Segmentation vs. Personas – What’s The Difference?

Summary: Segmentation and Personas are part of the segmentation analysis process. Personas creation is an essential step in segmentation analysis, but it is not a technique to create the actual segments. Personas are simply segments' profiles.

9 minutes to read. By author Michaela Mora on November 29, 2021
Topics: Analysis Techniques, Market Segmentation, Quantitative Research, User Experience, UX Research

Segmentation and Personas

Segmentation and Personas are part of the segmentation analysis process. They are not two different “research techniques” as many in the UX community want us to believe.

Segmentation includes procedures to identify the segments. Personas are just descriptions, profiles of those segments.

Market Segmentation Misconceptions

In market research, we often use “market segmentation” to talk about any segmentation analysis, but this can be confusing to UX researchers. The term “market” gets immediately associated with sales and marketing, leading many in the UX community to believe that market segmentation can’t be used in product development.

Nothing further from the truth. Market segmentation is often the first step in new product development in companies with mature customer research operations. Product development should be connected to the marketing strategy in any company, without products, there is nothing to market.

There are many misconceptions about market segmentation or segmentation in general among UX practitioners stemming from a lack of knowledge or bad experiences with this research approach.

Wrongfully, many think that market segmentation:

  • Involves only quantitative research
  • Is based only on demographic information
  • Fails to classify users accurately because it doesn’t consider their behaviors and needs
  • Can only be used in marketing and sales, not product development

Let me set the record straight. The most actionable market segmentation used variables relevant to the users’ needs, behaviors, and attitudes connected to the product or service category.

Market Segmentation for New Product Development

There is no one set of variables that will always discriminate between segments for every product category. Segmentation variables need to be specific to the goals of the segmentation and the desired business outcomes.

In new product development, we often need to consider the broad population of current and potential users in terms of:

  • User needs met or unmet by the product and competing alternatives
  • User behaviors triggered to meet needs along the customer journey
  • Channels of user interactions and touchpoints along the customer journey
  • Usage occasions
  • User roles activated by usage occasions and needs
  • Barriers and motivators to usage (financial, level of effort, impact on user’s environment, habits/inertia, expectations, etc.)
  • Demographics, when correlated with relevant behaviors connected to life stage (e.g., gender and age are sometimes associated with family status, usage of specific product categories, etc.).
  • Attitudes towards the product category, specific products, or brands (this could be motivators or barriers)
  • Users’ core values (family, politics, environment, money, religion, social status, material consumption, etc.)  that don’t change much over time and guide decision-making and behaviors within the product category
  • Situational factors that impact decision making (offers, competitor actions, unexpected situations triggering needs, etc.).

All of these variable categories can be used as segmentation variables. Very rarely, only one would work. We usually use a combination of several to find relevant and actionable segments. Users are humans like us, and we are complex creatures.

The Right Approach to Market Segmentation

Segmentations usually fall into one of two categories:

A Priori Segmentation: The variables used to segment users are already known and predetermined (e.g., customer status, age groups, user roles, company size, etc.). This approach is better suited when there is evidence that these variables separate users into meaningful segments to product development and get closer to the desired business outcomes. If the variable selection is based on best guesses or anecdotal data, this type of segmentation may become useless at best and misleading at worst.

Empirical Segmentation: The variables that would discriminate between segments are unknown. They could be any from the list mentioned above. This type of segmentation is exploratory by nature. We need to gather data from users on several dimensions and find patterns that indicate  different groups exists. If you have no data and just gut instinct, or sparse data, or qualitative data about your user segments, this is the best approach to uncover the segments and validate any hypothesis you may have about what they are.

How to Conduct Empirical Market Segmentation

Qualitative Research is a Must

Ideally, any empirical market segmentation should start with a qualitative research phase since we don’t know what differentiates the segments. If prior secondary research exists, assuming it is not too outdated, we can use it in place of qualitative research to formulate hypotheses about the segments.

If no other research exists, we design qualitative research to explore users’ needs, attitudes, and behaviors that may unveil motivators and barriers to using the product, specific features, and the circumstances that help or hinder the customer journey.

In this qualitative research phase, we go broad and deep and look for overall patterns along dimensions that become salient as we observe what users do and listen to their feedback.

The qualitative research should yield hypotheses about potentially discriminating variables. Some segment ideas may start forming, but it is best to leave them alone. We may still be wrong about the segments, as we could be paying attention to qualitative input that confirms our biases. Qualitative data is rich, messy, and hard to handle without quickly succumbing to our biases when we unknowingly impose patterns in trying to make sense of the messiness.

Focus on the variables that could reveal the segments and treat these variables as hypotheses that need validation.

Why Not Stop at Qualitative Research?

  1. Qualitative research samples are small and not representative of the user population. We don’t know which segment we may have excluded from the sample selection.
  2. Even if the segment ideas suggested by the qualitative research are on point, we don’t know the size of the segment and whether it is worth investing time and resources to develop products and features for any particular segment.

