Types of Market Segmentation: How To Build and Choose Segments

Market segmentation splits a total addressable market into groups of buyers who share traits that change what they want and how they respond to an offer. The main types of market segmentation are demographic, geographic, psychographic and behavioral for consumer markets, with firmographic and technographic added when the buyer is a company. This guide covers the main types, how to build segments from data, how to judge which segments are worth targeting, and a worked example that follows a hotel group from booking data to positioning.

What is market segmentation?

Market segmentation is a practice in which a team divides a market into smaller groups of buyers whose needs, characteristics or behavior differ enough that they may need separate products or marketing programs. The purpose is concentration. A team cannot serve every buyer equally well, so it chooses the groups it can serve best and builds offers, messages and channels for them.

Teams use segmentation to map buyer groups, targeting to choose which groups to serve, and positioning to decide what those groups should think the brand is.

The main types of market segmentation

Kotler, Keller and Chernev’s Marketing Management (16th edition) names demographic, geographic, behavioral and psychographic as the common types of market segmentation for consumers. Business teams use firmographic bases such as industry, company size and location, plus technographic data on software in use.

Demographic segmentation

Demographic segmentation groups buyers by age, gender, income, education, family size, life stage or occupation. The variables are observable and cheap to buy media against. Teams can find them in census data and panels, supplemented by CRM fields, which is why most segmentations start here.

T-Mobile’s Unlimited 55+ plan is a demographic segment with a published rationale. When T-Mobile announced the plan in August 2017 (two lines for $60 with AutoPay), its release cited 93 million Americans over 55 and said the company held 8% of that group against 18% of all US wireless customers; the plan still appears in the company’s 2025 10-K alongside military and first-responder plans. The limit of the base shows in Toyota’s Scion. The brand marketed to 18- to 24-year-olds, but the New York Times reported a median buyer age of 37. Toyota folded Scion into its own lineup in 2016. Age told Toyota who to chase. It said nothing about why a young buyer would prefer a Scion to a Toyota.

Geographic segmentation

Geographic segmentation groups buyers by world region or country, region within a country, city or metro size, population density and climate. Density distinguishes urban from rural areas, with suburban areas between them. It matters most where assortment, pricing or local regulation changes by place.

Macy’s My Macy’s program is the documented retail case. The 2008 pilot ran in 20 districts, expanded in February 2009 to 69 districts, and gave local teams decisions over roughly 10% to 15% of store inventory. The Bellevue, Washington store added extra-small sizes and shifted its jewelry mix for local demographics, and CEO Terry Lundgren cited a sales lift of about 250 basis points over the rest of the chain. Spotify applies the same base to price: its filings say subscription pricing is adapted to each local market’s purchasing power and willingness to pay.

Psychographic segmentation

Psychographic segmentation groups buyers by values, attitudes, interests and lifestyle, so it answers why where demographics answer who. Planet Fitness positions its whole brand on an attitude toward gyms, the “Judgement Free Zone” named in its FY2025 10-K, and reports about 20.8 million members across 2,896 clubs.

Strella’s guide to psychographic segmentation covers the five variables, their data sources and the decisions they inform (positioning, messaging, pricing, packaging and campaign planning), so this article does not repeat them.

Behavioral segmentation

Behavioral segmentation groups buyers by what they do with a product. The variables are usage rate (light, medium or heavy), loyalty status (hard-core loyal through switcher), occasion, benefits sought, user status and readiness to buy. Benefit-based segments usually carry the clearest marketing implications, because the benefits people seek are often what makes segments distinct.

Progressive’s Snapshot program is behavioral segmentation inside an insurance product. Progressive’s 2011 shareholder letter called telematics “an additional level of segmentation” layered on its existing pricing, with discounts of 0% to 30%. Carrier Management reported in 2023 that the participation discount had risen to 15% and the maximum surcharge was 60%. Renewal rates ran 6% higher for the biggest-discount drivers and 16% lower for drivers with no discount or surcharge. Netflix applies behavioral segmentation to recommendations. WIRED reported that it builds more than two thousand taste groups from viewing behavior and leaves age and gender out of its recommendation system.

