Divide to Conquer
Based on data from its call center, a warranty company thought its market was predominantly female. However, when that company commissioned marketing research, it found that its customer base was actually largely male. While the family members responsible for maintaining the warranty were typically female, those who made the decision to purchase the warranty in the first place were typically male. This information led the company to abandon plans for a major marketing campaign targeting women—which, it now understands, was not warranted and would not have been effective.
How was that possible? Segmentation! Segmentation is the process of dividing a customer base into distinct groups that share common characteristics or behaviors. It relies on advanced analytics to identify patterns within consumer data, ranging from psychographic and demographic attributes to behavioral and transactional indicators. Using statistical modeling and clustering algorithms, segmentation sorts through this data, moving beyond basic categorization to unveil nuanced consumer personas that drive strategic decision-making. This enables businesses to tailor their marketing, product development, and engagement more effectively to specific consumer groups.
For organizations, a segmentation project offers several unique advantages for targeting existing consumers. More thoroughly understanding consumers means identifying ongoing pain points. For example, if a segment prefers to use a mobile app to engage with the company rather than calling, then offering a mobile app can increase brand loyalty, make customers feel valued, and drive higher Net Promoter Scores or word-of-mouth advertising.
Segmentation also sharpens your approach to acquiring new consumers. Understanding what messaging might convert consumers to your organization is only one part of getting new business; the whole picture also involves understanding the channels through which potential consumers might convert. For example, if your customers tend to convert online, you may put more money into your web presence and reduce costly brick-and-mortar stores. Matching your consumers’ preferences and expectations to the right channel proportions can significantly enhance engagement, reduce friction in the purchasing journey, and increase overall ROI by increasing revenues and decreasing costs.
Not the Same
At its most basic level, segmentation relies on one fact: your consumers are not the same. Consumers within a market can vary in terms of their attitudes, behaviors, and demographics, but also differ in what messages might be most important and how you might reach them. Different groups of consumers within a given space or segment may also be more or less profitable than others for your organization.
Defining different consumer segments within a given space based on shared characteristics is one of the ways to keep your organization relevant and moving forward. The same things that make consumer segments unique will determine your marketing strategy. If your company provides a large portion of its products or services to college students, taking out ads in a print newspaper is likely frivolous. Likewise, if your highest-paying consumer segment prefers to listen to the radio, ads on a music streaming service are probably not the best way to reach them. To understand how to reach, convince, and ultimately gain consumers, you need to understand what they value, why they purchase, who they are, when they can be convinced, and where to reach them.
The information you want will guide the types of questions that are asked, which will then determine the insights you are able to glean from these consumers. Psychographic questions should explore motivations, values, and preferences without leading respondents. Behavioral segmentation benefits from self-reported habits and purchasing patterns, which you should cross-validate whenever possible. Demographic questions should be specific to allow for meaningful comparisons. Geographic and firmographic data should account for regional nuances in consumer behavior and business needs. Regardless of the data being collected, it is imperative to collect data accurately and ethically.
Ensuring data quality requires you to take a strategic approach to question design and survey length. Avoid questions that apply only to small subsets of consumers, as this can lead to gaps in data and limit the ability to analyze broader market trends. Instead, use follow-up questions selectively, reserving them for areas where deeper insights will meaningfully impact segmentation. It is tempting to ask every possible question, but overloading respondents with too many questions can result in lower-quality data and reduced engagement.
Moreover, questions with large amounts of missing data can pose difficulties with data wrangling, cleaning, and analysis in later stages. Prioritize questions that directly contribute to defining distinct segments rather than collecting excessive, non-essential information, especially from only some of your respondents. By aligning questions with specific segmentation goals, businesses can ensure their data remains actionable and their understanding of consumers stays sharp.
Tip #1: Use validated scales where applicable and consider segmenting by actual behavior rather than just intent, as consumers’ purchase decisions may differ from their survey responses. Additionally, ensure data sample diversity and representativeness to avoid skewed insights that disproportionately reflect a subset of the market.
