Market segmentation criteria

A company decides to launch a product. Basic question: who is going to buy it? The answer "the people who need it" is useless operationally. Saying "women aged 25 to 34 in big cities" is something, but it's still a group of millions of people with completely different lives, incomes, motivations and behaviors.

Between "everyone" and "one specific person" there's an intermediate task called market segmentation: dividing the audience into groups coherent enough that the same message, the same offer and the same channel work for the whole group. And choosing the right criteria for making that division is the difference between segmentation that changes business decisions and segmentation that just decorates a PowerPoint deck.

This article gathers the main criteria —where they come from, what they solve, where they fail— and proposes a way to choose them based on which decision you want to change.

Where the idea comes from

Segmentation as a formal marketing concept was born in 1956. Wendell R. Smith published an article in the Journal of Marketing that changed the discipline: "Product Differentiation and Market Segmentation as Alternative Marketing Strategies." Smith argued that offering a single homogeneous product to a mass market was ceasing to work; it was more profitable to identify subgroups with different preferences and serve them tailored offerings. That idea —obvious today, novel then— founded the field.

Throughout the 1960s, Philip Kotler systematized the approach in his editions of Marketing Management (first edition 1967) and formalized the STP framework —Segmentation, Targeting, Positioning— which remains the backbone of any serious marketing course. Theodore Levitt, in Marketing Myopia (Harvard Business Review, 1960), added the conceptual complement: companies that define themselves by their product rather than by the needs they solve fall behind when those needs are met some other way.

Since then, the criteria have multiplied. VALS (Values, Attitudes and Lifestyles), developed by Arnold Mitchell at SRI International in 1978, popularized psychographic segmentation. RFM (Recency, Frequency, Monetary), formalized by Arthur Hughes in Strategic Database Marketing (1995), took segmentation into the realm of transactional databases. Jobs-to-be-Done, articulated by Clayton Christensen and Bob Moesta in the 2000s building on Tony Ulwick's earlier work, proposed a shift of axis: segmenting not by who the person is but by what "job" they're trying to do when they "hire" a product.

The consequence of that history is that today there are many criteria available. The question is not "what criteria exist" but "which ones serve this business decision."

The six classic criteria

It's worth starting with the six that every textbook covers, with their use and their limit.

Demographic

Age, gender, income, education level, occupation, household size, life-cycle stage (single, couple, with young children, with adult children, retired).

When it works. Products where these variables genuinely predict the purchase. Baby diapers, products for the elderly, premium wine, home insurance. When age or income correlate strongly with the buying decision.

When it fails. Products where demographics don't explain behavior. Two 35-year-olds with the same salary can have radically different lives and buy opposite products. Demographics as the sole criterion produce segments too heterogeneous to be actionable.

Geographic

Country, region, city, postal code, urban/rural, climate.

When it works. Products with local or regional dependence. Food, seasonal clothing, proximity services, retail, local-language content. Brands expanding into a new market.

When it fails. Global digital products. Knowing that two customers are in different countries matters less than knowing what they do with the product. Geography is often a good initial filter but a poor sole segment.

Psychographic

Values, attitudes, lifestyle, interests, personality. The territory of VALS and of most elaborate "buyer personas."

When it works. Aspirational brands, products bought for identity affinity. Patagonia and environmental awareness; Apple and creative identity; Harley-Davidson and the freedom of the open road. When brand choice is an expression of identity.

When it fails. When it's built without data. The "personas" invented in a workshop without research —"María, 32, into yoga and authentic travel"— are fictions that don't predict behavior. For psychographics to work, you need real qualitative research (interviews, ethnography) or quantitative tools to back it up.

Behavioral

Frequency of use, timing of use, loyalty, price sensitivity, benefits sought, occasions (gift vs. personal use, travel vs. home).

When it works. Products where behavior is a reliable predictor. Subscriptions (daily user vs. occasional), retail (high-ticket customers vs. low, promo-sensitive vs. not), banking, telecom. Almost any digital product with telemetry.

When it fails. When behavioral data doesn't exist yet —product launches, new markets. It's hard to segment by behavior when there's no installed base.

By need / benefit

What the product solves for this person. One person buys an electric car for sustainability, another for savings, another for status, another for technology. Same product, four segments.

When it works. Almost always. It's the criterion most connected to the advertising message, because different needs require different promises.

When it fails. If it's confused with poorly defined "use cases." "The need to get around" is not a segmentation; it's the definition of the entire market.

By commercial value / RFM

How much the customer is worth to the business. RFM classifies each customer by Recency (when they last bought), Frequency (how often they buy) and Monetary (how much they spend). From there come segments like "champions" (buy often, recently, a lot), "at risk" (valuable customers who haven't returned in a while), "hibernating" (bought little, a long time ago).

When it works. Any business with a purchase history: e-commerce, retail, subscriptions, banking. It's probably the criterion that best predicts the ROI of retention and reactivation campaigns.

When it fails. For pure acquisition, it doesn't apply —prospects have no history. It's a customer criterion, not a market one.

Criteria for B2B

In business-to-business selling, the criteria change. The demographics of the individual buyer are less relevant and firmographics appear: descriptive variables of the client company.

