consumption of the product/service category or which types of consumers are going to the different distribution channels?

Constellation Wines U.S., Inc. (part of Constellation Wines, the largest wine business in the world), wanted to segment the consumer market for premium wine. But management wanted to use the segmentation to help its B-to-B customers—restaurants, bars, and liquor stores—better sell to the different types of buyers frequenting their locations. Management wanted the end result to predict brand preferences and responsiveness to different types of marketing/promotional efforts—to help bars and stores better merchandise and market to sell the most wine.

Perhaps the most important question that smart management should want answered is which customers and which segments of customers will be the most profitable for the firm to pursue for the long haul. Selden and Geoffrey Colvin, co-author of Angel Customers and Demon Customers: Discover Which is Which and Turbo-Charge Your Stock reveal in piece for Havard Business School, "M&A: The Value of a Customer," that "a surprisingly large percentage of executives we've talked to believe their companies have no unprofitable customers, which is virtually never true. When asked to name their most profitable and least profitable customers, most executives name the wrong ones or simply have no clue." Not surprisingly, the vast majority of segmentations on which we're brought in for a postmortem do not have a good measure of profitability built into the methodology.

Profitability by Proxy
"How many CMOs," Selden and Colvin wonder, "are partnering with their CFOs to create comprehensive and regular reporting on customer profitability? Note that we said comprehensive—that is, profitability for each customer based on total revenue and expenses, including capital costs. We also said regular as in monthly. Search as we may, we can find few companies that do this." So if companies don't have any existing data on profitability for current customers (never mind prospects), then what's a marketer to use to (1) assess the potential profitability and calculate the economic value of individual customers and (2) determine which of the myriad possible variables out there that could be used to divvy up the market are predictive of it? Enter profitability by proxy.

In addition to financial measures of the revenue (e.g., lifetime value, current spending in category in dollars, current share for your brand today) and cost (e.g., cost to reach and influence with sales force and media, cost to deliver and serve) sides of profitability, there are also several important stand-ins that reflect:


You don't want to predict who has bought the brand today; you want to see who might buy it tomorrow. Those measures provide a far more robust picture of the value of individuals—be they current customers or prospects—to the company, and they get much closer to the "comprehensive" ideal to which Selden and Colvin refer.

One of the hottest areas in marketing today, for instance, is the idea of responsiveness to your brand. There are folks who are willing to consider your brand if they are not using it already; they are willing to try it. If they are already using it, then they will try a line extension. They know your brand exists and have positive feelings about it. If someone hates your brand or has no interest in it, then that person is unlikely to ever move in your direction—no matter what you do. He will not even take the free sample you hand out on the street. He is unresponsive and a drain on profitability. So why waste your time on him?

Opinion leadership is another proxy. Again, we find a continuum; almost nothing in marketing is clearly black or white. At one extreme are the people so disengaged in the product category, or with such a small network of friends and acquaintances, that they don't exercise any more influence on behavior than what they buy and use. At the other extreme are the raging fans: people who not only buy and use the product and service but also are knowledgeable, have a wide circle of friends, and are both consulted and volunteer their expertise. By way of example, those are the car enthusiasts who just love cars, read everything about cars, and have a large network of acquaintances. And computer enthusiasts with many acquaintances affect many more sales than their own; they have value far in excess of what they buy.

Customer price sensitivity is another important indication of a buyer's value to the firm. The research we've done at Copernicus has consistently demonstrated that in almost every product category there are many people—as many as two-thirds—who are relatively price insensitive. While you probably can't e do not state that you can double the price, but in most cases you can increase the price by 10% or 15%. And obviously the higher the price you can charge, the better the margin and bottom line. Other possible proxies include interest in new products and services, decision-making power, growth potential, and magnitude of problems that—if solved—would lead the customer to switch.

Taking Your Mother's Advice
So now that we've got some measures of profitability and economic value, let's talk about what descriptive, definitive variables can and should be used to segment a market. The answer is we just don't know. We won't know, in any given market, what will be predictive of profitability and desirable behaviors until we talk to current and prospective customers, collect some data, and do some analysis. We imagine this is a scary prospect for most folks: to move ahead without knowing exactly what you're going to end up with. We suppose it's just human nature to want to select a certain set of variables—be they psychographics, attitudes, demographics, or behaviors—ahead of time, so oneto feels more in control of the final outcome. Particularly if the results sound like they'll be cool and fun (so, at least in theory, they're easier to shop around the company and show the CEO), why not just pick one set?

But didn't your mother ever tell you, "Don't put all your eggs in one basket"? Why bet only on psychographics or any other exclusive set of variables as the key predictors, particularly in the absence of any evidence that they are—investing significant time, money, and brand equity (not to mention personal credibility) in a market segmentation when you don't have to? Remember, only a fraction of the people who'd done a market segmentation study in the past few years felt they derived any value from it, and they ended up having to do another one. And we'd venture that a fair quantity of them didn't take their mothers' advice.

The kind of segmentation that forms the foundation of a great marketing strategy provides a detailed, well-balanced picture of the buyers in different groups. In other words, it tells you more about them than just their genders, ages, attitudes, or needs; it tells you all those things and more. The process that leads to that kind of segmentation involves testing hundreds of diverse variables including needs, psychographics, demographics, behaviors, and media preferences. We've found in our work for clients, for instance, that we might easily have 150 or more independent variables from a typical in-depth interview with a consumer. Those often consist of 15-20 demographics, 40-50 attitudes, 20-40 behaviors, 30-50 motivations to buy in the category and the brand, 20-30 media habits, and numerous database variables.

Some of those are relatively simple variables, such as gender, age, marital status, and purchasing behavior. Others are more complicated, such as relative income: the subject's income relative to the same ZIP code, block, or age group. Sometimes very complex variables are created, such as household income per capita compared with that of close friends and relatives. Factor analysis can reduce the large quantity of the variables; it can take the 150 and find the things they have in common to reduce the list to 30 or 40.

Test all those possible variables to see which are related to profitability or proxies for profitability. Then identify the key variables about people—or firms for a B-to-B product or service—that best predict what they are doing or will do (or both) in the marketplace. At that point, you can group consumers into segments based on their answers to those 5-25 variables. The company can use different methods here, such as cluster analyses, latent class analysis, and neural networks.

Finally, test again, looking at numerous segmentation schemes. Apply the managerial, statistical, and financial criteria you worked out up front to evaluate different solutions. Do they answer management's questions? Will the sales force be able to use them? Are there different brand preferences, different consumption levels, different channels, different media profiles, and—very important—different levels of future profitability? At the end of the process, if you aren't about to walk away with (at the very least) a set of segments that provide a clear understanding of (1) which group or groups represent the best profit opportunity and (2) what distinct needs they have that can be addressed and marketed to, then you need to keep working.

A Doorstop in the Making
The "original and true purpose" of a market segmentation exercise was "discovering customers whose behavior can be changed or whose needs are not being met" (per Yankelovich and Meer) and identifying a target group of buyers. It was not supposed to be a de facto manufacturing process for shelf déecor and doorstops.

To ensure that a company's segmentation efforts remain current and—you'll excuse the expression—on target, top management and CMOs need to ask themselves: "Have we segmented each market in which we operate to identify and describe the most profitable market targets to pursue?" A good segmentation provides a company with clear direction on which group represents the best target—one that had a high economic value to the company and can easily be identified in the population or in customer databases. If a segmentation meets those requirements, then it will pay for itself many times over. If it doesn't, then it will become a quickly discarded waste of resources—gathering dust on a shelf.

For more information about market segmentation and targeting, click here.

Back to News Room Index