Privacy rules and platform changes limit how much behavioral data teams can use. Durable segments rely on data customers agreed to share, such as purchases, loyalty records and app usage, and on research they chose to take part in.

Firmographic segmentation for B2B

Firmographic segmentation groups companies by industry, size, revenue, location and growth stage. Industry usually follows NAICS, the US federal standard with 20 sectors and 1,012 industries in the 2022 manual; size is measured by employees or revenue.

ADP’s bands show the base in use:

• Bands and products: Its 2025 Investor Day defines Small Business as 1 to 49 employees and Mid-Market as 50 to 999. Up-Market covers 1,000 and over, and the products follow the bands: RUN serves small businesses, Workforce Now serves mid-sized and large companies, and Lyric HCM serves large enterprises.
• Sales channels: ADP sells to small businesses through accountants and banks, as well as point-of-sale providers; up-market channels include ERP partners and system integrators.

Technographic segmentation adds what a company runs, which Gartner defines as the software solutions currently in use; a vendor that integrates with one data warehouse or CRM segments on exactly that.

A nested approach starts with these outer variables, because the data is visible and stable, then moves inward to purchasing approach and situational factors, and finally to the people on the buying committee. That inner work is where interviews pay: ProRata synthesized 100 in-depth interviews 10x faster with Strella to refine its ideal customer profile (customer stories).

B2C vs B2B segmentation at a glance

Consumer and business segmentation use different bases but the same logic, from observable traits outward to the harder-to-see reasons behind a purchase. The table maps the four consumer bases against the two business bases:

Business buyers are still people, so psychographic and behavioral variables apply to the buying committee once the outer nests are set.

How to segment a market step by step

A segmentation is only as good as the decision it was built to serve, so the process starts with the decision rather than the data, following Bain’s guidance to meet a concrete business objective. The steps below connect that objective to data collection and analysis, with sources mapped to each base and methods matched to the decision.

1. Define the market and the decision: State what the segmentation must decide, such as which sub-brand to launch, which customers to price differently or which accounts sales should call first. Set the market boundary it covers, including prospects and not only current customers. A segmentation with no decision attached becomes a deck.
2. Collect data against each base: Pull demographic and firmographic fields from the CRM or licensed data and behavioral records from transactions and product analytics. Collect needs and attitudes through a survey; Bain’s travel-company case started with a survey of about 3,000 consumers plus travel agent interviews. Run qualitative interviews before the survey so the questions measure needs the team has heard in people’s own words. Nielsen Norman Group suggests 5 to 30 interviews for that exploratory work and a survey of at least 100 respondents, ideally 500 or more, before claiming a statistical segmentation.
3. Build the segments: Run the analysis that fits the data (see the methods below) and profile each cluster with the descriptive variables. Name segments by the need or behavior that defines them. Reforge’s guidance is to segment on 2 to 3 attributes, which keeps every segment explainable to a sales or product lead in one sentence.
4. Choose targets: Score each segment on the viability checklist and on attractiveness, then pick the one or two the company can serve better than competitors.
5. Test before scaling: Confirm the segments exist and respond differently. Bain validated its two priority travel segments through six focus groups and survey callbacks before the client built products around them. Then set a refresh date, since Bain also notes that individual customers move from one segment to another.

