Customer-Driven Distribution Systems

Too often, distribution is the neglected side of marketing. Automobile companies, savvy in many aspects of strategy, have lost huge shares of the parts and service markets to NAPA, Midas, and Goodyear because they resist making changes in their dealer franchise networks. A great many other American companies—Tupperware springs to mind—are reaching their markets in similarly outmoded ways. It is hardly seemly for Tupperware to continue with its “parties” when more than half of American women are working outside their homes.

In contrast, a number of companies have outstripped their competition with imaginative strategies for getting products to their customers—and marketing executives can learn from them. The Federal Express system is so innovative and formidable that it might be considered a model even beyond the small-package delivery industry. American Hospital Supply has gained the edge over its competition by linking up to hospitals and clinics with a sophisticated system of data processing, while Steelcase has set a standard for delivering complex office furniture installations, complete and on time.

Although American companies have been ignoring the ways in which they deliver products and services, their customers are increasingly inclined to demand higher standards of performance. Customers want companies to value their time and trouble.

And so important opportunities for gaining a competitive advantage through distribution remain, and given the new technology, some companies may, as Federal Express has, achieve a breakthrough. Will the management of American companies (deregulated telecommunications companies included) make use of these opportunities or even recognize them for what they are? Just what process should a company use to select or structure the best possible distribution channels for its products?

We suggest eight steps to design a distribution system that really performs. The word process is key here because whatever the result of taking these steps, management will gain by clarifying what its customers want and how to serve them. Managers are always saying that they want their company to be “market driven.” In following these steps, they can give substance to what is too often merely corporate rhetoric.

Step 1: Find Out What Your Customers Want

Of all marketing decisions, the ones regarding distribution are the most far-reaching. A company can easily change its prices or its advertising. It can hire or fire a market research agency, revamp its sales promotion program, even modify its product line. But once a company sets up its distribution channels, it will generally find changing them to be difficult.

And so, the first step calls for researching what customers want from the buying process and then using their preferences to group customers into market segments. Managers conducting the research concentrate on learning what their ultimate customers—the end users—want in the way of service. It is these people, of course, who actually benefit from the products a company makes.

It is important for the researchers to emphasize that the product’s quality is not an issue. Nor should there be any question at this stage of what may or may not be most practical for the company, whether it be a service company, a manufacturer, or a middleman. Rather, respondents should be encouraged to consider the delivery of the service, the convenience of shopping for the product, and the kind of add-ons that are sold along with either.

There is, of course, no such thing as a truly homogeneous market, in which all customers view the company’s offerings in exactly the same way. Yet managers who routinely try to ascertain what market segments are worth preparing for when they design a product rarely try this when they make decisions about how to distribute it. This is a crucial mistake.

The preliminary research is meant to generate an inventory of customers’ desires, but it is important to exclude ideas too grand or trivial for consideration. Without restrictions of any kind, who wouldn’t ask for the moon? Needless to say, an overarching consideration is price: respondents should be made to realize that for every service (or lack of one) there will be a correspondingly higher or lower price. Equally important, however, respondents must be forced to weigh their preferences not only in relation to price but also in relation to one another.

Consider personal computers. The delivery of service might include such things as a demonstration of the product before sale or the provision of long-term warranties and flexible financing. After the sale, there might be training programs for using the equipment and a program to install and repair it. Customers might appreciate “loaners” while their equipment is being repaired or technical advice over a telephone hot line. They should be prepared to make trade-offs among these inducements.

Services, we’ve found, usually fall into five categories:

1. Lot size. Do customers want to buy in units of one or in multiple units?

2. Market decentralization. Do customers value around-the-corner convenience, or are they willing to deal across great distances, say via an 800 number?

3. Waiting time. Do customers want immediate delivery, or are they more concerned about the assurance of delivery?

4. Product variety. Do customers value having the choice of many related products, or do they prefer the store to specialize?

5. Service backup. Do customers want immediate, in-house repair and technical help, or can they wait and choose their own local repair services?

