The Hard Road to Good Customer Segmentation

Anthony Tjan’s Harvard Business Review blog post, Approximately Correct Is Better than Precisely Incorrect, perfectly illustrates the value of customer segmentation.  Your customers are different, and treating them all as if they were the same results in poorly targeted offers, lower sales, and dissatisfied customers.

The trick is that customer segmentation is hard.  Otherwise, everyone would be doing it well!  The ideal segments are cohesive (everyone in a segment has similar behavior and preferences) and actionable (you can identify which segment a customer is in and respond appropriately).  There are more or less three generic flavors of customer segmentation, and all of them have drawbacks.


Demographic-based segments use information like age, gender, and geography to segment customers.  For business markets, the equivalents are industry and company size.  These segments are easy to implement because that information is accessible through existing customer information, surveys, and third-party sources.  Targeting these segments is also straightforward because media, salespeople, and other channels can easily discern which segment a consumer or business is in.

However, demographic segments often don’t correlate very well to customer needs and behavior.  In other words, they’re not very cohesive.  Perhaps middle-aged males aren’t a good segment for your company overall, but some slice of that demographic group may be your best customers.  There’s no good way to tell without using additional segmentation dimensions, and companies using demographics often end up throwing out the baby with the bathwater (or vice versa, trying to sell products to the bathwater).


As the name suggests, behavioral segmentation is based on customers’ actual behavior.  This approach excels at analyzing and predicting the behavior of existing customers, since past behavior often does help predict future behavior.  Identifying the right metrics to use for segmenting customers is a bit harder.  The true challenge, however, lies in targeting new customers.  Since the segmentation is based on existing customer behavior, it becomes quite difficult to tie it back to criteria that can be used to identify potential customers in the same segments.  So while it helps you work with current customers effectively, behavioral segmentation often falls short in customer acquisition.


Psychographic segmentation is certainly the coolest sounding of the three approaches.  It focuses on understanding the different motivators (and often lifestyles, for consumers) of customers.  By understanding these factors, companies try to achieve a deeper understanding of how to better serve different psychographic segments.  Psychographic segmentation tries to get directly at customer needs, something that demographic segmentation can only crudely approximate.

While psychographics can provide unique insights, they suffer from difficulties in targeting, similar to behavioral segmentation.  In addition, the link between psychographics and behavior is often a weak one because it’s difficult to predict how general attitudes and lifestyles will impact specific product decisions.

An Ongoing Process

So what’s the solution to this three-part dilemma?  The most important aspect is to treat segmentation as an ongoing process of rather than a destination.  Any attempt at segmentation is likely to be imperfect, so it should always be seen as a set of useful hypotheses rather than ultimate truth.  More specifically:

  • Start simple:  Don’t necessarily try for the most sophisticated segmentation in one step.  Refine it over time.
  • Gather data:  Start with basic demographics.  As your segmentation becomes more sophisticated, layer on behavioral and psychographic elements.
  • Ask questions:  Any segmentation has areas of ambiguity or even error.  Just because your statistics have a 95% confidence interval doesn’t mean that you’re not operating in the 5% occasionally.  Take a skeptical eye to your data and figure out which areas need to be confirmed.
  • Use qualitative insight to form hypotheses:  Every business has insights into how individual customers behave.  Try to expand those into hypotheses about segments.
  • Conduct experiments to test them:  Collect additional data to test your hypotheses, and make sure the data collection is designed in response to specific questions.
  • Address flaws in the segmentation model:  Make your model of customer decisions and behavior more robust based on your new data and analysis.
  • Repeat

As you can see, there’s no quick fix to the challenges of customer segmentation.  Of course, even a rough segmentation is a huge improvement over treating all your customers the same way.  By using a thoughtful approach and not treating an initial segmentation as gospel, you can continue to gain even more insights into your customers and make better decisions.

Related Posts:

  • Anonymous

    This is an extremely insightful post, and definitely worth careful consideration.  Valid customer segmentation analysis is certainly one of the fruitful avenues for research, but it is hard to do well.  The key, as you say, is to treat this type of analysis as an ongoing process.     

  • MHoopes

    My experience is that the only way to segment your customers is to ignore what they look like, where they live, or how much they make.  All that matters is what they buy and what triggered that buy (behavorial segmentation).

blog comments powered by Disqus