Use Strategic Market Models to Predict Customer Behavior

Reading Time: 19 min 

Topics

Buy
Like what you’re reading?
Join our community
Member
Free

5 free articles per month, $6.95/article thereafter, free newsletter.

Subscribe
$89 $45/Year

Unlimited digital content, quarterly magazine, free newsletter, entire archive.

Sign me up

In the late 1980s, the computer systems industry embarked on a perilous but thrilling white-water journey of change that left many previous giants upset or beached on the rocks, while young, unknown companies emerged from the chute as celebrities. In retrospect, the undercurrents of change seem clear, but, at the time, the level of confusion was tremendous. Hewlett-Packard’s successful navigation through this treacherous passage is attributable to many factors, including its new approach, called strategic market modeling (SMM), which HP’s computer systems organization used to generate and test alternative business strategies both for itself and for competitors.

The computer systems industry (traditional mini-computers and mainframes) in the late 1980s was already experiencing the impact of powerful desktop PCs and workstations when it was hit from a different direction by the growing acceptance of Unix and open systems. For HP, while the emergence of commercial Unix represented a tremendous opportunity, it was also a threat. Unix had been a popular computer operating system among scientists and engineers for many years, due to its portability across different platforms and its networking features. It began to penetrate the commercial market, as MIS managers introduced Unix into their departments for the same reasons. But Unix was only a small part of a larger technology shift to the “open systems” that have subsequently revolutionized the computer industry. Open systems have meant the standardization and interchangeability of components, with companies leveraging each other’s R&D investments, resulting in great economies of scale for computer customers.

HP managers hoped that Unix would thaw the minicomputer marketplace, where most customers were already frozen into one or two proprietary vendors, primarily IBM or DEC. Market shares between vendors seldom changed by more than 1 percent per year. Even though HP had only a 6 percent market share in the very large commercial minicomputer market, HP managers viewed the proprietary HP 3000 business as a success; it was highly profitable and growing in excess of 15 percent per year. The introduction of a competing Unix offering, the HP 9000, would cannibalize this business.

This problem was compounded by the fact that if HP were to sell both product lines to information technology directors around the world, the same sales force would have to do the selling. HP could not have two sales representatives calling on the same customers and recommending two different solutions. Consequently, the sales-people needed to understand both the relative strengths and the future direction of each product line, so they could help each customer make the best decision. Thus a very serious question was whether HP should offer more than one product line.

In other industries, such as the auto industry, positioning multiple product lines within a single market is a common practice. The concept of similar, yet differentiated, overlapping product lines is a way to increase market coverage. But, at HP, where one group manager described managing his divisions as “herding cats,” the natural inclination was for both the HP 3000 and HP 9000 businesses to look at the market independently and to simultaneously target the same part of the market, thereby positioning themselves on top of each other. This situation had already developed, and the intent of the SMM project was to correct it.

In essence, the issue was to position two competing businesses, one new and high-growth, the other established and mature, in order to maximize shareholder return. HP had two principal objectives:

  1. To accurately and impartially determine potential alternative strategies that would maximize the combined profitability of the two businesses.
  2. To achieve agreement across HP’s computer systems organization on the position of the two businesses.

As valuable as this endeavor was for HP, its greatest potential value lies in its lessons for the future. My purpose here is to describe both HP’s methodology and, more important, what we have learned.

What Is Strategic Market Modeling?

SMM uses several market research methodologies, including choice modeling, cluster analysis, and needs segmentation, which have all had value in consumer marketing. Their application in high-priced, complex industrial markets like the mid-range computer systems business is relatively uncommon. HP’s implementation was unique in two important respects. First, the issues required a higher level of abstraction than the product or feature focus with which choice modeling is typically applied. Second, the business is so complex that it requires a series of tightly coupled choice models (one for the U.S. market, one for each of three European countries, and one for U.S. distributors, resellers, and consultants) to capture the marketplace dynamics.