Tackle the Nuances
Now that you have identified the questions you want to ask and the questions you want your segmentation to be built on, you must ensure that you are prepared to tackle the nuances of the data. Demographic data is some of the most sought-after and may help draw out consumer insights, but it can also be difficult to work with. That is because demographic data comes in all forms (e.g., continuous, nominal, ordinal) that may need to be recoded or reverse-coded and can have large variances when compared to another. Each of these unique, inherent nuances may cause segmentation models to be incorrect, inaccurate, or worse, invalid. Although each requires a tailored approach, having a plan in place ahead of time will mitigate any severe data-related problems.
Once your data is collected and cleaned, you are ready to begin the technical aspects of the segmentation process. The first step is to construct a number of factors that will help differentiate and identify your consumer segments, which are integral to segmenting your consumers. This process involves trial and error, as well as flexibility about what data to include. There are multiple ways to construct and identify these factors, commonly using a modeling approach called exploratory factor analysis. This approach can help identify what questions or data might be removed from or added to the model, the relative importance of each question or data point, what the underlying factor structure should be, whether factors are orthogonal, and, ultimately, what latent concept each factor represents.
For example, if you added 20 questions into your exploratory factor analysis and the analysis identified four different factors, the next step should evaluate how questions are grouped together, such as if there are questions that do not go well with the others, and confirm that each factor explains a unique amount of variance, separate from the others.
Perhaps you noticed that one of your factors is comprised of three questions asking about how the consumer prefers to communicate with your organization, and one question asking about the degree to which the consumer wants to be included in decisions. You could consider this factor one of consumer “involvement.” This process continues until all your key factors are sorted and identified.
Cluster analysis is the recommended way to put your consumers in larger groups. Based on a set of algorithms, consumers will be grouped into segments that are similar in scores on your key factors. Although there are different cluster analyses, one question is always asked: What is the correct number of clusters? Unfortunately, there is no one answer for all segmentations. The correct number for you is based on the overlap, or lack thereof, between the clusters. Although cluster analysis is an objective assessment of the data, this also requires some subjective evaluation on the part of the researcher. If your cluster solution has too many segments, you may have too much overlap between segments.
Identifying the perfect amount of overlap, along with the ideal number of consumer segments, requires time, effort, and experience. The best way to get started is by using a software approach to get you in the ballpark. From there, it is up to you, the data, and lots of time. To compare different cluster numbers or cluster solutions (e.g., to compare a six-segment solution to a seven-segment solution), you can begin by assessing differences in scores on the key factors you constructed. If there are still several competing solutions, it may be worthwhile to bring other data points—such as other psychographic measures, behaviors, or even demographic questions or other questions that may have almost made the cut—into the factor analysis.
As you are evaluating a potential solution, it may become apparent that there is too little or too much overlap occurring and that solution can be disregarded. This process repeats and continues until the best solution among the competing solutions is identified and you have your consumer segments. Lastly, get the software to implement that solution and bucket your respondents into their respective segments.
Tip #2: Be aware of highly niche or polarized consumer segments! For example, if your organization provides a service, you are probably familiar with a group of consumers who can be either extremely pleased with your service or extremely displeased with it. These segments can be highly differentiated from your other segments, although not always in a good way.
Lasting Impacts
We highly recommend involving key stakeholders from across the organization to ensure that every part of the process is a success. However, this is an especially important step at which to include other stakeholders, as the decisions made here will have meaningful and lasting impacts on the entire organization and often require data or insights from each department anyway.
With your key factors defined and segments determined, you are ready to generate actionable insights. Identify your segments’ scores on key differentiating factors, and then start assessing each segment’s scores on every other data point that you collected. Looking at each data point broken down by segment may lead to unexpected or unanticipated insights. For example, you may have expected that the different segments of your customer base view financial aspects differently, but now you can sort through the data for less obvious information. Does one segment insure their household electronics at much higher rates than others? Does one segment switch providers constantly? Does one segment tend to join your customer base from one specific company or organization? As you filter through your data, you may find trends that were not exposed while determining your key factors or your segments.
