Market, Customer and Product Analytics

Introduction

Product managers make better decisions when they replace unsupported assumptions with structured evidence. Market, customer and product analytics provide that evidence. Together, they help determine which customers to serve, what problems to solve, how to design an offer, how much to charge, how large an opportunity may be, and how to respond to competitors.

These activities are related but not identical. Market analytics examines the external environment and opportunity size. Customer analytics studies customer characteristics, behaviour, needs and experiences. Product analytics examines how people discover, use, adopt, retain and value a product. Marketing research supplies the primary and secondary data required by all three.

Analytics does not eliminate managerial judgement. Surveys may contain bias, behavioural data may lack context, and market estimates may depend on uncertain assumptions. The professional task is therefore to combine qualitative understanding, quantitative evidence and strategic judgement in a transparent decision process.

Learning Objectives

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

1. Segment markets using meaningful and actionable variables.
2. Evaluate and select target segments.
3. Develop evidence-based customer personas.
4. Map customer journeys and identify friction points.
5. Design basic concept tests and interpret their results.
6. Explain the logic of conjoint analysis and its uses.
7. Apply principles of pricing research and demand estimation.
8. Analyse competitors using structured frameworks.
9. Estimate market size using TAM, SAM and SOM.
10. Integrate analytical findings into product decisions.

1. From Market Information to Product Decisions

A useful analytical chain is:

Market context → customer problem → segment opportunity → product concept → value proposition → price and route to market → adoption and retention outcomes.

Each link requires different evidence. Industry reports may indicate market growth but cannot prove that a particular customer will buy. Interviews may reveal an important problem but cannot estimate its frequency in the population. Product usage data may reveal abandonment but not necessarily the reason for it. Strong decisions therefore triangulate multiple sources.

Primary and secondary research

Primary research is collected for the current decision, including interviews, observation, surveys, experiments and concept tests. Secondary research uses existing sources, such as government statistics, trade publications, company reports, academic studies and internal records.

Primary research offers relevance and control but may be expensive and affected by sampling or question-design problems. Secondary research is often faster and cheaper, but its definitions, dates, populations and methods may not match the manager's question.

Descriptive, diagnostic and predictive analytics

• Descriptive analytics asks what happened: sales by region, conversion rate or churn rate.
• Diagnostic analytics asks why it happened: an onboarding step may cause abandonment, or a competitor may have changed its price.
• Predictive analytics estimates what may happen: expected demand, probability of churn or likely adoption.
• Prescriptive analysis evaluates what should be done: which segment to target, which feature bundle to launch or which price to test.

These categories are not substitutes for one another. A predictive model built on poorly defined data can produce precise-looking but misleading results.

2. Market Segmentation and Targeting

Meaning of segmentation

Market segmentation divides a broad market into groups whose members have sufficiently similar needs, behaviours or responses to justify differentiated product or marketing decisions. A segment is not merely a demographic category. It must have a meaningful relationship to the product problem and the firm's ability to serve it.

Common segmentation bases include:

The strongest segmentation often combines variables. For example, “small businesses” is broad; “owner-managed cafés with fewer than ten employees that need simple same-day inventory visibility” is more decision-relevant.

Requirements for useful segments

A segment should generally be:

• Identifiable: its members can be described and measured.
• Substantial: it is large or valuable enough to justify attention.
• Accessible: the organisation can reach and serve it.
• Differentiable: it responds differently from other groups.
• Actionable: the organisation can design a realistic offering for it.
• Stable enough: it will not disappear before the strategy can be implemented.

A segment can be statistically distinct but commercially useless. For example, a cluster-analysis solution may produce four groups with different attitudes, but if the groups cannot be reached through different channels or served profitably, the distinction has little managerial value.

Targeting and positioning

Targeting evaluates segments against strategic criteria such as growth, competitive intensity, customer accessibility, fit with capabilities, willingness to pay and expected profitability. A simple prioritisation score may be created by assigning weights to criteria, provided that the scoring assumptions are documented and tested.

A positioning statement clarifies the intended meaning of the product in the target customer's mind. A practical structure is:

For [target customer] who needs [important problem or outcome], [product] is a [category or frame of reference] that provides [distinctive benefit] because [reason to believe].

Positioning is not merely a slogan. It should guide product scope, experience design, pricing and communication.

3. Customer Personas and Jobs to Be Done

A persona is a research-based representation of a meaningful customer type. It should synthesise evidence rather than describe an imaginary individual. A useful persona normally includes:

• context and responsibilities;
• goals and desired outcomes;
• current alternatives and workarounds;
• barriers, anxieties and triggers;
• buying authority and stakeholders;
• relevant behaviours and usage situations;
• evidence supporting the claims.

Demographic decoration should be limited. A fictional name, photograph or quote can aid communication, but it does not make a persona valid. The essential question is whether the persona helps the team make a different product decision.

The jobs-to-be-done perspective focuses on what customers are trying to accomplish in a particular situation. A job may be functional, emotional or social. For example, a commuter may “track transport arrival times” functionally, “feel in control” emotionally and “appear reliable to colleagues” socially. Product research should examine desired outcomes, not only stated feature requests.

Avoiding persona bias

Personas may reproduce internal stereotypes if they are based only on sales opinions or a few interviews. To improve quality:

1. define the research question before recruiting participants;
2. include variation in experience, access and usage;
3. distinguish observed behaviour from interpretation;
4. record contradictory evidence;
5. validate the persona against behavioural or market data;
6. review it when the product, market or customer population changes.

4. Customer Journey Mapping

A customer journey map describes the sequence of stages through which a customer moves before, during and after an interaction with an organisation. Typical stages include awareness, consideration, purchase, onboarding, use, support and renewal or advocacy.

