Segmentation, Targeting, Positioning, and Personas

Introduction

Product managers rarely serve a completely uniform market. Customers differ in the problems they experience, the outcomes they seek, the contexts in which they buy and use products, their willingness to pay, and the value they create for the firm. Segmentation, targeting, positioning, and personas provide a disciplined way to convert these differences into product strategy.

Segmentation identifies meaningful groups within a broader market. Targeting determines which groups deserve strategic priority. Positioning defines the distinctive value a product should occupy in the minds of selected customers relative to alternatives. Personas translate research findings into vivid but evidence-based representations of important users and buyers.

These activities are not merely communication exercises. They influence product scope, research priorities, pricing, channels, service design, sales enablement, retention programmes, and the allocation of scarce organisational resources. Poor segmentation can lead to undifferentiated products, conflicting road maps, inefficient acquisition, and misleading customer narratives. Good segmentation links customer evidence to choices.

Learning Objectives

By the end of this chapter, the reader should be able to:

1. Distinguish demographic, needs-based, behavioural, value-based, contextual, and firmographic segmentation.
2. Develop segments using qualitative research, quantitative data, and analytics.
3. Evaluate segment attractiveness and strategic fit.
4. Select target segments while recognising trade-offs and cannibalisation risks.
5. Construct responsible personas from evidence rather than stereotypes.
6. Develop and interpret positioning statements and perceptual maps.
7. Recognise ethical, statistical, technological, and governance limitations in segmentation practice.

1. The Strategic Logic of STP

The STP framework follows a sequence, although in practice the process is iterative:

Segmentation → Targeting → Positioning → Product and market actions → Measurement and learning

Segmentation asks: What meaningful differences exist among customers or users? Targeting asks: Which differences matter sufficiently for us to serve? Positioning asks: What distinctive and credible value should the chosen audience associate with our product?

A segment is not simply a demographic category. A useful segment has a recognisable pattern of needs or behaviour, can be identified and reached, is sufficiently substantial or strategically important, is distinguishable from other groups, and can be served with a credible value proposition. These conditions are often summarised as measurability, substantiality, accessibility, differentiability, and actionability. They are decision criteria, not guarantees of success.

Segmentation may be performed at several levels:

• Market level: broad groups with different problems or desired outcomes.
• Account or customer level: organisations or households with different economic value and relationship characteristics.
• User level: people who interact with the product.
• Usage-occasion level: situations in which the product is used.
• Need-state level: immediate objectives, anxieties, or jobs to be done.

A buyer, economic decision-maker, user, administrator, and beneficiary may be different people. A product manager must therefore avoid treating “the customer” as a single actor.

2. Foundations of Market Segmentation

2.1 Traditional bases of segmentation

Demographic variables are often useful for description, but they should not be assumed to explain behaviour without evidence. For example, age may correlate with technology adoption in one category but be irrelevant in another. Needs-based and behavioural variables are generally more actionable for product decisions because they connect directly to problems and usage.

2.2 Jobs, needs, and outcomes

A needs-based segment groups people according to the problem they are trying to solve and the outcomes they value. The “job” may be functional, emotional, or social. A commuter may be trying to arrive reliably, reduce travel anxiety, or signal environmental commitment. One person may perform several jobs at different times, so need-state segmentation should include usage context and occasion.

Research should distinguish:

• Stated needs: what people say they want.
• Latent needs: needs that are difficult for people to articulate.
• Unmet needs: important needs that current alternatives address poorly.
• Must-have needs: conditions for adoption or continued use.
• Delighters: benefits that exceed expectations but may not be expected.

Jobs-to-be-done thinking is valuable because it shifts analysis from product categories to desired progress. However, it should not become a slogan replacing research. Jobs must be evidenced through interviews, observation, behavioural data, diary studies, support records, and experiments.

2.3 Behavioural and value-based segmentation

Behavioural segmentation uses actions rather than only self-description. Common variables include acquisition source, activation, frequency, recency, feature adoption, purchase occasion, switching, renewal, referral, and support contact.