Once customers have traded off, say, demands for convenient location against product variety or variety against expert sales assistance, researchers can group these preferences into market segments and look for links between the segments suggested by the survey and the segments that may be generated by analysis of independent demographic or other marketing data.

We suspect, for example, that a segment of small businesses would be much more concerned about one-stop shopping than large businesses; big companies have purchasing specialists with the time to choose complementary products from different sources and to secure the lowest prices within various quality ranges. If a company sells to people who want one-stop shopping, it might want to know whether this segment coincides with self-employed accountants, for example. This small market segment is likely to be substantial, and it has needs quite different from those of a segment consisting of start-up scientific research companies.

A number of marketing research techniques are available to researchers at this step in the process, among them conjoint analysis, hybrid modeling, and constant-sum scales.1 Unfortunately, most of these techniques have been developed to elicit choices among the tangible properties of product design: gas mileage versus size of car, size versus model, and so on. The things people want from a distribution system tend to be less tangible and more difficult to visualize and make judgments about (convenience of location versus depth of assortment, for example). Survey instruments ought to be designed with this challenge in mind.

Step 2: Decide on Appropriate Outlets

At this stage researchers focus on the relation between market segments—defined as clusters of demands for service—and the outlets where services are normally delivered. Suppose, for example, that customers for a home computer indicate a desire for self-service, a somewhat narrow assortment of merchandise, limited after-sale service, and a relatively Spartan atmosphere—so long as the prices are low. Clearly, this segment consists of people who would put up with a discount store operation—a 47th St. Photo, for instance—and trade off the amenities of upscale service or nearby location.

The fame of a store such as 47th St. Photo can be an asset in the analysis. Using the names of such well known existing outlets or suggesting a hybrid of two or more kinds of such outlets, researchers can label potential clusters of service attributes. Respondents are asked about the service outlets they visualize, and researchers label the clusters constituting a segment precisely and vividly. On the other hand, labels are merely points of reference. They suggest existing kinds of retail outlets without limiting the possibilities.

For clusters suggesting no existing kinds of outlets, short descriptions of hypothetical outlets may be of help. Researchers may coin new names and, in analyzing the data, position the various segments along a wide continuum. The chemical industry, for example, may have no analog to a discount store or a rack jobber. But if many respondents indicate that they would like to see something along these lines, then the research team might, in the course of the survey, develop an appropriate option, describe it, think of a label for it, and present it to new respondents for consideration.

Venturesome financial institutions such as Merrill Lynch, Bank One, and GE Credit have scored impressive gains with just such distribution ideas. How else did we get to “financial supermarkets” and “discount brokerage”? In contrast, many marketing strategists in the personal computer industry have failed to predict the significance of value-added resellers or retail outlets with multiple but highly focused assortments. Obviously, they did not start by conceiving their distribution channels according to the shopping needs of potential customers.

Do not be hamstrung by industry experience. The more creative researchers are with their labels, the better step 2 will work.

Step 3: Find Out About the Costs

Up to this point, the customer is sovereign: the process aims to determine what customers perceive to be optional shopping conditions among the many pertaining to distribution and related services. In the first part of step 3, however, it is essential to obtain an impartial assessment of whether the things that customers want (more precisely, the “clusters” of things they want) are feasible for the company. This is the first reality check, one that is made before management as a whole gets involved in the process.

It may be made by selected members of the corporation’s staff, assuming they are professionals who can be objective about the company’s line operations. Otherwise, the company must turn to executives from unaffected wholesaling or retailing enterprises or to academic authorities.

Researchers have already asked customers to trade off their demands for service against price, so utterly implausible combinations of shopping conditions—outlets combining small-lot purchases and low unit prices, for example—have been eliminated from further consideration. But less obviously implausible combinations may remain. Suppose a group of customers for personal computers claim they are willing to pay any price for a hypothetical shopping outlet combining custom tailoring with quick delivery. Are these two shopping conditions ever practical in combination?

The second part of step 3 aims to determine what kind of support will be needed from suppliers or other “up-channel” participants for any hypothetical outlet suggested by the data. Distribution outlets do not operate in isolation; there is always a distribution system backing them up.