In constructing your segment profiles, keep in mind that information does not exist in a vacuum. Consider understanding unique segment differences alongside broader, contextual information. For example, if there is a segment that: a) spends more on your product’s or service’s category in general, b) has the highest income, c) and has fewer financial responsibilities in their lives, that segment is likely of higher priority to your organization than a segment that is unaware of your product or service, has the lowest income and has more financial responsibilities. Using a simpler, more straightforward approach, such as relative indexing, or a more sophisticated approach, such as regression-based modeling, you can rank-order your segments based on their scores on whatever data points your organization prioritizes most.
Always consider segment overlap and preference together! Perhaps one of your less-attractive segments also happens to highly overlap with your most-attractive segment in that they spend the most hours watching a particular television show. Although you may not be attempting to directly target the less-attractive segment, you could increase your marketing spend in this instance, due to the marketing phenomenon called the spillover effect. By targeting the most attractive segment, you will also inherently convert consumers of the other segment—the spillover effect.
Tip #3: Visualize your information! These segment profiles contain the exact information foundational to why you conducted your segmentation – make sure to engage others and display information in a variety of ways.
Efficient and Effective
Using the information surrounding your newly created segments, you can tailor messaging, creative strategies, and media placement to ensure that the right message reaches the right consumer at the right time. Combined with such strategies as digital ad targeting, personalized e-mail campaigns, or strategic media buys, segmentation provides the foundation for more efficient and effective marketing efforts. Your brand can focus resources on the consumers most likely to engage, convert, and remain loyal—maximizing ROI and strengthening market positioning.
Segmentation should not be seen as the end of the entire process but instead as the beginning of a whole new one. In addition to giving you invaluable information about the market and the consumers in your category, segmentation can also provide information on what new product or service lines could be introduced and how they may integrate with your organization’s current portfolio. This opens the door for other marketing exercises, such as A/B testing or concept testing, to include ways in which you may best reach your consumers about this new product or service line. Through small-scale pilot testing, A/B testing can highlight which would be the preferred service offering. All of this is an extension of your initial segmentation efforts.
Lay the Groundwork
Segmentation is valuable beyond informing marketing and advertising strategies in the present. One way to ensure its continued usefulness is to create a tool that can help determine segments in future research. A classification tool, also called a typing tool, is an instrument that can determine a consumer’s segment based on a few key questions or elements. It allows future research to focus on gathering other insights while still being able to determine which segment consumers belong to. Rather than running the whole segmentation again, you can send a shorter questionnaire to more respondents to classify them and assess your strategic marketing and advertising goals.
There are multiple methods for constructing the typing tool or classification tool; your specific circumstances dictate which method would be easiest, most accurate, and the most logical to use. No matter what you use, the goals should be: a) not re-segmenting respondents but using the new model to extend the old model to predict the classifications of the new data, b) optimizing the trade-off between the number of questionnaire items and accuracy of classification, and c) creating a repeatable and verifiable model.
Thanks to the evolution and refinement of large language models (LLMs), you may even create typing tools in web-based interfaces. For example, it is possible to use LLMs to integrate a typing tool into a survey platform, such that consumers are segmented, based on previous data, before they even complete the questionnaire. However, it is imperative to have your researchers check every aspect of the process, as these technologies are still in their infancy and are known to make mistakes.
Tip #4: Do you need a lot of items to create a reasonable degree of accuracy? Dichotomize your classifications! You may consider reducing your classification scheme from the overall number of consumer segments to instead be a binary 0/1 variable, where a value of 0 would indicate not belonging to a target segment and a value of 1 would. Because multinomial logistic regression is an extension of the simpler binary logistic regression, your analysis remains the same, yet model accuracy drastically increases.
Defining the Market
Another advantage of segmentation is that it permits you to size your market. Market-sizing is an estimate of the potential customer base for your product or service, which helps your organization understand its place in the competitive space, provides an honest assessment of where you are and identifies growth areas. This is a standalone process, although it is often integrated with a segmentation effort.
Markets can be sized by different metrics, such as the number of customers or revenues. Each method offers distinct insights into the market and its impact on your organization. For example, if you were interested in selling a product to male consumers over 18 in the United States, you may start your market-sizing with the most recent US Census estimates for that population (165 million in 2024). Taking it a step further, if you knew that one in every five men bought this product once in their lifetime, for an average price of $30, the market size in terms of revenues would be $990,000,000. Although this is a simplistic example, this is the basis for market-sizing.