• Industry (manufacturing, professional services, retail, healthcare, technology).
• Size (number of employees, revenue).
• Digital maturity (what tools they already use, what level of adoption they have).
• Company stage (startup, scale-up, established).
• Geography and markets served.
• Buyer's role and function (CEO, CMO, operations director, head of procurement).
• Buying committee (who decides, who influences, who vetoes).
• Available budget and approval cycle.
• Installed technology (which CRM, which CMS, which stack — useful because it reveals compatibility or need).

Jobs-to-be-Done segmentation also works very well for B2B: two CFOs with the same demographics can have radically different "jobs" —one needs to close the month faster, another needs to explain results to the board better. Same role, different segments.

Jobs-to-be-Done: the shift of axis

Traditional segmentation classifies people: age, income, values, behavior. Christensen and Moesta proposed classifying circumstances. The question is not "who is this customer?" but "what are they trying to do when they decide to buy?"

The canonical example is the one about McDonald's milkshakes, studied by Christensen. Traditionally, McDonald's segmented milkshake buyers by demographics (ages, gender). A different investigation found that 40% of milkshakes were sold first thing in the morning to people driving to work: they wanted something that would last the trip, keep one hand busy, not make a mess, and be slow to "eat." The milkshake's "job" was to keep them entertained on a long commute. The competition wasn't other desserts but bananas, doughnuts and snack bars. Same person, different "job" at another time, different product hired.

Applying JTBD requires specific research —interviews that reconstruct real purchase episodes— but it produces segmentations that pure demographics can't see.

How to choose: the actionable-decision rule

There's a meta-criterion that helps choose among the above: a segment is useful if separating it changes a marketing decision. If two segments are going to receive the same message, in the same channel, with the same offer, they're not segments —they're the same audience described with different words.

Concretely, a segment deserves to exist if it justifies at least one of:

• A different message. The promise or the angle changes for this group.
• A different channel. This group is reached through a different medium.
• A different offer or price. This group pays a different price or accesses a different option.
• A different product or experience. This group receives variations of the product.
• A different page or journey. This group enters the funnel by its own route.

If none of the five changes, the "segment" is ornamental.

The other practical criterion is measurable and operable. A brilliant but invisible segment in the available data can't be activated. If you can't tell who belongs to the segment by looking at your CRM, web analytics, paid-media tools or research, it's not an actionable segment.

Mistakes every team repeats

Too many segments. Producing 12 buyer personas is not producing good segmentation. It's producing 12 fictions that nobody uses afterward. Three to five segments are usually the operational limit: what a team can keep in mind and serve with different assets.

Personas without data. Inventing segments in a workshop with sticky notes and no follow-up research. Invented personas tend to resemble the team that invented them: if the team is urban, professional and of a certain profile, the personas are too —and real customers may have nothing to do with them.

Confusing description with segmentation. "Women aged 30 to 45, urban, interested in wellness" describes 50 million people in Europe. Actionable segmentation needs variables that narrow and distinguish, not ones that merely describe in general.

Segments without size. A perfectly tuned segment that's 0.3% of the market doesn't justify dedicated marketing effort. Before investing, estimate the segment's size and economic value.

Ignoring the channel. A "perfect" segment you can't reach through any available channel is theoretical. Useful segments have at least one accessible and profitable channel.

Treating segments as immutable castes. People move between segments. An "at risk" customer can become a "champion" with the right campaign; a "champion" can fall to "hibernating" if neglected. Segmentation is a snapshot of the moment, not a fixed destination.

Demographics as the only lens. It's the easiest criterion to obtain and the least predictive of behavior. It should almost never be the sole criterion; it can be an initial filter.

Confusing market with customer. Segmenting the total available market —to enter or not enter— is different from segmenting the existing customer base —to retain, cross-sell, recover. RFM answers the second; demographics and needs answer the first better.

How it fits into the workflow

Segmentation criteria, badly managed, live in a deck nobody revisits. Well managed, they shape the whole operation: the editorial calendar, the SEO architecture, the paid campaigns, the landing pages, the tracking.

Creative operations is what keeps a system of segments from staying theoretical. At Polimake, Studio translates segments into message and content architecture (which pages, which keywords, which assets for each segment); Studio coordinates execution and measures which segments respond best; Media produces the creative variants needed so the same concept adapts without turning into fifteen disconnected campaigns.

This relates to market segmentation as a general concept, to the market research that feeds the segments with real data, and to the positioning decision that comes next.

To close

Choosing segmentation criteria means deciding which differences between customers deserve to translate into different messages, offers, channels or experiences. The frequent trap —producing many criteria "to complete the list"— ends in segmentations that are pretty and operationally useless.

The practice that ages best: start with the marketing decision you want to change, choose the few criteria that actually change that decision, validate that they can be measured and reached, and review quarterly whether the segments still make sense. What serves one company stage may stop serving when the product, the channel or the market changes.

Quick references

• Demographics: fast and cheap, but rarely explains behavior on its own.
• Geography: useful initial filter, weak sole segment.
• Psychographics: powerful with research; fiction without it.
• Behavior: predicts better when there's data; useful for retention.
• Need / benefit: directly connected to the message.
• RFM: the gold standard for existing customers.
• Firmographics: the basis of B2B.
• Jobs-to-be-Done: when the circumstance matters more than the person.
• The rule: if it doesn't change message, channel, offer, product or route, it's not a segment.
• Three to five segments are usually the operational limit.