Data sources and tools

Each segmentation type has a home source. A CRM and a CDP supply different parts of the customer record; a survey platform supplies research responses. The CDP Institute’s 2026 definition of a CDP is “software that creates and maintains a persistent, unified customer record that is accessible to other systems”; a CRM stores sales and service interactions, mostly from its own system, for staff. The sources by type:

• Demographic: Census data and panel targeting attributes, supplemented by CRM fields.
• Geographic: Billing and shipping addresses, supplemented by store catchments and territory maps.
• Psychographic: Survey attitude scales and open-ended interviews.
• Behavioral: Transactions and loyalty records (usually in the CRM or CDP) and app and web analytics, with MarTech’s caution that a CDP labeled “real-time” may be segmenting on data that is not.
• Firmographic and technographic: CRM account records and licensed company and technology data, supplemented by what sales reps hear on calls.

Researchers use surveys and transaction data to identify which groups exist and how big they are. Through qualitative interviews, they learn why people buy and use those reasons to write a positioning statement.

The fitness chain [solidcore] shows the sequence. Its team already knew that clients arrive for different reasons: some for the community and energy, some for the structure and routine, and some to push themselves. To go deeper, it used AI-moderated interviews on Strella with prospects, active members and past clients. Participants described practical and emotional responses that had not appeared in previous research. Gillian Almeida, [solidcore]’s VP of Strategy & Insights, described the result: “Strella helps us hear from clients on an ongoing basis and share those insights with the team quickly, so we can make decisions with real client voices behind them.” The full account is in how [solidcore] uses Strella.

Teams can add the same why-focused step alongside their survey and transaction data. Participants join open-ended interviews on their own time and answer in their own words. The AI moderator follows up in the moment on what each person says. Many interviews run in parallel, and themes across all of them surface as the interviews finish, in hours, not weeks. Teams weighing the method can try an interview to hear how the moderator follows up.

Methods behind segments

Three methods do most segmentation work, and each fits a different kind of data.

• K-means cluster analysis: The analyst picks a number of segments (k), the algorithm places k centers, assigns every respondent to the nearest center, moves each center to the middle of its members and repeats until nothing changes. It suits survey needs or attitude items after each variable is standardized so no scale dominates; expect round, similar-sized clusters and sensitivity to outliers. Choosing k is a judgment call supported by fit statistics and by rerunning from different random starts to see whether the solution holds. Consumer data rarely contains natural, well-separated segments, and a cluster analysis returns clusters whether or not that structure is there, so stability checks are part of the method rather than an extra.
• RFM: Score each customer on purchase recency and frequency, then on monetary value, typically in quintiles, which yields 125 cells. Analysts need only a purchase history, so it is the fastest way to segment an existing customer base for retention and reactivation. They cannot assess prospects with RFM and use only three predictors, so teams should use it to support a needs-based segmentation rather than replace one.
• Conjoint analysis: Respondents choose among product descriptions built from attributes and levels, and the analysis estimates how much each level is worth to them (part-worths). Analysts use latent-class estimation to group respondents by the trade-offs they make, producing needs-based segments tied directly to pricing and feature decisions. Keep attributes to about six and make levels concrete; a level like “superior performance” leaves each respondent to guess what it means.

Use RFM on transactions to identify existing buyers and their spending. If the question is which needs divide the market, cluster survey data and check stability. If the question is what people will trade for what, and at which price, run conjoint and segment on the part-worths.

How segmentation fits with targeting and positioning (STP)

Segmentation is the first of the three STP steps, and nothing downstream works without it. Teams divide the market into meaningful customer groups, choose which groups to serve, then differentiate and position the offer for them.

Targeting takes the segments and weighs two things: the segment’s overall attractiveness (size, growth, profitability, scale economies and risk) and fit with the company’s objectives and resources. Teams then position the offer to occupy a clear, distinctive and desirable place in the target’s mind relative to competitors.

Verizon ran the sequence at the scale of an org chart. From April 2019 it reported Consumer and Business segments and told analysts it was moving from a technology-driven to a customer-driven structure, with business customers grouped into Global Enterprise, Small and Medium Business, Public Sector and Wholesale.

