Blog

The segmentation graveyard is full of sad creatures. Most large organizations – and many small ones – have buried such a graveyard in previous research and strategic plans. Among its disused residents are programs that:

• Lack of explanatory power for observed behavior
• Are incomprehensible because of the more than a dozen variables that were used to define them
• Include large "unexplained" buckets for groups outside of statistically inferred clusters
• Have significant overlap between segments

These pitfalls are not just the domain of the uncomplicated. We once worked with a large medical technology company whose segments were defined by a research firm, including Overwhelmed (who were dissatisfied in almost all dimensions) and Simplifiers (who were happy with everything as it is). What on earth are they going to do with it?

There is a better way. As I researched at Branding Strategy Insider, there are several steps that must be followed to reliably create actionable segmentation. However, we are also often asked which features define such a segmentation. This piece provides that answer through five simple properties:

1. Segments should be neatly linked to tasks to be done. My mentor, Clayton Christensen, first defined the concept of tasks to be done, which relates to how a client might want to progress under certain circumstances. This concept, covered in more detail in my book with this title, explains why people do what they do. Ultimately, people buy a Land Rover instead of a Tahoe for a comparable price because they want to achieve something (both functionally and emotionally), not because they belong to a certain population group or relate to an archypal person. These latter factors correlate, but do not cause, actual behavior. Understanding the core motivations and the circumstances in which they arise can provide explanatory power for a segmentation.

2. Segments should predict preferences. The acid test for a segmentation survey is to work in short concept tests at the end. The segments that were defined in some of the previous questions in the survey should have different levels of interest in different concepts. If it doesn't, the segments aren't going to be of much help to product developers. While segments have other uses as well – for example, targeting salespeople based on buying processes, or managing advertising spend based on media consumption – product development is usually primarily a primary motive for creating good segmentation.

3. Segments should relate to findable data. Early in my career, I suggested elegant segmentation to a bank employee based on psychographic data. The scheme made a very intuitive sense and has been statistically validated. His recoil was straightforward: "How do I find out the different psychologies of all people who are standing in line in a branch in a row?" I didn't really have an answer and that experience stuck with me. If you can't define customers based on discoverable data, you may have wonderful segmentation for product development, but it does little for your go-to-market efforts.

4. Segments should be understandable. I recently interviewed a mobile game developer who created 60 segments that were used to determine how players can be monetized based on their different in-game behavior. That was sensible; The person lived in a data-rich environment where only they needed to understand the hyper-complex scheme. Very few of us develop video games like this. We have limited leverage that we can pull to benefit from segmentation and we need to show larger organizations how to use these frameworks. If segments are too numerous to be easily remembered, or if they are defined by more than 4-5 variables, the red flags should wave high. A segmentation is only as good as it is actually used. Make it understandable.

5. Segments should focus on the addressable market. A staggering number of segmentations are meant to explain the entire market just to define just a few segments in the domain that can really be addressed by the company. Just as a scheme with 60 segments is generally of no use, so is one with only one pair. Marketers need to find the right balance and focus on the addressable market first. If 80% of the market isn't interested in your Organic Cheez Whiz, you don't really need to segment it heavily. Focus on where it matters.

These are simple features, but they are all too rare. Why? In part because companies refuse to do the groundwork to discover jobs that need to be rigorously done. It's actually not that much work, but it requires stepping down from prejudice and a detailed framework so that blurred insights are not interpreted through the everyday mental lenses. Another major illness is that people are both confused and enchanted by statistics that are often used by providers who use fancy terms (e.g. standard deviation) without considering the key details (e.g. standard deviation is only then a valid measure if the answers follow a normal bell curve, which is very often not the case). Insist that vendors focus not on unclear statistical metrics but on their everyday meaning and how that relates to predicting what segmentation really needs to predict.

Actionable segmentations are definitely within reach. Don't bury another expensive project. Rather, you can use these five simple tests to check early on whether a segmentation will be successful.

For more information on this approach, see my book JOBS TO BE DONE: A Roadmap for Customer-centric Innovation.

The Blake Project can help you create a bolder competitive future with a workshop to do

Brand Strategy Insider is a service from The Blake Project: A strategic brand consultancy specializing in brand research, brand strategy, brand growth and branding

Free publications and resources for marketers