The courage to change
Editor’s note: Kathryn Korostoff is an independent marketing research consultant based in Boston.
Market segmentation benefits are pretty obvious. Most marketing professionals appreciate that it is better to divide your target market into logical groups rather than try to conquer it as a generic whole. Products, messages, channels and pricing can be optimized to each group’s needs through segmentation. Rather than advertise products broadly to a generic customer group, seasoned marketers seek to focus their marketing budget on a) the most attractive groups with b) messages/products/channels/pricing that are most likely to optimize success.
OK, so that part is easy.
Here’s the not-so-easy part. Creating an actionable segmentation scheme that is truly embraced throughout a company is often a costly, time-consuming task - one that has a high risk of failing. However, with careful attention to six of the most common hazards, I have seen clients create market segmentation models that do work, are embraced, and yield improved marketing ROI.
Hazard 1: Unwavering dedication to a single approach
In the world of segmentation, we often talk about specific segmentation approaches. Most commonly:
Needs-based segments are created by grouping members based on similar needs and attitudes. I tend to see this approach with larger companies seeking a model that will inform long-term strategic planning in addition to shorter-term sales/marketing improvements.
Product-based segments are created based on use of or likelihood to purchase a specific product. I see this approach with a lot of technology companies - and for good reason. Very often, they simply want to divide a broad market into subgroups based on likelihood to buy their products. They may end up with a five-group model - of which two groups are attractive but need to be targeted differently.
Attitudinal-based segments are based on members having similar attitudes. These attitudes can be quite diverse. For example, maybe it is a combination of attitudes about television viewing, teenage fashion and automobile brand preferences. Or as a B2B case, a combination of attitudes about employee productivity, time-to-market and regulatory compliance.
Firmographics/vertical-based (for B2B) segments are a very practical, a priori approach.
Demographics-based (B2C) segments are, again, a very practical approach. For some clients, it does make sense to simply divide a broad market into groups based on combinations of household income and, say, number of family members.
Even from these descriptions we can see that different approaches have their pros and cons. So why not combine them?
In one recent case, a client used a single market segmentation study of 600 target customers to create two models. First, a needs-based model that identified naturally-occurring groups of needs - this really paints an actionable landscape relevant to emerging customer needs and likely combinations of needs. Second, the client used the same data set to create a firmographics-based model that could be used for direct mail and other highly-precise applications. For example, direct mail campaigns were customized to each of the segments and sent to lists selected using the precise firmographics-based profiling criteria.
Hazard 2: Failing to get executive-level participation from day one
You’ve spent a lot of money and time creating a gorgeous segmentation scheme that will clearly improve marketing ROI. But did all of the relevant leaders have an opportunity to offer their ideas before the questionnaire was designed? Were relevant executives asked for their hypotheses about what type of model would be most actionable for their needs? Was any creative brainstorming done such that existing knowledge about market realities could inform the questionnaire design? If the answer to any of these questions is no, I guarantee you that when the time comes to roll out the model, you will encounter resistance if not absolute roadblocks.
Getting executives in a room is never easy. Still, in my experience, a group discussion to share existing knowledge and generate hypotheses is best. But one-on-one interactions are certainly better than nothing. Getting early-stage participation is an obvious, but often poorly-executed, phase to any segmentation study.
Hazard 3: Overreliance on statistics
Statistics are great. And powerful. And useful. But we can’t blindly allow one analytic approach to dictate “the” best model. Nor should we limit ourselves to what the math tells us. If we do, we risk creating a model that is academically interesting but impractical.
One problem I have seen is when a model emerges that looks great, but only from a math point of view. Survey respondents are grouped into buckets based on similar behaviors and needs and distributed well in terms of size. But is it actionable? Very often, especially in B2B studies, a model results in profiling information that is not distinct. That is, you have segments, but you have no way to find them. In terms of needs, the segments are distinct. But in terms of firmographics, they are muddy. This is a classic problem and is a great example of why you can’t let the statistics dictate. It’s imperative that the team conducting the analysis understand what is required to make the model actionable for real-world marketing applications.
In other cases, some market researchers become accustomed to a given tool and it becomes a crutch for them. I once hired a reasonably seasoned project manager. On his very first project, he came to me with a segmentation result that he put forth as the best model for our client. However, I could quickly see that the model didn’t jibe with my experience segmenting similar markets. Upon questioning, I found out he had used only one technique and didn’t think there was any value in looking at the data set in different ways. He didn’t last long.
If you do consider engaging a market research firm to conduct a segmentation study, make sure they are well-versed in various techniques. An agency that relies solely on cluster analysis, for example, simply isn’t going to uncover a deep enough view into recurring themes in the data. The market segmentation statistical tool chest should include multiple options, including latent class analysis, CHAID, and, perhaps, more traditional cluster analyses.
Hazard 4: Creating an overly complex model
I once had a client come to me for a refresh on an existing model. The existing model had 23 segments. Twenty-three! Who could remember them all, let alone evaluate or use that many? When my team completed the refresh, we had eight segments - of varying degrees of attractiveness for the client.
Now, can there be cases where a 23-group model works? Maybe, but in my experience, not many.
Segmentation models can be overly complex in other ways, too. I have seen models that required insanely difficult combinations of variables to assign people to segments. I have also seen models that can’t be communicated easily due to their many nuances.
Personally, I am a big fan of simplicity. If you have a segmentation model that cannot be clearly explained and easily retained, it probably isn’t going to get used.
Hazard 5: Stumbling when developing a global model
Planning for global studies requires a whole set of planning considerations and questionnaire design sensitivities. For example: Are we looking for a single model that works well globally, or does each region want a highly-optimized, region-specific model? Will the segmentation model be used for different purposes around the world? For example, in Latin America is it more for strategic planning, while the focus in the rest of the world is on direct mail efforts?
Hazard 6: Fear of change
From the research agency perspective, the most frustrating situation is an all-too-common one: A well-planned, comprehensive methodology is executed - and yields a fresh, actionable model. The final presentation goes well, the model is well-received by the client’s management team. And then, a few weeks go by and the client gets cold feet.
Yes, new segmentation models can be scary. Clients in both the B2C and B2B worlds have historically entrenched ways of looking at their markets. Changing that is a huge task. Sometimes we end up spending as much time helping the client apply the model as we did in developing it in the first pace!
Let me share a recent example. Working with a technology company that sells to consumers, we created a needs-based segmentation model. Management was behind the effort from day one. The model we delivered had seven segments, which were different in terms of attractiveness, fit, messaging opportunities and channel preferences. Sounds good, right? Well, it was. But the client’s product development team was under so much pressure to react to competitors’ moves that it was unable to take swift action. The segmentation model gave them the potential for a huge competitive advantage - insight into how they could meet the needs of the most attractive customer segments - but the firm couldn’t fully capitalize.
One exercise I did with the client to help them tweak their brains was as follows. At a re-re-presentation of the model, I had the audience focus on one of the attractive segments. And then I said, “If I gave you $100 million to start a company aimed at this segment, what would you do?” This helped them take off the common filters of “But what about our current customers?” and “But our sales people say we need to respond to competitor X.” Look, of course those things are considerations, but they often get in the way of leapfrog potential. And after all, we can’t leapfrog the competition if we are following them.
Fantastic tool
Market segmentation is a fantastic tool for optimizing marketing success. But it takes time to do it right. It takes money to create the model. It takes planning to avoid common pitfalls. And it takes a good dose of courage to actually apply it. But for companies unafraid of change, it can provide the fuel needed to not just jump ahead of, but to soar above, the competition.