For example, if an attribute cluster suggests a “limited line, full function, vertically oriented industrial distributor,” the question would be this: What backup system ensures that this kind of distribution will satisfy customers as well as possible? The answers should be concrete: high-technology distribution centers, training programs, catalog expertise. Sometimes existing distribution systems enjoy the necessary support, sometimes not. If not, the division of labor among suppliers will have to be restructured so that what customers desire may be delivered by the most capable up-channel participant.

Step 3 is a good time to get insights from people out in the distributive trades. It is also the time to tap in-house knowledge, the opinions of salespersons and others who stay in contact with customers.

The third and final part of step 3 is to project the cost of support systems feasible for each outlet type, on the assumption that the company may be able to contract with third parties to perform the outlet functions. Researchers cost out the new support systems on an incremental basis, starting with the company’s existing distribution system. Costing requires informed guesswork; any change in one element of a distribution system has ramifications for another. But if, for example, the data suggest that customers want rapid delivery, local inventories will have to be maintained. Distribution centers may have to be constructed to support the local inventories. Cost accountants familiar with distribution may provide estimates, although they may have trouble dealing with the more theoretical scenarios. In the end, the question to be answered is this: What increase in market share is required to offset the added costs of the new distribution alternatives?

It is important to collect these cost estimates during step 3 because they are backup material for step 4. The figures may well reveal that certain systems of distribution are prohibitively expensive and should be removed from further consideration. We know one manufacturer of specialty medical supplies and equipment that was losing sales to competitors selling via mail order. But the added cost of establishing a competing catalog system did not make sense, so the company abandoned the option at this stage.

Step 4: Bound the “Ideal”

At this point the researchers have come as close as they can to discerning an ideal market-driven system. Top management has been obliged to keep its hands off. Researchers have had a chance to find out, perhaps for the first time, what it really takes to please customers.

Step 4 gives a cross section of the company’s executives an opportunity to subject the research findings to their own hard tests. Researchers invite these executives to investigate how any existing or hypothetical channel of distribution would affect company efficiency (costs, revenues, and profits), effectiveness (especially market share), and adaptability (fluidity of capital invested, ability to accept new products or adjust to new technologies). At the same time, executives give their impressions of what distribution is or is not doing. Though this part of the process is meant to generate reliable numbers, discussions with managers should be open-ended. They may even bring up their pet peeves.

Finally, researchers develop a list of company objectives for distribution based on their conversations with top management. They turn this list into a survey instrument and send it to every executive in the company who has a stake in distribution matters. Executives trade off objectives in the same way that customers trade off their requirements for outlet design. The result is a list of weighted objectives that are the constraints bounding the system.

Inevitably, at this stage, some executives want to impose constraints on the distribution design, which they justify not so much by numbers as by industrial tradition. There are rigidities and prejudices in most industries, some of which are reinforced by law, some of which are perceived to have the force of law.

The faltering car dealer system has not been altered since the mid-1920s, in part because of peculiarities in the legal structure of auto distribution (franchise laws, dealer-day-in-court laws). But there is also an industry folklore that gets in the way of change, even though auto companies face a shift in power to consolidating dealers. How much longer before the executives of Chrysler, GM, Toyota, and other companies will be forced to compare their old objectives with new options?

The Coca-Cola Company and PepsiCo, in contrast, are consolidating their traditionally independent franchise bottler networks into distribution systems with greater maneuverability. At IBM, distribution by means of a direct sales force had been a sacred principle essentially until the company started making personal computers. It finally began to use third parties but only after great internal strain, after which the personal computer division was accorded the status of an independent business unit.

Step 5: Compare Your Options

With the completion of step 4, company researchers will have a weighted list of management’s objectives and constraints on the one hand and on the other a roster of the various ideal, market-driven distribution systems generated earlier in the process. Step 5 requires them to compare these two sets of data with each other and also with the system of distribution already in place. The researchers will, of course, consult with distribution managers about the company’s present system: structure, functions performed by various channel participants, costs, discounts, and the like. It may be necessary for researchers to undertake an analysis of volume flows by channels as well as by margins, functions, and value-added at each level. A reasonably detailed map of this type can be very illuminating.