Value-based segmentation asks whether different customers generate different economic or strategic value. A simplified customer lifetime value model for a recurring-revenue product is:

Here, m is contribution margin per period, r is the retention probability per period, and d is the discount rate per period. The formula assumes constant margin, constant retention, a steady-state relationship, and a one-period timing convention. It is a simplified perpetuity model and should not be used when margins, retention, acquisition costs, or customer behaviour vary materially over time.

For a non-recurring or finite-horizon relationship, a more general expression is:

R_t is revenue in period t, C_t is cost to serve in that period, p_t is the probability the customer remains active in that period, d is the discount rate, T is the evaluation horizon, and CAC is customer acquisition cost. All monetary values must use the same currency and time basis. The model is useful for comparing segments, but predictions should be tested against realised cohorts.

Value should not be reduced to current revenue. A small current account may have high strategic value because it provides learning, network effects, credibility, or future expansion. Conversely, a high-revenue account may impose substantial service costs or create concentration risk.

3. Developing Segments from Evidence

3.1 Begin with a decision, not a dataset

Segmentation should be designed around a decision such as:

• Which users should receive a redesigned onboarding journey?
• Which small-business segment should be prioritised for expansion?
• Which use case should define the next product release?
• Which customers are at risk of churn and require differentiated intervention?

A dataset may contain hundreds of variables, but not all are strategically relevant. The team should define the market, unit of analysis, decision horizon, inclusion criteria, and success measures before selecting methods.

3.2 Qualitative discovery

Qualitative research is particularly important before quantitative segmentation. Useful methods include semi-structured interviews, contextual observation, diary studies, service safaris, customer-support analysis, sales-call review, and ethnographic research.

A strong interview explores actual episodes rather than general opinions:

1. What happened the last time the problem occurred?
2. What triggered the search for a solution?
3. What alternatives were considered?
4. What made adoption difficult?
5. What consequences followed from success or failure?
6. Who else influenced the decision?

Researchers should seek variation across contexts and avoid recruiting only enthusiastic existing customers. Qualitative data can reveal segment dimensions, language, barriers, and causal mechanisms, but it cannot by itself estimate segment prevalence.

3.3 Quantitative validation

A survey can measure needs, attitudes, barriers, behaviours, willingness to pay, and stated preferences across a larger sample. Sampling should reflect the intended market, and questionnaire items should be clear, non-leading, and suitably scaled.

Common approaches include:

• A priori segmentation: groups are defined in advance, such as enterprise versus small business.
• Post hoc segmentation: groups are discovered from measured similarities.
• Hybrid segmentation: theory defines relevant variables and analytics identifies patterns.

Cluster analysis can be useful, but it is not a machine that discovers objectively true segments. Results depend on variable selection, scaling, missing-data treatment, distance measure, algorithm, number of clusters, and validation sample. Standardising variables may be appropriate when variables are measured on different scales, but standardisation should not erase meaningful business differences.

For a data vector x_i and a cluster centre μ_k, squared Euclidean distance is commonly written as:

Here, x_ij is the value of observation i on variable j, μ_kj is the centre of cluster k on that variable, and p is the number of variables. This distance is appropriate only when numeric variables and the chosen scaling make such differences substantively meaningful. It is not automatically suitable for categorical or mixed data.

The number of clusters should be justified using statistical diagnostics, stability across samples, interpretability, operational usefulness, and managerial consequences. A statistically neat cluster that cannot be reached or served is not a useful market segment.

3.4 Segment profiles

Each candidate segment should be documented using a standard evidence profile:

A segment label should describe a meaningful pattern, such as “time-pressured compliance teams needing auditable workflows,” rather than a vague demographic label such as “busy professionals.”

4. Assessing Segment Attractiveness

A segment is attractive only relative to organisational objectives and capabilities. Relevant criteria include:

1. Size and growth: current demand and credible future potential.
2. Need intensity: importance and frequency of the problem.
3. Willingness and ability to pay: economic value of solving the problem.
4. Competitive intensity: strength of incumbent and substitute offerings.
5. Accessibility: ability to identify, reach, sell to, and support the segment.
6. Strategic fit: alignment with capabilities, brand, technology, and mission.
7. Profitability: expected contribution after acquisition and service costs.
8. Risk: regulatory, reputational, concentration, privacy, and execution risk.
9. Learning and option value: ability to generate knowledge or open adjacent markets.