SMM relies on quantitative research to build a database for simulating the marketplace. Each customer’s unique needs and perceptions are captured via survey, preserved intact (not averaged), and ultimately used to predict each respondent’s purchase decision, given various hypothetical choices of products and services. Once a company establishes the customer needs and perceptions database, in addition to the more traditional survey analyses, the company can design “what if” scenarios and run simulations to predict market behavior, testing what would happen if it or a competitor improved a product or service in various ways or introduced a totally new product or service. These simulations of potential changes in market behavior can be done in addition to more traditional survey analyses.

Why Did HP Use SMM?

SMM’s creation was inspired by a sense of desperation: the two businesses were locked in a bitter struggle for the market, the customer base, and the channel, and no one knew how to resolve it acceptably. The positions were clear. The upstart HP 9000 division managers wanted to be unfettered in their pursuit of growth, regardless of the consequences for the HP 3000 business. The HP 3000 managers did not want the HP 9000 business to cannibalize the customers and markets they had spent years developing. All previous efforts at mediation had failed, and the company had run out of approaches to resolve the issue. For example, the latest task force had negotiated an agreement to partition the market by industries — the HP 3000 would be sold to some industries, the HP 9000 to others. The agreement had lasted less than a month when a leading HP 3000 software supplier in an HP 3000 industry decided to shift to Unix — either the HP 9000 or a competitor’s product.

HP required an impartial, fact-based method to put customer and shareholder needs above internal politics. The group marketing manager had the responsibility for resolving the issue in a manner acceptable to both product divisions and championed the SMM approach to division general managers after the group strategic planning manager had proposed it. SMM promised to be a data-driven methodology so rigorous, objective, and impartial that all parties would accept its results, and it was specifically designed to meet this requirement. If all parties could collaborate on the design of the research and watch it meticulously executed, with the resulting data used to test alternative strategies, then all would accept the conclusions. The intent was not to surrender the company’s business strategy to computer algorithms, but rather to use the tool to explore choices and directions impartially. Ultimately, those accountable for the results would need to make the tough decisions.

How SMM Works

SMM uses survey techniques to capture three types of data from a customer sample: demographic (classification) data, user needs data, and competitive perception data. For this project, we targeted a written survey at certain percentages of HP and non-HP customers: CIOs and IT directors, companies in various industries and of various sizes, Unix and non-Unix users, and so on. Demographic data from each respondent included categories that were subsequently used in sorting the data, such as industry, job, organization size, application information, first choice, present usage information, Unix usage, networking environment, and adaptation sequence (early adopters, early and late majorities, late adopters, and so on).

We obtained user needs and competitor perceptions data using the research vendor’s proprietary techniques (see Table 1, Part A). The relative importance of user needs was captured in a method similar to the $100 test, in which the total weighting of each respondent’s needs equals $100, and the perception of how well each competitor meets each need is scored on a ten-point scale. Once each customer’s unique response was preserved in a database, we could predict each customer’s purchasing behavior using a variety of algorithms, such as the simple “winner takes all” algorithm, in which the company that comes first in meeting a particular need gets the total value of that need (see Table 1, Part B). In this case, Competitor A wins because its total score is $46, whereas the HP 3000 score is only $33.

The testing of alternative positioning strategies is perhaps SMM’s most valuable use. To illustrate, using the previous example, suppose the HP 3000 was suddenly improved sufficiently so that everyone in the database immediately perceived a 20 percent improvement in its performance. How many customers would change their purchase decision? Alternatively, what would be the effect of a 20 percent improvement in availability? By using the classification data, we could profile the customers whose purchase decisions would have changed and the competitors who would have lost business as a result of this strategy. For example, a 20 percent improvement in this particular customer’s perception of HP 3000 performance would have caused the customer to switch its purchase decision to HP, whereas a 20 percent improvement in availability would have had no effect, since the customer already perceived HP to be first in availability. Conversely, we could test a product line’s vulnerability to a decline in each variable.