How to tell if a segment is worth targeting

A segment is worth targeting when it passes five standard tests and then scores well on attractiveness against the company’s ability to serve it. The five tests:

• Measurable: The team can estimate the segment’s size, purchasing power and characteristics from data it has or can buy.
• Accessible: The segment can be reached and served through channels the company can use at a cost that makes sense.
• Substantial: The segment is large and profitable enough to justify a separate program. No automaker builds cars for people under four feet tall, for example.
• Differentiable: The segment responds differently to a different marketing mix. If married and unmarried women respond the same way to a perfume promotion, they are one segment, not two.
• Actionable: The company can design a program that attracts and serves the segment with the resources it has.

Passing all five earns a seat in the attractiveness scoring, not a target. Malcolm McDonald’s directional policy matrix worksheet keeps that scoring simple: choose six factors or fewer, split 100 weighting points among them, score each segment 0 to 10 per factor, and treat a weighted score of 7.0 or higher as highly attractive. Growth rate and accessible segment size are usual headings; profit potential is another. He also argues attractiveness should be judged with data external to the company, estimated about three years ahead, before the team scores its own competitiveness.

One more check belongs here because teams cannot judge profitability from needs-based segments alone: McKinsey’s credit card segmentation work notes that needs-based segmentation does not divide customers into profitable and unprofitable groups, so a value lens (RFM or margin by segment) runs alongside the needs lens.

If a segment fails measurable or accessible, drop it before scoring. If it passes all five and scores 7.0 or above on weighted attractiveness, it goes on the shortlist for the company-fit comparison.

Common market segmentation mistakes

Teams struggle to use segments when they create too many or leave them unchanged as buyers’ needs shift. Other failures come from ignoring profitability or leaving no one responsible for acting on the study.

• Too many segments: A Quirk’s case describes a 23-segment model refreshed down to eight, and a seven-segment model whose product team was unable to take swift action. Teresa Torres’s rule from Product Talk is that if two segments’ needs are the same, teams remove the split from the tree.
• Stale data: Most segmentations are a snapshot. Strategy+Business describes a telecom segmentation that was out of date barely six months later because the long-distance market had shifted, and Reforge suggests a starting window of no more than a year of data.
• Ignoring segment size and profit: Quirk’s states the test plainly: a segment must be large enough to be worth investing in. Teams tailor a message to a distinct but small segment rather than build a product line for it.
• Attitudes that do not predict purchase: Yankelovich and Meer’s 2006 HBR article “Rediscovering Market Segmentation” argues psychographic profiling has become a diversion from segmentation’s purpose, which is finding customers whose behavior can be changed or whose needs are not being met. Teams must check every attitudinal segment against purchasing behavior in the category.
• No owner after the study: Quirk’s practitioners put activation first among the factors that define a segmentation’s legacy and recommend a C-suite champion from conception through implementation, and the first action once a segmentation lands is naming who owns each segment’s refresh date.

Market segmentation example from data to positioning

Picture a regional hotel group with 40 midscale properties deciding whether to launch a second brand. This scenario is called the example hotel group from here on. Every figure in it is illustrative, chosen to show the mechanics rather than to describe a real company.

The team would draw on three data sources. From about 18 months of booking records in the property system and loyalty database, it would pull stay length, day of week, booking lead time, rate paid, channel and repeat visits for hundreds of thousands of stays. It would survey a few thousand recent guests about trip purpose and the three things that mattered most in choosing the property, and the survey would include a bank of needs statements rated on a scale. To explore the reasons behind the ratings, the example hotel group would then run around 60 AI-moderated interviews with guests drawn across the stay-length spread.