A weighted score can structure discussion:

A_s is the attractiveness score for segment s, w_j is the weight assigned to criterion j, and z_sj is the segment’s standardised or normalised score on that criterion. The weights should sum to 1 if the score is intended as a weighted average. Scores are decision aids, not objective truths; sensitivity analysis should examine how rankings change when weights or assumptions change.

For example, a firm may assign weights of 0.30 to need intensity, 0.25 to economic potential, 0.20 to accessibility, 0.15 to strategic fit, and 0.10 to risk-adjusted feasibility. Segment A scores 8, 7, 6, 9, and 7 on a ten-point scale. Its score is:

A_A = (0.30×8) + (0.25×7) + (0.20×6) + (0.15×9) + (0.10×7) = 7.45

The calculation does not prove that A is the best target. It makes assumptions visible and allows comparison with alternatives.

5. Targeting Decisions

Targeting is a choice about where to concentrate product and organisational effort. Four traditional approaches are:

• Undifferentiated targeting: one broad offer for the whole market.
• Differentiated targeting: distinct offers or propositions for multiple segments.
• Concentrated targeting: deep focus on one segment or niche.
• Individual or micro-targeting: highly customised treatment based on individual or small-group data.

A product may use a portfolio strategy: one core platform with differentiated packaging, workflows, pricing, or communications. Targeting should consider the total system. Serving many segments may increase revenue but also create complexity, conflicting requirements, and diluted positioning.

Important trade-offs include:

• reach versus depth;
• growth versus profitability;
• standardisation versus customisation;
• present value versus future potential;
• acquisition efficiency versus inclusion;
• segment focus versus network effects.

Targeting decisions should state who is not being prioritised and why. Explicit non-targeting reduces roadmap conflict and prevents every stakeholder from claiming that every feature is essential.

6. Personas: Useful Representations, Not Fictional Customers

A persona is a research-based archetype representing a meaningful pattern of users, buyers, or decision-makers. It is not a fictional biography intended to make a presentation attractive. A responsible persona includes evidence, uncertainty, context, goals, behaviours, constraints, triggers, alternatives, and success criteria.

A practical persona structure includes:

• role and relevant context;
• primary job or desired outcome;
• triggering situation;
• current workflow and alternatives;
• key pains and anxieties;
• decision criteria;
• barriers to adoption;
• information and purchasing channels;
• accessibility and inclusion considerations;
• evidence sources and confidence level;
• implications for product, service, and measurement.

Personas should be constructed from multiple sources. For example, an enterprise persona may combine interviews, product telemetry, procurement records, support cases, and win-loss analysis. A persona should identify variation rather than imply that every member behaves identically.

Responsible persona practice

Personas can reproduce bias when they rely on stereotypes, selective quotations, or overly narrow samples. Teams should:

• distinguish observed facts from interpretation;
• record sample size, date, and source quality;
• avoid unnecessary sensitive attributes;
• test whether the persona excludes users with disabilities or limited access;
• represent non-users and unsuccessful users where relevant;
• review personas periodically as markets and products change;
• avoid using personas as a substitute for direct user contact.

A useful persona is a boundary object: it enables designers, engineers, marketers, sales teams, and executives to discuss evidence consistently. It becomes harmful when treated as an individual-level prediction or as permission to ignore contradictory evidence.

7. Positioning and Perceptual Maps

Positioning is the intended meaning of a product in a target customer’s consideration set. It answers four questions:

1. For whom is the product relevant?
2. What problem or desired outcome does it address?
3. What category or frame of reference helps customers understand it?
4. Why should customers believe it is meaningfully better or different?

A positioning statement can use this structure:

For [target customer and context], [product] is a [frame of reference] that delivers [primary outcome] because [distinctive reason to believe], unlike [relevant alternative].

The statement is an internal strategic tool, not necessarily advertising copy. It should be specific enough to guide product choices and testable enough to be challenged.

