Rules of effective segmentations, and why we believe automation can fall flat!

Author profile picture

Simon Harris

13 Oct, 2022 | 4 minutes

When maximizing reach and driving mental and physical availability is considered key to sales success, it can be tempting to think that most product innovation should be designed for mass appeal. The problem is, with markets increasingly saturated and new products required to displace entrenched competition, this approach often leads to watered-down propositions – favoring the lowest common denominator and rather average products!

If the goal is to maximize trial and repeat, then our experience indicates that you’re better off targeting a well-defined set of consumers. In short, it’s nearly always better to uncover and understand the different preferences that exist amongst your target population, and then make informed choices on who to target, with a view to creating something that delights them, rather than works fine for the masses.

And whilst we’ve always specialized in applying this focused, segmentation-led approach to product development, the principle applies across the development and execution of the entire proposition!

“In a world of digital media, knowing the right message for a specific group of consumers can be hugely powerful; with agile content creation and pinpoint targeting driving higher engagement, purchase and improved ROI.”

Guy Grimsley, Director Huxly, reflecting on his days as an Insights Director

That said, we concur that segmentations only add value if they have been designed and implemented correctly. Which is why we believe automated and DIY solutions can be a false economy, for the reasons summarized below:

  • If you throw data at any segmentation tool it will always find a solution, but that doesn’t mean the strongest statistical solution is the best for your business, far from it! How do you know that the solution is really the best route forward?
  • There are multiple segmentation techniques, each prioritizing data in different ways. Different methods suit different situations and data sets, so how do you know you’re on the best path?
  • Some segmentation techniques require the commissioner to make choices, such as the desired number of segments, which all impact the final solution. How do you know if you have made the best choices? What other solutions could have emerged, and would they have been better?
  • Inputs ultimately determine segmentation results, this should necessitate an interrogation and refinement of the data. How do you know if you have selected the best inputs?
  • For a segmentation to land and thrive in an organization, employees need to be taken on the journey, to understand how the final solution has been achieved so that they might have confidence in it. Do you understand how the solution was derived, how it is better than other possible solutions to confidently bring others with you?

To this end, having a skilled analyst involved throughout the process, from research design through to analysis and interpretation, is essential. We called on such experts from around our business to identify our rules for an effective segmentation.

Know what you’re trying to achieve

You need a clear goal from the outset. What are you trying to achieve? Remember to be flexible and adapt to whatever the data reveals.

Having a clear end goal is key not only in terms of selecting the right variables to input into the model but also, more fundamentally, what questions and scales to use at the data collection phase, get this wrong and any segmentation analysis could fall flat - Rubbish in, rubbish out!

Selecting the right variables at the outset can be tricky, as you might not know what really matters until you see the results (that’s one of the beauties of a good segmentation, unlocking previously unknown insight!). But the design process can benefit hugely from some preliminary work to determine the key dynamics of the area you are investigating, be it primary qual, some desk research or a review of previous studies.

Simon POV 2 Image 2

“Some of the best segmentations I have done have been based on a prior piece of qual research to understand the key topics and the consumer language.”

Simon Harris, MMR Product Testing Excellence Director

Knowing what attributes you need to assess and how consumers might react to them is also key in determining the types of scales you use. If flat data is likely to be an issue than trade off or choice based techniques can be used to force consumers to prioritize attributes. If consumers need to react to a lot of stimulus than using scales that allow enough points for consumers to discriminate and provide a hierarchy is key.

Even with high levels of thought and planning you might still fall short of 100% accuracy, but as long as you have allowed for consumers to discriminate (if they want to) and captured everything that might matter, our statisticians can, and will, take it from here. Through the segmentation process, our analysts explore which inputs within the data are differentiating and which aren’t and make recommendations on how to refine the inputs to create more meaningful and actionable outputs. Pivotal to this is understanding how the segmentation techniques work, and importantly what the techniques prioritizes when making classifications.

Simon POV 2 Image 3

Whilst we are talking about design, a few other things to think about at the set up stage:

  • Think about how you’d like to target the people in the emerging segments so that you can ensure you collect the right information upfront, or sufficient information to allow for follow up deep dives later on.
  • Think about what the business will be comfortable with as this can influence what data the segmentation needs to be based on (behavioral, attitudinal, needs based, taste based etc).
Simon POV 2 Image 4

Interrogate solutions and explore alternatives

This is where art meets the science, and it’s why we believe you should be wary of automated alternatives. As we said earlier, if you throw data at any segmentation tool it will always find a solution, but that doesn’t mean it’s the right one.