Quantitative Research Is Required

It is not uncommon to get a long list of potential segmentation variables out of qualitative research or other prior secondary research that may be available. Not all of these variables will have a strong discriminatory power.

When tested in a larger sample, some variables may lead to “universal truths,” showing many people behave similarly or share common attitudes and preferences on a particular dimension. Other variables will show more significant variation in a large sample, suggesting that distinct segments may exist.  

If budget and timing allow it, we do a “pilot” segmentation to eliminate the variables that don’t discriminate and then run a second study with the rest of the variables and new ones (discovered during the analysis of pilot data) that may separate the groups. However, this can get expensive and time-consuming. In practice, we often do all of this in one step using multivariate statistical techniques such as cluster analysis, convergent cluster and ensemble analysis, latent class segmentation, factor analysis, and discriminant analysis.

Segmentation techniques yield probabilistic models, which means that each user has a probability of belonging to each segment. We compare their probabilities and assign them to the segment in which they have the highest probability score. The segment classification accuracy depends on our ability to include relevant variables that discriminate, which is why the initial qualitative research phase is so important.

No segmentation solution is 100% accurate. There is always a level of error in probabilistic models. The question is whether the level of accuracy helps minimize the risk of bad decisions in product development. In practice, segmentation models with a 70% or higher accuracy level provide solid guidance in risk mitigation.

How Often Should You Update a Segmentation?

The shelf-life of a segmentation study varies by product category and how dynamic the competitive landscape is. Segmentation studies have very short shelf-lives in product categories with low entry barriers, allowing new competitors to change the landscape, or in constantly evolving categories due to new technologies, cultural trends, or significant events (e.g., COVID pandemic).

The longevity of the segmentation will also depend on the variables used to identify the segments. Specific needs, attitudes, and behaviors may change when faced with different options offered by competitors (e.g., free shipping for Amazon’s Prime members) and circumstances (e.g., digital transformation driven by the COVID pandemic).

Segmentation solutions based on more stable variables (e.g., core values) may last longer, but they will eventually need to be revised as well. Even changes in unrelated categories that impact socioeconomic and cultural trends will ultimately reach most product categories (physical or digital). Expecting a segmentation solution to be valid forever is unrealistic, and frankly, a lazy assumption. Change is constant.

How are Personas Different from Segmentation?

An essential step in segmentation analysis is to create profiles of the segments. We pay attention to variables that discriminate between the segments and search for patterns that tell the story of who the average user in that segment is. We use those differentiating criteria to create names for the segments to easily talk about them and create empathy among internal stakeholders. How rich or sparse that profile is, depends on the data available and the team’s creativity in presenting the segmentation results.

With the increased use of storytelling and graphic design in research reports and presentations to help the C-suite, marketing, and product teams digest segmentation insights, segment profiling has morphed into a Personas design exercise in many organizations. Segments are often described as people who do this or that, are of a certain age and gender, show this or that behavior, are in a specific user role, or face a particular problem, etc., depending on the variables included in the study. They sometimes even get a personal name.

There is nothing wrong with that. Personas are just another name for segment profiles describing a user type based on relevant criteria from the segmentation research (needs, behaviors, attitudes, demographics, etc.). Personas are a deliverable from segmentation analysis, not a research technique in itself.

 

Segmentation and Personas Definitions

The Problem with Personas

Given the rich profiles that can be derived from segmentation studies, Personas may look like a qualitative research exercise to the non-researcher. In companies with no market research/insights operation or one working in a silo, many sales and marketing teams and UX teams have taken over the Personas concept and decoupled it from the research needed to develop them.

In many organizations, Personas are developed in internal sales meetings or UX design sprints and workshops based on anecdotal evidence or a few user interviews. There is no consideration of whether they are worth pursuing based on their size and generalizability.

Among UX research practitioners, I often hear disdain for quantitative analysis, which comes from a lack of knowledge about quantitative methods.

As of 2021, UX research is mainly conducted using qualitative research methods, not because it is the best methodology for every research question, but because that’s what most UX researchers know. Researchers who only know one way will try to use it for everything, even when inappropriate.

Unfortunately, this practice of creating Personas out of thin air or basing them only on qualitative research hurts their perceived value among internal stakeholders. When Personas have no foundation in data, they quickly become obsolete and misleading. They are hard to defend and use to guide product development decisions.

In Conclusion

Personas are simply segmentation profiles based on a set of differentiating criteria. These criteria and any other variables used to add color to the segments to make them easier to communicate should stem from the segmentation analysis.

The goal of persona development should be to provide vivid pictures of actionable and meaningful segments to guide product development decisions that support desired business outcomes. A mixed methodology approach based on qualitative and quantitative segmentation research is the best way to develop valuable Personas.

To conduct a segmentation analysis and develop personas that give your business a competitive edge either in product development and/or marketing to target the right segments, contact us to learn more about how we can help with our segmentation research services.