One of three conclusions will emerge from these comparisons. First, the existing system, the management-bounded system, and the ideal system may closely resemble each other. If this is the case, then management knows for sure that the existing system is about as good as it can get. If customer satisfaction is mediocre nevertheless, the message should be clear: the fault lies not in the design of the system but in its implementation.

Second, the existing and management-bounded systems may be similar to each other but substantially different from the ideal. This outcome may mean that the objectives and constraints adopted by management are causing the gap. Such a finding calls for a careful investigation of management’s perceptions, the purpose of step 6.

Third and especially sobering, all three systems may be substantially different. Assuming that the management-bounded system is positioned somewhere between the existing and the ideal, it may be possible to improve customer satisfaction without relaxing management’s objectives. This is the time to ask if relaxation of certain management constraints might not produce even greater benefits.

By 1980, IBM’s direct sales force and sales branches had formed the core of the distribution network for its existing line—mainframe and word processors. These channels could not, however, be cost-effective in delivering personal computers to the small business market—not, in any case, at the standard for customer satisfaction that IBM’s executives considered their company’s hallmark.

The ideal would have been a network of highly decentralized, service-intensive specialty stores carrying an assortment of personal computer brands and models as well as other types of office equipment. Because some IBM executives were convinced that the company could not maintain control over the quality of service without ownership, the company opened its own retail outlets to sell IBM equipment alone. IBM product centers offered the consumer a variety of equipment, but comparison shopping within them was impossible. In 1986, IBM sold off its product center network to NYNEX. (Interestingly enough, IBM has since come to realize that the small business market is so heterogeneous that it consists of multiple segments.)

And so the ideal system acts as a stake in the ground. If the management- bounded options are not reasonably similar to the ideal, then researchers will ultimately have to confront managers with the fact that the company has been sacrificing customer satisfaction to other objectives.

In the long run, some of these other objectives may be critical and may even supersede the effort to satisfy customers via distribution. When management decides on any new strategy, it will simultaneously establish a hurdle rate—a minimum projected return on investment that justifies going ahead. Managers may, of course, set hurdles incorrectly, not only because they miscalculate costs but because they acquire a prejudice for or against particular channels of distribution. In any case, distribution strategies that do not clear their hurdles should be dropped from consideration in step 5.

Step 6: Review Your Pet Assumptions

This step is meant to help distinguish a serious constraint from an ordinary prejudice. It entails bringing in outsiders—lawyers, political consultants, distribution experts from other industries—who will call management’s assumptions into question. Management often protects the status quo, for example, by claiming that changes might violate the law or encourage shadowy activities. Outsiders can look at the relevant laws and ask if they are what they seem. Can’t they be changed? Does holding to one value force the company to sacrifice another?

The automobile industry has clung steadfastly to the dealer franchise system, in part out of fear of legal tangles. Porsche’s attempt to implement a more consumer-responsive approach to distribution in the early 1980s turned into a fiasco largely because Porsche’s dealers made clear that it would keep the company tied up in the courts for a generation. Alas, Porsche was on the right track.

But the impulse to stand pat does not always stem from anxiety about the law. The use of authorized third-party outlets for personal computers is an example: it often portends gray market activities. Some time ago, top managers at IBM indicated that they had been worried about the price cutting and “footballing” that would result if they authorized third-party outlets—a concern that proved justified. Had they let this serious concern paralyze them, their personal computer division would never have expanded as quickly as it did.

And so during step 6, outside authorities should be called on to check whether legal and other constraints exist and, if they do, whether they can be overcome. Of course reliance on outside experts can be risky. Who is to say top management doesn’t know what it is talking about? Who can tell what course a lawsuit will take or what laws Congress and state legislatures will enact?

Business decisions are based on judgments, not certainties. Merrill Lynch would never have launched its highly successful cash management account program if it had not altered its assumptions about how the SEC would enforce federal banking laws. What are other companies missing?