A perceptual map represents how customers compare alternatives on selected attributes. The axes should be based on research, not on attributes that merely flatter the firm. A two-dimensional map is a simplification; customers may evaluate products on many dimensions, and perceptions may differ from objective performance.

Figure: An illustrative perceptual map. Actual positions should be estimated from customer research and validated data.

To build a map, researchers may ask respondents to rate or compare alternatives on relevant attributes, or use similarity judgments and multidimensional scaling. The resulting map describes perceived relationships, not objective product specifications. A gap on a map is not automatically an opportunity: it may represent low demand, technical infeasibility, or a position competitors can quickly copy.

8. Integrated Case: Northstar, an Illustrative B2B Product

Background

Northstar is an illustrative software company offering workflow automation to medium-sized professional-services firms. Its initial product serves many industries but has weak retention and a crowded sales pipeline. Management believes the product is “for any team that wants to work more efficiently.”

Problem

Research shows that acquisition is relatively inexpensive, but activation and renewal differ sharply. The product team must decide which market to prioritise and how to reposition the product before investing in major features.

Analysis

The team conducts 24 interviews across existing customers, churned accounts, prospects, and non-users. It reviews onboarding data, support tickets, sales losses, and renewal records. Three recurring need patterns emerge:

1. General task coordination: values ease of adoption and low price but frequently substitutes spreadsheets.
2. Client-delivery control: needs deadlines, approvals, audit trails, and visibility across projects.
3. Regulated workflow governance: needs detailed permissions, evidence retention, and formal compliance reporting.

The second group shows the strongest retention and expansion signals. The third has high potential but long procurement cycles and substantial product gaps. The first is large but price-sensitive and expensive to support relative to revenue.

Approach and solution

Northstar selects client-delivery teams as its primary target, while treating regulated workflow governance as a longer-term adjacency. It changes its positioning from “easy workflow automation for everyone” to “auditable delivery coordination for client-service teams.”

The roadmap prioritises approval trails, deadline risk alerts, client-facing status views, and templates for recurring engagements. The team creates two personas: a delivery manager who needs control and visibility, and an account leader who needs confidence that commitments will be met. It does not create a demographic persona because age, gender, and location did not explain the relevant needs.

A quarterly dashboard tracks activation of a first workflow, completion of an approval, time to value, retention by segment, expansion, support burden, and exclusion indicators. Sales and marketing are trained to qualify the target context rather than merely collect firm size.

Outcome and learning points

In this illustrative case, the major improvement is strategic coherence: research, targeting, product priorities, personas, positioning, and measurement now reinforce one another. The case does not claim that focus alone guarantees growth. It demonstrates that a segment is valuable when it combines a meaningful need with evidence of reachable demand, economic potential, and organisational fit.

Key lessons are:

• broad market language can conceal incompatible needs;
• retention and service data can reveal segment quality;
• the largest segment is not always the best target;
• positioning must be reflected in the product experience;
• personas should guide decisions, not decorate presentations;
• adjacent segments should be evaluated rather than automatically pursued.

9. Measurement, Experimentation, and Governance

STP decisions require ongoing validation. Useful measures include segment-level acquisition cost, activation, retention, expansion, contribution margin, conversion, referral, satisfaction, and support cost. Results should be analysed by cohort and context rather than only by aggregate averages.

Experiments can test positioning claims, onboarding paths, packaging, messages, and feature value. Randomised experiments are strongest for causal questions when assignment is feasible and ethical. Observational comparisons require caution because segments differ in many ways beyond the intervention.

AI and machine learning can assist with text coding, propensity modelling, recommendation, look-alike modelling, and anomaly detection. They do not remove the need for human judgment. A model may optimise for historical conversion and thereby reinforce exclusion, target vulnerable people, use proxy variables for protected characteristics, or confuse correlation with need.

Responsible governance should include:

• a documented purpose and lawful basis for personal-data use;
• data minimisation and retention controls;
• transparency about profiling where required;
• review for disparate impact and accessibility;
• human oversight for consequential decisions;
• audit trails for model and segment changes;
• opt-out or contest mechanisms where appropriate;
• restrictions on sensitive attributes and impermissible proxies.

