Like most sciences, segmentation keeps asking more questions

I love science. In my early years I planned to become an astronomer. That was in 1960s and 70s, so I blame John F. Kennedy and his ‘man on the moon’ promise for building a hope an aspiration that proved well beyond my mathematical capabilities.

Fortunately, I was not totally mathematically challenged, so have been able to participate and enjoy the almost universal application of numeracy in our digital world.

One thing that is great about science is that it never ceases to ask the next question. When you think you’ve got it all wrapped up and are close to answering the big questions about the meaning of life and where we came from, science will always throw another curve ball. That is why I always find myself pondering the science and methodology of segmentation and, in particular, why I’m confronted by different carve ups of our community.

In the old days, we were happy with demographic segmentation. It was reliable. People were a certain gender, age, lived in specific locations, earned a known income and so on. You could populate the spreadsheet then slice and dice your audience in any way you wanted using rock solid data sets.

Now things are not so clear. We have the far less certain psychographic, behavioural and other overlays. Personally, I think these provide far more powerful perspectives on the way people will react to communications and other stimuli. Getting the insights right and, more importantly, configured into a commercially useful tool, is a long term and painstaking process.

Different research or branding companies will give you a community view based on a new and proprietary segmentation of the population. To illustrate the point without mentioning any names, I have over the past couple of years encountered the following:

  • Research company - 3 segments based on financial literacy and approach to money (behavioural);

  • Financial services business - 23 segments based on consultant’s 'secret sauce recipe (behavioural, attitudinal?);

  • Research company - 60-plus based on a complex matrix of data (psychographic by location);

  • Brand agency - 4 (psychographic) which I could have derived myself with about 30 minutes to think about it.

The reason for this is commercially available models comprise many varieties of segmentation - from demographic to psychographic and everything in between. Some models are combinations of these. The challenge is selecting what is going to best fit your purpose. How do you sort through this morass of what superficially appear to be conflicted analyses of the people who comprise your current and prospective customer base?

Let’s assume now that there is one off-the-shelf ‘master’ view of the population to which we’re ready to subscribe. The question then becomes how do you overlay this model onto your own customer base?

We’ll make up a fictitious segment that a major research company might invent to describe a segment - ‘Gregarious Urbanite’ (believe me, that’s the sort of dross you’ll occasionally encounter on this journey). The descriptor might say things like:

‘Fun loving millennial. Tech-savvy. Paying rent or repaying a mortgage (insightful heh? doesn’t everyone do this?). Often lives in inner city. Frequents restaurants and consumes news from multiple sources on different devices.’

How informative would this be? In a customer based of 100,000, you could probably ascribe this to thousands. Faced with the same question or purchase decision as another covered by that segment, would the buying decision be the same, or the motive or circumstance of it? Very unlikely.

You need to add some more meat to the bone, some attitudinal or other data that can enrich your view of the propensity of these two people to purchase your product or service.

So where do you get that? The answer is in your customer data, your own source of truth. Before you look at enrichment using third party data, this is where your segmentation work should always commence. There is no richer seam of gold to mine than data on how customers have engaged with your brand, products and services in the past. It is the finest indicator of how they and others similar to them are likely to do so in future.

A deep understanding of your existing customers, their interactions with your business, their behaviours and preferences, enables you to identify exactly the type of data that will enrich your customer insights. In other words, it becomes one of the filters you can apply to the range of enrichment options available to you.

A cynical person could say that the lack of a universal view of the segments that make up our communities is due to the need for consultancies to evolve proprietary views of the world that they can market to their own prospects. While there is possibly a modicum of truth to this, the reality is that the bewildering array of segmentation models is largely due to the multiple ways and purposes for which they are created.

Your challenge is to know your brand strategy and customers well enough to select the one that is going to enable you to align your messages with what resonates with your audiences.

Here’s some articles that you could take a look at for different perspectives on segmentation:

Smart Insights: How to use Segmentation, Targeting and Positioning (STP) to develop marketing strategies, by Annmarie Hanlon. 31 October 2018.

Science Alert, Market Segmentation Models to Obtain Different Kinds of Customer Loyalty, Mario Montinaro and Ivan Sciascia. (Very technical data science stuff with a retention focus and lots of links to other academic articles)

Openview, Customer Segmentation: A step-by-step guide for B2B, by Tien Anh Nguyen, 3 July 2018. (Another technical article, but the opening sections provide some good pointers to getting started)