The team would build segments in two passes. Using RFM on the bookings, it would find a small core of returning guests, roughly one in eight, accounting for around two-fifths of room nights. It would test k-means on the standardized needs statements at three to six clusters, rerun it from different starts and settle on four. It would then profile each cluster with booking data to define the segments:

• Weeknight regulars (just under a third of guests, about a third of room nights): These guests would make Monday to Thursday stays of one or two nights, booked inside a week, with high loyalty-program enrollment, and they would want reliable Wi-Fi, the same breakfast every time and check-in in under a minute.
• Weekend escapers (just over a third of guests, about a quarter of room nights): These guests would stay Friday and Saturday, book 3 to 6 weeks out and choose on location and rate, with little loyalty to any chain.
• Event groups (about a quarter of guests, roughly a sixth of room nights): These guests would book weddings, tournaments and conferences as room blocks, where the organizer decides, not the guest.
• Extended-stay movers (under a tenth of guests, about a quarter of room nights): These guests would stay 14 nights or more for relocations, project work or medical treatment, booked by the guest or an employer, with the highest nights per booking in the data.

In a study like this, the survey would likely show extended-stay movers rating a kitchen in the room far above other guests. The interviews would then explain why. Movers would describe a month of restaurant meals, a wish for a morning routine that feels like home, and the awkwardness of daily housekeeping when the room is where someone lives. Weekend escapers, in the same interviews, would talk almost entirely about the neighborhood and the price.

Two candidates would reach the viability scoring for the example hotel group, weekend escapers for their headcount and extended-stay movers for their room nights:

On this scoring, the example hotel group would choose extended-stay movers. The segment is smaller but underserved, and an underserved segment is the kind worth targeting. The channels would be cheaper, and the conversion would use properties the group already owns. Scored on weighted attractiveness across growth, size and profit, movers would clear the 7.0 bar for a highly attractive segment, while escapers would land around 5.

The example hotel group’s team would write its positioning statement in a common positioning format, To (target segment and need) our (brand) is (a concept) that (point of difference), filled in as follows: To people living away from home for a month or more who need a place that runs like home, the new brand is a weekly-rate extended-stay hotel that gives every guest a full kitchen and in-room laundry, with housekeeping on their own schedule, unlike corporate apartments that require a lease and hotels that charge by the night. The team’s next step would be a six-property pilot that reruns the same interview guide at 90 days to check whether the segment’s reasons still hold.

If your team needs the why behind a decision but cannot wait weeks for fieldwork and synthesis, talk to the Strella team about your research question. See Strella in action.

FAQ

What are the 4 types of market segmentation?

Demographic, geographic, psychographic and behavioral are the four, the standard set for consumer markets. A quick way to hold them in mind is who, where, why and how. Firmographic takes the place of demographic when the buyer is a company.

What are the benefits of market segmentation?

Teams use segmentation to concentrate spend and product effort on the buyers most likely to respond, gaining higher response rates and revenue per customer than they would with a uniform offer.

The best-documented numbers are for personalization, segmentation’s downstream cousin: McKinsey’s 2021 Next in Personalization research, built on a survey of 1,013 US adults, found personalization most often drives a 10% to 15% revenue lift, with company-specific results from 5% to 25%.

What are the 6 types of market segmentation?

There is no agreed list of six. Six-item lists start from the classic four and add two of firmographic, technographic, needs-based and value-based, and different publishers choose different pairs. Needs-based and value-based are better understood as ways of cutting a market than as new bases: Gartner describes needs and values as core approaches to building segments, with clusters as a third approach. It uses value segmentation to mean segmenting on lifetime value.

What is the difference between market segmentation and customer segmentation?

Market segmentation covers the whole market, including prospects and groups the company will never target, while customer segmentation usually covers only existing customers, in TechTarget’s distinction. Many textbooks and analysts use the terms interchangeably. The practical difference is the data: a market segmentation needs survey, panel or interview data that reaches non-customers. A customer segmentation can run on CRM and transaction records alone, which is why RFM is a customer segmentation method and conjoint can serve either.

How does market segmentation fit into STP?

It is the first step and the input to the other two, and a team that skips to positioning without a chosen target writes a message aimed at everyone. The loop runs more than once in practice: targets change as segments shift, and the fix is a refresh on a schedule rather than a new segmentation every few years.