Customer Segmentation Needs a Behaviorally Informed Upgrade

In the 1970s, the clothing company Laura Ashley was the pinnacle of women’s fashion. Known for their instantly recognizable and classic feminine floral-print aesthetic, the retail giant had more than 220 stores worldwide, including 94 in North America. They were undisputed leaders in the market and a firm favorite of upper and upper-middle class woman, with a fan base that included Lady Diana Spencer. Then, more women started going to work.

Laura Ashley’s target customers were affluent, married women over 30 years old. However, as more married women entered the labor force, their needs changed. They wanted clothing that was stylish and functional for work. But the company failed to adapt. It continued to target the same group of women with the same clothing options. By 1999, Laura Ashley sold all of its United States retail shops in a management buyout for $1.

Organizations including businesses, not-for-profits, and governments use segmentation to identify their target market and develop products, programs, and policies to fit that market. Companies often segment their customers using demographics like age, income, or gender. The act of segmentation makes sense. It helps companies focus on designing the product and developing a marketing strategy for whom they believe will benefit most. Marketing Laura Ashley’s products to a 62-year-old male retiree is not a good business strategy. Targeting middle-aged married women is—until their needs change and the company fails to adapt.

Our science and capabilities have evolved dramatically over the past few years, but the mechanics of segmentation haven’t kept up.

So how can companies understand their customer segments and avoid a fate like Laura Ashley? Segmentation has traditionally relied on the belief that people are different from each other, they can only belong to one segment, and that they rarely change segments. Many organizations still lean heavily on this philosophy. But advances in our understanding of human behavior and the technology at our disposal mean segmentation is ready for an update. We can improve how companies segment their customers, in three key ways.

  1. Move from people are different from one another to people are different from one another and themselves over time and across situations.
  2. Move from segmenting based on correlates or predictors of behavior to segmenting based on observed and actual behavior.
  3. Move from static, one-time categorizations to real-time, continuous adjustment where customers could belong to different groups at different points in time.

The figure below illustrates these three improvements to segmentation—by usage, by behavior, and dynamically, respectively.

Segmentation by usage

People are different from each other. We agree! However, research suggests that people might also be different from themselves over time and across situations. In other words, human behavior is time and context specific. Consider the following examples.

Motivation to accomplish tasks waxes and wanes with the passage of time. People are mostly motivated to take on virtuous tasks like eating healthy or starting a new assignment at salient points of time, which is known as the fresh-start effect. This could be the beginning of a new year, a birthday, anniversary, or other critical life events. The same individual might react differently to a request to join a fitness program if they’re asked in January versus in April.

Research on the hot-cold empathy gap suggests that people’s behaviors, attitudes, emotions, and judgments are dramatically different when they are in a hot state (for example, emotional or aroused) than when they are in a cold state. These differences can result in impulsive behaviors on the one hand, and detached systematic thinking on the other hand by the very same person, depending on the state they’re in.

Likewise, research shows that decision-making online is very different from offline. The same customer would order extra cheese and extra toppings when they place orders online at a pizza place versus ordering in person. And spending decisions by the same person could be different as a function of how they pay—by cash or card.

Given that human behavior is time and context specific, it doesn’t make sense to tether people to a particular segment that is time and context independent.

Segmentation by behavior

Traditional segmentation relies on correlates of behavior. For example, a marketer might infer that every millennial will exhibit similar purchasing behavior when shopping for furniture simply because they are in the same demographic group; or that all residents of a particular city behave differently from all residents of another city. With advances in data science and the fact that a lot of behavior can now be observed online, companies can use actual behavior to understand their customers. Online retailers like Amazon and search engines like Google already use actual behaviors rather than cues—the ads you see, the products that are recommended to you, and, in some cases, even the prices that you get are all a function of these companies’ observations of your behavior. What does all this mean for the practice of segmentation? It means the unit of analysis should shift from an individual to specific instances of an individual’s behavior (i.e., an individual-situation combination).

For example, an online florist could segment based on whether web visitors were looking for a gift or flowers for a home, and not on age, income, or geography. Gift purchasers care about the aesthetics of the arrangement, the speed and timing of delivery, and are relatively less price sensitive than purchasers of flowers for home who care more about the longevity of the flowers. A bank website could merge its customer data with browsing behavior on a given day to customize the layout of the web page. A government website could provide different messages based on whether constituents are accessing the message on a computer or a phone. And a charitable donations website could offer different messages and layouts as a function of when and from where people come to their website. In this sense, segmentation is based on actual behavior and is dynamic. If their behaviors change, the web layout that this customer sees today might differ from the one they see tomorrow.

And, rather than a rigid view of segmentation in which the same segmentation scheme is applied across the organization, there could be different segmentation schemes for where you are in the process—intervention design, communication design, and last-mile distribution.

We have the elements to begin approaching segmentation as something that’s ongoing and dynamic, rather than set once and static.

Dynamic segmentation

Given that human behavior is time and context specific, it doesn’t make sense to tether people to a particular segment that is time and context independent. By understanding that people are different from each other and themselves and using actual behavior as an indicator, we have the elements to begin approaching segmentation as something that’s ongoing and dynamic, rather than set once and static.

Had Laura Ashley had this knowledge and the ability to collect data over time at their disposal, the company could have observed that fewer women would enter its shops during working hours and accordingly produced more practical clothing to suit customers’ needs. Alternatively, its pricing and displays might have looked different on weekends (when working women visited stores) than weekdays. While this was logistically and economically difficult given the world of the 1970s, the technological and behavioral tools in the 2020s make it easier for companies like Laura Ashley to adapt to their customer’s changing tastes and preferences by adopting a dynamic approach to segmentation.

What’s next

Our science and capabilities have evolved dramatically over the past few years, but the mechanics of segmentation haven’t kept up. Behavioral science tells us that while people are different from each other, they are also different from themselves. And our data science now allows us to observe behavior and immediately customize marketing interventions rather than rely on surrogates of behavior. They conspire to give us a dynamic approach to segmentation, which might be the difference between understanding your customers or failing to do so.

Disclosure: Dilip Soman is a member of the Behavioral Economics in Action at Rotman (BEAR) Research Center, which provided financial support to Behavioral Scientist as a 2021 organizational partner. Organizational partners do not play a role in the editorial decisions of the magazine.