We could also test combinations of improvements. For example, what would be the effect of a 20 percent increase in the first and fourth variables, as opposed to a 20 percent increase in the second and third? Thus we could test the relative attractiveness of alternative positioning strategies based on multiple variables. Each simulation revealed not only share increase but also share decrease for each competitor — who lost as well as who gained.

The premise behind the SMM project was that this technique could be used simultaneously on multiple product lines. For instance, suppose the HP 3000 were improved in performance, availability, and ease of use, while the HP 9000 were improved in service and support, operating cost, and growth path. What would be the net effect on the individual shares and the combined HP share in the sample? HP tested alternative strategies for the two product lines in exactly this way. It also tested competitor scenarios in a similar manner. What would be the optimal strategy for a competitor, and, alternatively, where would it be most vulnerable, i.e., where would a deterioration in its relative performance hurt the most? What if the HP 3000 were improved in one way, but Competitor A improved in another?

How HP Used SMM

At HP and elsewhere, regardless of the brilliance or quality of analysis in business planning, success or failure is usually determined by how well a company addresses human issues. If the purpose of a strategic planning project is to set an organization’s strategic direction, how can the direction be changed unless people change their behavior, and how can behavior change unless their beliefs about the market and the business also change? We believe strategic planning to be a battle for people’s hearts and minds (at all organizational levels), rather than an analytical exercise. Mintzberg asserts that strategic planning doesn’t work because it is done by left-brain planners rather than right-brain decision makers.1 Overall, Mintzberg may be right, but the reason that strategic planning works at HP is because the decision makers usually do it. Consequently, HP business plans are not elaborate but are well executed.

For the SMM project, the group marketing manager formed a steering committee with the two division general managers as a board of governance for the overall project. Joint working teams (representing both divisions) and a core team of the group planning manager, the division planning managers, and the group marketing communications manager did the actual work. The division planning managers decided that the original design for the research was too narrowly defined, changed it, and took ownership of it. The division planning teams ran their own analyses of the data separately and together in the core team. The division marketing managers decided to become directly involved and joined the core team in a series of all-day working sessions in which they used the model themselves.

When the joint teams designed the surveys, they carefully chose classification questions to support every segmentation model that the businesses proposed. An important feature of SMM is that it supports a dynamic segmentation process by which the data can easily be sorted and analyzed in various ways. The teams tested and found wanting dozens of segmentation models because, when they analyzed the needs profiles of the different segments, none was appreciably different from the overall profile (see Figure 1).

Did this mean that segments did not exist — that, in fact, the computer systems market was a single homogeneous market without segments? The team began to panic, since an implicit assumption in the process was that the two businesses could be differentiated by attuning them to the needs of different segments. If there were no segments, there was no basis for their differentiation.

Fortunately, the project plan had budgeted for statistical analysis to identify naturally occurring clusters or segments of customers. This analysis resulted in the only segmentation method that revealed substantially unique customer segments (see Figure 2 for profiles of the four segments we identified). The profiles are very different from each other; traditional segmentation techniques — classification data — had revealed segments that looked remarkably similar to the overall profile (see Figure 1).

The team further studied these segments and their demographic profiles to construct a picture of the segmentation model at a higher level of abstraction by including classification information about the various segments (see Figure 3). This needs-based segmentation model was excellent for “inbound marketing” — guiding engineering on how to design the product — but it was much less suitable for “outbound marketing” — designing tactical sales and marketing programs. To this point, we had always assumed that inbound and outbound marketing needed to work from the same segmentation model. We reexamined this assumption and realized that we could use different segmentation models for inbound and outbound marketing as long as they were consistent. For example, a factory could build products for the availability/performance segment, while outbound marketing could craft a mainframe-downsizing program for specific industry or application segments. (Other parts of HP reached the same conclusion, and some now use three or four segmentation models simultaneously, each focused on a particular functional problem, i.e., product design, marketing communications, or sales support.)

The segmentation model gave us confidence that both businesses could be differentiated. For example, the HP 3000 business could focus on the availability/performance or the solution segments, and the HP 9000 could focus on the transparent access or the value segments. SMM provided the means for testing the relative merits of these strategies. We could simulate the strategies, and the model would predict how much market share increase HP could achieve.