Often, market-sizing is overlooked due to misconceptions or assumptions about the market. However, one way to convince stakeholders to pursue a segmentation project is to discuss the additional advantages, without any downsides, of a market-sizing analysis. For example, market-sizing gives your organization critical information as to how many consumers are in the total addressable market for your product or service. Presenting this information can help frame the information provided from a segmentation. When you know how many customers are in your market, you can generalize those estimates to how many customers or revenues may exist in each segment. In future segmentation efforts, this can provide specifics about how well your marketing approaches worked. In most cases, the data already being collected for segmentation, along with third-party data, can get your organization to an estimate of your market size.
Two of the most common approaches to market sizing are the top-down approach and the bottom-up approach. With the top-down approach, you have the widest part of your market at the top, such as the population of a given region. As you work through your approach, you apply filters and become more specific, such as a certain age group within that region. Continue this process until you have reached the most specific target demographic of interest. The last step in a top-down approach is determining the portion of the final number that your organization occupies.
As a hypothetical example, Company A is interested in a top-down approach for its hockey skates. Beginning at the broadest part, Company A is willing to do business in Canada and the US, which equates to 366,576,000 people. From there, we can add filters and increase specificity. Sources estimate that 11.6 million people are engaged in ice skating sports, with nearly 2.5 million playing ice hockey. If we extrapolate that percentage from the US to Canada, while increasing it for Canada’s generally longer, colder winters and higher levels of winter sports participation, we get 3.09 million people between the countries. Lastly, hockey requires more gear and has a higher likelihood of individuals owning their gear versus renting. Applying a conservative filter of 95%, with 5% renting, our final market size is 2.94 million people. If Company A sold 300,000 units, we can estimate that it would own 10.2% of the market.
The bottom-up approach is nearly the opposite: you start with your organization’s revenues or customer base and generalize it “up” a step or two. Commonly, the last step in a bottom-up approach is calculating the number of people or the number of revenues in your target demographic. Although this approach relies heavily on survey and internal data, there are fewer assumptions to be made, and the assumptions to be made have a higher degree of accuracy. Keep in mind that generalizing too far is inaccurate and unnecessary (e.g., there is no need to calculate your way to the widest part; the US census already provides that).
Now, the bottom-up approach, although sharing some of the same steps, will be much more targeted and precise than the top-down approach. Using a combination of data collected from our survey and third-party information, Company B has determined that it owns 14% of the market for consumer electronic insurance. With a customer base of 780,000, Company B can generalize up a step to determine that the relevant market size is approximately 5,571,000. Company B can do the same for revenues by extrapolating the ratio of its share of the market to the relevant market size overall.
Each approach has its own advantages. With the top-down approach, you can use very detailed information about your market and tailor it as you see fit. If there’s an estimate for the filter you want to provide, you can simply keep adding filters until you have reached your desired target demographic. However, the more filters you add, the more you introduce what is called measurement error, or the amount of difference between the numbers you used and the exact number. The bottom-up approach may serve as the inverse. Although it relies more on your organization’s specific data, it also more strongly controls the measurement error in your approach; if your organization sells a product, it is likely you know, within a very small margin of error, how many customers you have.
Tip #5: Regardless of approach, once you have your own market size for your company, you can also compute market sizes for your competitors if data surrounding their sales or consumer base is known—a process known as competitive market sizing.
A Strategic Imperative
Consumer segmentation is far more than a data or questionnaire exercise. It is a strategic imperative that empowers your company or organization to understand its audience deeply, engage them meaningfully, and grow intelligently. The process demands rigor, cross-functional collaboration, and continuous refinement, but the payoff is clear: smarter decisions, stronger connections, and sustainable success. As markets evolve and new tools emerge, so too must your segmentation efforts—ensuring that your organization remains not only relevant but also resonant with the people who matter most.
Tip #6: Even outside of macro-level events that may disrupt markets and purchasing behaviors, such as a global pandemic or economic recession, segmentations wear out or become outdated due to constantly changing marketplace conditions. Make sure to refresh your segmentation every so often to keep up!
This article originally appeared in the July/August 2025 issue of Quirk’s Marketing Research Review; republished with permission.