Legal requirements vary by jurisdiction, but ethical responsibility extends beyond legal compliance. A segment should not be targeted merely because it is vulnerable, easy to manipulate, or poorly able to challenge an automated decision. Product managers must consider autonomy, fairness, privacy, security, and social consequences.

10. Common Failure Modes and Best Practices

Failure modes

Demographic determinism: assuming demographic categories explain needs without evidence.

Over-segmentation: creating many small groups that cannot support differentiated action.

Under-segmentation: treating materially different users as one market and producing a compromised product.

Data-driven reification: treating algorithmic clusters as natural social facts.

Vanity personas: inventing names, photographs, and biographies without reliable evidence.

Attribute-led positioning: claiming superiority on features customers do not value.

Ignoring non-users: studying only current customers and missing adoption barriers.

Static strategy: failing to update segments as competition, technology, regulation, and behaviour change.

Best practices

1. Start with a strategic decision and a clear unit of analysis.
2. Combine qualitative understanding with quantitative validation.
3. Segment on variables linked to needs, behaviour, value, or context.
4. Test segment stability and business actionability.
5. Make assumptions, uncertainty, and evidence quality explicit.
6. Evaluate both attractiveness and organisational capability.
7. Define target and non-target segments clearly.
8. Connect positioning to product proof points and customer outcomes.
9. Treat personas as hypotheses that require updating.
10. Monitor outcomes by segment and examine fairness consequences.

Chapter Summary

Segmentation identifies meaningful differences among customers, users, occasions, and needs. Effective segments are measurable, distinguishable, reachable, substantial or strategically important, and actionable. Needs-based, behavioural, contextual, and value-based approaches generally provide stronger product guidance than demographics alone, although demographic and firmographic variables may remain useful descriptors.

Evidence-based segmentation combines qualitative discovery, quantitative measurement, behavioural data, and careful analytics. Cluster methods can reveal patterns, but their outputs depend on methodological choices and must be judged by stability, interpretability, and usefulness. Segment attractiveness should incorporate need intensity, economics, accessibility, competitive conditions, strategic fit, risk, and future option value.

Targeting is an explicit allocation decision involving trade-offs among reach, depth, growth, profitability, standardisation, and customisation. Personas should represent researched patterns rather than stereotypes, and should record uncertainty and evidence. Positioning defines the intended meaning of a product for a selected audience and must be supported by credible product experience. Perceptual maps help compare alternatives, but they simplify complex perceptions and do not guarantee an opportunity.

Finally, STP is iterative. Segment performance, customer behaviour, competitive conditions, and ethical risks must be monitored. AI can support analysis but cannot substitute for research judgment, governance, or responsibility.

Key Terms

Segmentation: Dividing a market into meaningful groups with shared characteristics or needs.

Targeting: Selecting segments for strategic attention and resource allocation.

Positioning: The intended distinctive meaning of a product in a target customer’s consideration set.

Needs-based segmentation: Grouping customers by desired outcomes, problems, or jobs.

Behavioural segmentation: Grouping customers by observed actions and usage patterns.

Value-based segmentation: Grouping customers according to economic or strategic value.

Contextual segmentation: Grouping customers by situation, occasion, location, time, or environment.

Firmographics: Organisational characteristics such as industry, size, and revenue.

Need state: A specific situation and associated problem or desired outcome.

Job to be done: The progress a customer is trying to make in a particular context.

Customer lifetime value (CLV): The expected discounted contribution from a customer relationship, usually after relevant costs.

Customer acquisition cost (CAC): The cost of acquiring a customer.

Persona: An evidence-based archetype representing a meaningful user or buyer pattern.

Perceptual map: A visual representation of how customers perceive relationships among alternatives.

Cluster analysis: A set of methods for grouping observations according to selected similarity measures.

Activation: The point at which a user first experiences meaningful product value.

Cannibalisation: The reduction of one product’s demand caused by another product or offer from the same firm.