In our experience, it is not unusual for us to run numerous different solutions before we land on the one that’s fit for purpose. There is no substitute for careful, considered analysis. Trust us when we say that one size most definitely does not fit all (which is the whole point of a segmentation after all!!). To this end, having a skilled analyst involved throughout the process, from research design through to analysis and interpretation, is worth its weight in gold.

As well as evaluating different solutions, our analysts will often explore multiple segmentation techniques on any given project. Be aware that different segmentation techniques do sometimes throw out wildly different solutions – which is a red flag for us and not something you would realize when running only one solution.

Be aware that it isn’t uncommon for the same method to throw out different solutions in different software packages. This is because segmentation techniques rely on making decisions, decisions which can be programmed but if you don’t know these decisions are being made then you may not realize how it is impacting your resulting solution or what other direction the solution could have gone in if a different decision was made.

“A client came to us asking us to replicate a segmentation solution they had created, to provide reassurance and collateral for stakeholder sell in. We used the same segmentation tool and parameters but used a different software package and it produced a different result. The issue was caused by how the software dealt with ties which occur in hierarchical clustering techniques when there is more two or more options for how the next cluster is formed. The result was 2 different solutions, if the client hadn’t been made aware of this they may have blindly proceeded with their solution not knowing other equally viable solutions existed.”

James Godfrey, MMR Head of Statistics and Analysis

In other cases, segmentation techniques can throw out different results on the same dataset as they require the user to define some parameters upfront, such as the desired number of segments, selecting that you want 3 instead of 4 doesn’t just mean you get one less segment, the 3 segments you get might not bear any resemblance to any of the segments you would have got in the 4 segment solution. Having an analyst on hand to explore and compare all the options means that you don’t miss out on essential learnings simply because you typed 3 instead of 4!

And on this note, it is crucial that you think about what the desired complexity of the solution is. Sometimes the best value comes from segmentations that are relatively simple – after all, too many segments can sometimes result in a lack of actionability. Now is the time to work through your options and the implications.

Take your people on a journey

Ideally, segmentations should thrive within organizations for many years. However, for that to happen everyone needs to have confidence in how the solution was derived and that it really is the best solution. Getting to this point requires getting stakeholders involved early and taking them on the research journey. Our teams are experienced at involving many stakeholders in this way, often presenting several solutions to the clients and then interrogating these further based on stakeholder feedback before making a recommendation on how to proceed.

How results are presented back to the business can have a huge impact on whether a segmentation lands or slides into oblivion. Sometimes a little theatre is required, which is something the MMR family of companies have plenty of experience in – roping in our creative colleagues at design consultants The Together Agency and video production specialists Ideal Insight, as well as our brand consultants at Huxly. The final result means that the segments are bought to life and the findings get maximum impact.

“We’re looking for the little things that can make a big difference…..if you just throw AI/ML at this sort of thing it just finds the patterns, then everyone finds the same patterns and we get back into the averages=average circle…..again room for a bit of art here, and when we find the nuggets then really think about how we can bring them to life for the client.”

Mat Lintern, MMR Global Managing Director

It is all in the detail

Our final rule relies on the tiny detail. To be actionable, segments need to be identifiable and targetable and this can require an incredibly granular level of detail, much of which might not have been collected as part of the original study or included in the original data analysis as its importance might only become apparent until later in the process.

Our segmentation tools can be applied in flexible ways, either designed into the research from the start or retrospectively applied to a data set. We can identify respondents from segments to recontact for follow up interviews or build algorithms to allow for new consumers to be recruited and allocated to the segments for follow up studies. In this way, the segmentation can play its traditional powerful role in innovation and ATL marketing as well as fulfil the modern requirement for micro targeting and dynamic media.

It’s a mantra here at MMR, to look beyond the average consumer and what they claim is important to them. We strive to uncover what is going on under the surface. We think that we have some of the best people and tools in the business – and we have actively steered clear of automating the process because of the risks we believe it poses. As soon as human involvement is removed, segmentations can lose all strategic significance, resulting in poorly understood solutions which then don’t have the solid underpinnings to confidently impact and live on in the organization.

Simon POV 2 Image 5

If you’re considering the power of segmentations and want to discuss your situation in a non-pressured ways, please reach out and we’ll set something up.

Chat to us today