Step 7: Confront the Gap

This is the climax of the process. It requires top management to confront the gap between its practices or objectives and the ideal. For the first and only time, managers conducting the research bring together all executives responsible for distribution to determine the shape of a new system. To underline its significance, the company holds the meeting somewhere offsite.

The researchers get things going by presenting the ideal distribution system. Then they share the results of steps 4 and 5. In the course of this discussion, researchers outline for top management the objectives and constraints that were used to bound the ideal and show their effect, if any, in limiting what customers really desire. Next, researchers present the data and expert opinions challenging the validity of management’s objectives and constraints—what was gained from step 6.

All this information serves as background for what usually proves to be a provocative discussion. We have found that researchers can prompt openness to it if they use computers to readjust weightings or other data and display the results instantaneously. This session brings top management face-to- face with the folklore restricting its thinking. Executives compare alternatives, weigh opportunity costs in relation to risk and exposure, and consider a host of other quantitative and subjective variables that are all too easily buried under day-to-day affairs. Most important, they make decisions in a new context—one in which an attainable ideal has been delineated, the intervening distance between the ideal and the reality has been measured, and the obstacles to closing the gap have been made explicit.

Such was the case for a personal care products company, whose ideal suggested the elimination of one level in its system—the brokers. It was a big step for managers to contemplate. When the company’s brand lacked visibility and strong consumer demand, brokers had played a key role in providing access to the retail trade. Management felt a strong sense of loyalty and indebtedness to them. Over the years the brand had emerged as the best seller in its category; now the brokers contributed little to volume. Indeed, a growing price sensitivity on the part of consumers, coupled with the inefficiency of the broker system, placed the manufacturer in a vulnerable position.

It’s not important to know this company’s final decision. What is important is how the process teased out the lines of a crucial choice. Apple Computer, for example, would not likely have experimented with mail order channels in its early history had it followed this line of investigation to its conclusion. It would have found that the amount of hand-holding required to make a personal computer plus a software sale is extremely high. Similarly, IBM would not have been so surprised to find that dealers with outbound sales forces have greater staying power than those who simply rely on inbound retail sales. IBM retail showrooms cannot provide the kind of in-depth analysis and training that visits to a customer’s premises can.

Step 8: Prepare to Implement

The final step in the process modifies the ideal distribution system emerging from step 3 according to the final objectives and constraints established in step 7. What managers are left with is a good system—not ideal, perhaps, but optimal.

This should be the subject of intensive implementation planning. And it is important for senior managers to help implement the system, if for no other reason than to give them a personal stake in the outcome. Besides, having confronted the ideal and having tested it against the other options, management has a full understanding of the trade-offs as well as the obstacles to implementation.

When it comes time to change the existing distribution system or to scrap it entirely, managers should test modifications on a small scale before committing resources to them. The major problem is that word will spread quickly. The gossip network among dealers and distributors is one of the busiest around.

The process we lay out in this article is not a simple one. Managers are required to focus on something as insubstantial as quality, or the ideal system, and then to come up with hard numbers to project a reasonable ratio of return to expense. They must even anticipate how adaptable their ideal might be to changes in the law or the political environment. Clearly, there is as much art here as science.

Still, none of the eight steps we outline should be skipped in the interest of apparent expediency. Managerial sophistication will speed things along, but sophistication alone will be no substitute for going through the entire process. With all the effort reevaluating the system requires, readers may assume that the process always justifies itself by the constructive changes it brings about. In fact, its real value is in the clarity it brings to a critical aspect of doing business.

A specialty grocery products manufacturer discovered that it was getting its products onto supermarket shelves in ways that on the surface looked Rube Goldbergian. It was using an array of third-party players, including food brokers, grocery wholesalers, and health food distributors, some of whom carried out a remarkable range of functions between the manufacturing and the retail level of the distribution chain. When the company drew a structural diagram, it looked like a bowl of spaghetti. Nevertheless, further analysis revealed that the system met all the criteria of an ideal.

The recommendation? “Don’t mess with it! Don’t touch a thing!” Sometimes the eight-step process explains precisely why you should do nothing to change the distribution system you already have.

A version of this article appeared in the July 1987 issue of Harvard Business Review.