SMM proved to be only a starting point for identifying viable strategies. It allowed us to move beyond segmentation as a basis for strategy formulation because we could test any positioning proposal regardless of how it was formulated. For example, we could try out positioning strategies that were intuitively derived, based on the experience and insight of the general managers, marketing managers, and business teams. The team members canvassed the divisions, brainstormed, and ultimately, tested more than ninety different positionings. As a consequence, the model suggested six equally viable strategies. The question was no longer what the best strategy was, but which of six apparently attractive strategies we should pick. Each strategy was expressed in terms of a combination of user needs.

Other surprises emerged from using SMM. For example, by comparing what customers said they would buy with what the model predicted they should buy (based on their individual needs and perceptions), we discovered that roughly a quarter of the market should be buying PC servers, not mid-range systems or mainframes. Furthermore, many people who said they would buy mainframes should not have, according to the model. In this way, the model quantitatively forecast the two most important recent developments in the industry: the decline of the mainframe and the ascendance of PC server networks.

We were also surprised by the importance of awareness. While analyzing how different competitors were perceived, we noticed an apparent relationship between how many people were familiar with a competitor’s product and how highly the product was regarded by those who were familiar with it. SMM allowed us to vary both awareness and perceptions. “Awareness” is the measurement of what percentage of the market feels it has sufficient knowledge of a company’s product to have an opinion of it; “perception” is the measurement of that percentage’s opinion of the company. For example, suppose only 60 percent of the market was aware of the HP 9000 product line. We could test what would happen if we changed how the product was perceived and if we increased awareness. In other words, we could assess the trade-off of investing in marketing communications to increase awareness versus investing in improving the products themselves in order to improve perception levels.

Many years ago, researchers explained how market share leads to various economic advantages of scale.2 However, another consequence of market share is awareness, and if levels of awareness affect levels of perception, this would suggest that increasing awareness is a force multiplier; i.e., if awareness can be increased by 10 percent, this might lead to 10 percent better market coverage and to better perception overall. If this is the case, then HP’s products would have to be substantially better than competitors’ for the overall market to perceive them as equivalent, due to their advantage in awareness (see Figure 4 for the possible relationship between awareness and perceptions).

Yet another surprise had to do with price sensitivity. The model suggested that neither product line was as price sensitive as we had thought, and that one was twice as price sensitive as the other. By constructing a financial model, we were able to estimate the effect of different strategies on the profitability of the combined businesses.

The Project Results

The most immediate impact of the SMM project was closure on the question of whether to combine the product lines. The model proved to everyone’s satisfaction the benefit of two product lines in meeting diverse marketplace needs. In positioning, both businesses now share a common view of the market, which allows them to decide where to cooperate and where to compete. Both businesses jointly attacked the mainframe-based availability/performance segment with a mainframe downsizing program. Elsewhere, the HP 9000 business pushed its products’ value, while the HP 3000 pushed its products’ ease of use.

The businesses addressed the PC server challenge in various ways. The HP 9000 division created a new family of products focused on the PC server market and Novell Netware (the leading PC networking operating system). Meanwhile, the HP 3000 business, in keeping with its ease-of-use emphasis, introduced diskless PCs bundled with its servers.

HP had already fully appreciated its awareness problem, but SMM provided additional evidence of its importance. The company responded with new advertising and public relations programs; consequently, HP is sometimes referred to as “the IBM alternative.”

SMM became an infrastructure investment. Once in place, the groups involved in its design continued to use it for their individual purposes. But the tool was also available to the more than thirty other groups that form HP’s computer systems and support businesses.

What HP Learned

SMM is a learning tool not unlike an archaeologist’s pick and shovel. At HP, we discovered answers to questions we had not asked and were left with a rich set of questions and deeper insight.