Look-alike modelling: Identifying prospects who resemble a selected customer group using observed data.

Actionability: The extent to which a segment can be reached and served with differentiated action.

Review Questions

Short Answer Questions

1. Why is a demographic variable not automatically a useful segmentation variable?
2. Distinguish a buyer, user, and economic decision-maker.
3. What are the main purposes of a persona?
4. Why should a perceptual map not be interpreted as an objective product-performance map?
5. Identify two ethical risks associated with algorithmic segmentation.

Long Answer Questions

1. Explain the STP process and show how it connects customer research to product strategy.
2. Compare needs-based, behavioural, and value-based segmentation, including the contexts in which each is most useful.
3. Discuss the assumptions and limitations of the simplified customer lifetime value model.
4. Explain how a product team should evaluate segment attractiveness and strategic fit.
5. Describe how to construct and govern a responsible persona system for a complex B2B product.

Analytical / Application Questions

1. A product has three customer groups: a large low-retention group, a smaller high-retention group, and a small high-potential group with long procurement cycles. Design a decision framework for selecting a target and justify the data you would collect.
2. Create a research plan to test whether a proposed positioning statement is meaningful, distinctive, credible, and relevant to its intended audience.
3. A machine-learning model identifies a profitable segment partly through location and device variables. Explain how you would test for proxy discrimination, validate business value, and decide whether the segment may ethically be targeted.

Multiple Choice Questions with Answers

Question 1

Which characteristic most directly indicates that a market segment is actionable?

A. It has a memorable name.

B. It can be reached and served with differentiated activities.

C. It contains people of the same age.

D. It is the largest group in the dataset.

Correct Answer: B. It can be reached and served with differentiated activities.

Question 2

Which segmentation basis focuses most directly on desired outcomes and problems?

A. Geographic segmentation

B. Firmographic segmentation

C. Needs-based segmentation

D. Demographic segmentation

Correct Answer: C. Needs-based segmentation

Question 3

In the simplified recurring-revenue CLV formula, what does r represent?

A. Revenue per period

B. Retention probability per period

C. Discount rate per period

D. Acquisition cost

Correct Answer: B. Retention probability per period

Question 4

Why should cluster-analysis results be treated cautiously?

A. Clusters are always entirely subjective.

B. The results depend on variables, scaling, algorithm, and validation.

C. Cluster analysis can use only demographic data.

D. Clusters cannot be compared with business outcomes.

Correct Answer: B. The results depend on variables, scaling, algorithm, and validation.

Question 5

Which statement best describes a responsible persona?

A. A fictional biography designed to make a presentation engaging

B. A demographic stereotype used for advertising

C. An evidence-based archetype that records context, behaviour, and uncertainty

D. A permanent description of every customer

Correct Answer: C. An evidence-based archetype that records context, behaviour, and uncertainty

Question 6

What is the primary purpose of positioning?

A. To calculate customer lifetime value

B. To define the intended distinctive meaning of a product for a target audience

C. To classify all customers by income

D. To replace product testing

Correct Answer: B. To define the intended distinctive meaning of a product for a target audience

Question 7

A perceptual map primarily represents:

A. Customers’ perceived relationships among alternatives

B. A product’s audited technical specifications

C. The legal market share of each competitor

D. The exact future demand for each product

Correct Answer: A. Customers’ perceived relationships among alternatives

Question 8

If the weights in a weighted attractiveness score are intended to form a weighted average, they should generally:

A. All be equal to zero

B. Sum to 1

C. Be larger than the segment scores

D. Use different measurement units

Correct Answer: B. Sum to 1

Question 9

Which is a common failure mode in persona development?

A. Recording evidence sources

B. Including uncertainty

C. Using observed behaviour

D. Inventing biographies without reliable research

Correct Answer: D. Inventing biographies without reliable research

Question 10

Why is human oversight important when AI is used for segmentation?

A. AI cannot process any customer data.

B. Models may reproduce bias or use inappropriate proxy variables.

C. AI always creates too few segments.

D. Human oversight makes validation unnecessary.

Correct Answer: B. Models may reproduce bias or use inappropriate proxy variables.