Hitting a golf ball down the middle of the fairway may be good in golf, but not so good in marketing. As with B2C, B2B marketers see value in understanding their customers and potential customers through segmentation. We need to ask whether there are different groups of customers with different needs or other characteristics that require different go-to-market strategies and possibly different products or services. Segmenting businesses, however, presents some complications that require careful thinking to navigate successfully.
Suppose we survey 1000 people in 1000 companies. We then segment those companies and develop more targeted marking strategies for our product. Our segmentation is driven by wants and needs relevant to our product and we use state-of-the-art methods (let’s say, latent class clustering). Suppose we also have too much time and money on our hands, and we immediately repeat the study with a different 1000 respondents.
- Will we arrive at the same segmentation?
- Will the same companies end up in the same segments?
- In either study, can we trust that the respondents accurately represent their companies? In other words, did we in fact segment companies or merely individuals?
Doing the same study twice as described is simply not done in business research, so conclusive answers to the above questions are elusive. Let’s consider, however, some issues surrounding B2B segmentation and develop tentative answers and recommendations.
Companies differ in visible ways…
…and a segmentation should not ignore those differences. We recently ran segmented data that included very small and very large companies to determine their needs for various financial services. To no one’s surprise, the enterprises tended not to cluster with the 10-employee businesses—the small and large companies were so different that they arguably comprised different markets altogether. Yet, one or two very large companies did end up with the small companies according to the clustering algorithm, and were conspicuous by comparison. In the final results, we grouped the large companies despite what the segmentation told us (and had plenty of other statistical evidence to support that decision).
Company size is an easy and obvious factor for segmentation schemes. In general, we recommend segmenting among otherwise similar companies (with similarity defined by whatever factors are known and relevant in a given market) rather than across dissimilar companies. The result will be cleaner segmentation and does not blur the boundaries between companies known to be different. Segments will be easier to personify, more actionable and more stable.
In addition, acknowledging the differences among companies allows for different surveys and different segmentation drivers tailored to different businesses. Even research design can be different—perhaps we would survey multiple respondents in large companies or have different screening criteria for who is a qualified respondent. Most importantly, we may see that some categories should not be treated with a standard segmentation study. If, for example, your company has 10 Fortune 100 clients, we would suggest in-depth case studies for each client rather than a segmentation.
Too many segmentation drivers is a bad thing
In B2B research, we often research small universes such as very large companies, companies who buy or sell specialized products, etc. Conventional wisdom is that with a small universe, a small sample is less of a problem because even a small sample can be representative. While true, that wisdom breaks down in segmentation studies. I once taught a class in a 600 seat lecture hall. During a snowstorm, the university elected not to cancel classes but instead to ask students to make their best effort to attend. Needless to say, my class had perhaps 50 people that day spread out over the entire space. One small clump of students sat in front, the rest were scattered throughout. In other words, at most one segment was represented, plus a number of loners.
Now consider a study with only 12 segmentation drivers where each driver is a dichotomous variable. A respondent could answer those 12 driver questions in 4096 different ways. If we survey 100 respondents (a reasonable sample size for a B2B study with a small universe), those 100 can be spread far more thinly than my 50 students across 600 seats. Yes, they could cluster such that four or five groups give similar patterns of response, but they may also spread themselves out. The 100 respondents could have 100 different patterns of response over 12 survey items such that no real clusters exist, or that there are dozens of very small clusters.
Stable, actionable, and repeatable segmentations of businesses should focus on a smaller, more tightly focused set of drivers. As much as possible, those drivers should address characteristics of businesses rather than attitudes of individuals in businesses. Unlike consumers and consumer segmentation, sampling a respondent in business and then developing an attitudinal segmentation confronts some intricacies.
Businesses are not people
A successful segmentation will drive strategy—whether marketing, sales, product development, or something else. The paradox for business segmentation is that individuals -- not businesses -- react to campaigns, sales pitches, products. However, unless the respondent is the only person in the business, there could well be others who would answer differently and lead to different conclusions for how to characterize that particular company. When planning a B2B segmentation, then, we need to answer very clearly whether our goal is to understand people or organizations.
For organizations, we can design a study around business needs and characteristics, and respondents are merely our best source for that information. For people, our segmentation can also include individual attitudes and beliefs. The key is that mixing the two (business features and individual characteristics) will result in a vague segmentation within B2B segmentation. The segments are less likely to be stable, less likely to have clear personas, and more difficult to act on.
Conclusion
We began by wondering how surveys of individuals can result in a segmentation of businesses. We believe that B2B segmentations can be improved and can be more accurate and usable, by careful planning and knowing the audience. Three points are critical:
- Acknowledge that a useful segmentation will be a mix of a priori segments (separating organizations into groups by size and other key features) and clustering techniques applied within those groups.
- Choose segmentation drivers carefully. Too many will result in the appearance of a segmentation but may not represent any meaningful clustering of respondents.
- Understand the questions that the segmentation research is supposed to answer—will you act on the segmentation at an individual level or an organizational level?
If you would like to discuss how Hansa can help your organization with an appropriate segmentation strategy for your company, please contact us.c