The results from our first implementation of SMM provide a sound indication of its ultimate potential. Even though the company had considerable experience in choice modeling, combining multiple models on a problem of this abstraction and strategic importance presented some important lessons:

  1. Address all the decision-making issues before proceeding. Key players have to accept the process first. Establish mechanisms (steering committees, joint working teams, off-site meetings, and so on) that will sufficiently involve all players.
  1. Use this tool only for really big problems. In the realm of strategic planning, SMM is more akin to an atom smasher than to a general-purpose tool. A strategic market model is time consuming, rigorous, and expensive, but once in place, it is very powerful. It is most appropriate for high-value situations that are daunting in their complexity but where the potential returns easily justify the effort. SMM may be a good indicator of the tools that must become commonplace if decision making is to keep pace with the accelerating competitive intensity and sophistication that HP and other companies are encountering in their businesses. Five years ago, a solid understanding of customer needs in the aggregate sense was sufficient for success. Today’s environment demands analysis of microclimates within the marketplace, where overall success or failure may hinge on appreciation of subtle nuances in customer needs and perceptions. SMM provides the precision to handle this complexity.
  2. Carefully evaluate the market research vendor’s capabilities. We chose a vendor for its preeminence in choice modeling and reputation for quality in survey design and execution. But we did not consider the adequacy of its software; as a result, the analysis was too difficult, and we lost precious time, energy, and enthusiasm.
  3. Manage expectations. Explorers and pioneers through the ages have experienced the painful consequences of overselling a vision. A person enthralled with a new tool’s potential may overstate or oversell it, and followers may generate their own unrealistic expectations. In either case, the result is the same: lost credibility and a sense of failure. The best defense is to explicitly articulate up front what a tool such as SMM will not do.
  4. Leave room for detours and side trips. While the value of planning the analysis is very important, so is the capability to probe anomalies and investigate surprises as they emerge. Although there are time and budget pressures, there is also a need for unscheduled exploration. Some of the most valuable revelations came from independent analyses late in the process, such as mainframes’ vulnerability and PC servers’ potential, the price insensitivity of the product lines, and the relationship between awareness and perceptions, to name a few. None of these discoveries would have been made without a few discretionary resources and a lot of curiosity.
  5. Needs segmentation and strategic market modeling are powerful, but they are not enough. When we embarked on this project, we expected to identify segments based on demographic questions, such as industry, application, size of company, and so on — “who the customer is.” We moved beyond this level of understanding through needs segmentation and modeling to define customer segments by “what the customer needs.” However, this level of understanding begs a deeper level of understanding and segmentation: “why the customer needs it.” In other words, why do some customers in one segment want performance and availability, while others want transparent access? Do all of the customers who want performance and availability want it for the same reason?

Knowing customers’ needs is not enough. To have what we call a truly imaginative understanding of user needs, we must know customers so well that we fully comprehend both their spoken and unspoken needs — now and in the future. We need to know what new products, features, and services will surprise and delight them. We need to understand their world so well that we can bring new technology to problems that customers may not yet truly realize they have. Our ultimate goal is this deeper, richer level of understanding.

Topics

References

1. H. Mintzberg, The Rise and Fall of Strategic Planning (New York: Free Press, 1994), pp. 324–325.

2. See, for example:

S. Schoeffler, R.D. Buzzell, and D.F. Heany, “Impact of Strategic Planning on Profit Performance,” Harvard Business Review, volume 52, March–April 1974, pp. 137–145;

R.D. Buzzell, B.T. Gale, and R.G.M. Sultan, “Market Share — A Key to Profitability,” Harvard Business Review, volume 53, January–February 1975, pp. 97–106;

W.J. Abernathy and K. Wayne, “Limits of the Learning Curve,” Harvard Business Review, volume 52, September–October 1974, pp. 109–119; and

P. Ghemawat, “Building Strategy on the Experience Curve,” Harvard Business Review, volume 63, March–April 1985, pp. 143–149.

Reprint #:

3737

More Like This

Add a comment

You must to post a comment.

First time here? Sign up for a free account: Comment on articles and get access to many more articles.