A robust map may include the customer's goal, actions, questions, emotions, touchpoints, backstage processes, pain points and opportunities. It should distinguish the customer's experience from the organisation's internal process.

Figure: A simplified customer journey showing goals and friction across stages.

Journey mapping is most valuable when linked to metrics. For example, consideration may be measured by qualified trial starts, purchase by checkout completion, use by activation or task success, and renewal by retained accounts. A map without evidence can become a workshop poster rather than an analytical instrument.

5. Concept Testing and Product Validation

Concept testing evaluates a proposed product, feature, service or value proposition before substantial investment. A concept may be presented as text, storyboard, prototype, landing page or interactive mock-up. The test should match the maturity of the decision: early research may test the problem and appeal; later research may test comprehension, usability, preference and purchase likelihood.

A sound concept test defines:

1. the decision to be made;
2. the target population and sampling method;
3. the stimulus and exposure conditions;
4. the measures, such as relevance, uniqueness, credibility, intent and barriers;
5. comparison concepts or a control where appropriate;
6. the threshold for action.

Stated purchase intention is not the same as actual purchase. Respondents may be polite, may misunderstand the offer or may not face the real cost and effort of adoption. Behavioural indicators—clicks, sign-ups, task completion, deposits, repeat use or experiment-based conversion—usually provide stronger validation.

Basic experimental logic

When possible, compare a treatment group exposed to a new concept with a suitable control group. If p_T is the outcome proportion in the treatment group and p_C is the outcome proportion in the control group, the absolute difference is:

Here, ∆p is measured in proportion points. The comparison is informative only when assignment, exposure, measurement and relevant conditions are sufficiently comparable. A difference may still be caused by sampling variation, seasonality, novelty or another simultaneous change; statistical and managerial interpretation must therefore be combined.

6. Conjoint Principles

Conjoint analysis studies how people value combinations of product attributes and levels. For example, a streaming service may vary content access, advertising, device support and monthly price. Respondents evaluate complete alternatives rather than rating each feature in isolation.

A simplified part-worth model represents the utility of an alternative as:

U_j is the utility of alternative j; β₀ is an intercept when applicable; β_k is the estimated contribution of attribute level k; x_jk indicates whether that level is present; β_p is the price coefficient; p_j is price; and ε_j is unobserved variation. The signs and scale depend on coding and model specification. The equation is a conceptual representation, not a guarantee that customer choice is linear or additive.

In a choice-based conjoint study, the probability of choosing alternative j under a multinomial logit model is commonly represented as:

J is the number of alternatives in the choice set. This formulation assumes that the utility specification is appropriate and that the error structure supports the logit model. It should not be interpreted as a direct observation of market share unless realistic alternatives, awareness, availability and competitive conditions are incorporated.

Conjoint analysis can estimate relative attribute importance, preference for levels, likely choice under specified scenarios and—when a credible price coefficient exists—an indicative willingness to pay. It is limited by attribute selection, hypothetical bias, respondent fatigue, unrealistic choice sets and model assumptions. It should support, not replace, product strategy and market experiments.

7. Pricing Research and Demand Estimation

Price is both a financial variable and a signal of value, quality and positioning. Pricing research should investigate willingness to pay, perceived value, competitive reference prices, affordability, switching costs and the consequences of different pricing structures.

Useful methods include:

• Van Westendorp price sensitivity analysis: asks threshold questions about prices perceived as cheap, expensive, too cheap and too expensive. It is useful for exploring perceptions but does not directly predict purchase or profit.
• Gabor–Granger analysis: presents prices and measures purchase likelihood at each level. It can show a price-response pattern but is vulnerable to hypothetical bias.
• Conjoint analysis: estimates trade-offs between price and other attributes.
• Monadic experiments or A/B tests: expose comparable groups to different prices and measure behaviour. These require careful attention to fairness, legal constraints and long-term effects.
• Value-in-use research: estimates the economic or functional benefit created for a customer, especially in business markets.

Demand, revenue and elasticity

Demand is the quantity customers are willing and able to purchase at different prices under specified conditions. A simple linear demand function is:

where Q is quantity demanded per period, P is price per unit, a is the intercept and b is the price-response coefficient. The model assumes a stable context and a linear approximation over the relevant price range. It should not be extrapolated indefinitely, because real demand may have thresholds, saturation and nonlinear effects.

Revenue is:

where R is revenue, P is price per unit and Q is units sold. Profit requires costs, so revenue should not be mistaken for profitability.

Price elasticity of demand is:

It measures the percentage change in quantity associated with a percentage change in price. Absolute elasticity above 1 is commonly described as elastic; below 1 as inelastic. The interpretation assumes that other important influences are held sufficiently constant and that the measurement interval is clear.

Worked example: linear demand

Suppose a research model estimates Q = 10,000 − 200P, where price is in currency units and quantity is units per month. At a price of 20:

Q = 10,000 − 200(20) = 6,000 units.

Revenue is therefore R = 20 × 6,000 = 120,000 currency units per month. At a price of 25, demand is 5,000 units and revenue is 25 × 5,000 = 125,000. Under this simplified model, the higher price produces greater revenue, although costs, customer acquisition, retention and competitive reactions must still be considered before making a pricing decision.

8. Competitor Analysis

Competitor analysis should examine not only named rivals but also substitutes, manual workarounds, internal solutions and the option of doing nothing. A competitor matrix can compare:

• target customers and use cases;
• core job addressed;
• features and performance;
• price and monetisation model;
• distribution and acquisition channels;
• brand and trust signals;
• switching costs and ecosystem effects;
• strengths, weaknesses and likely responses.

The competitive frame of reference identifies the alternatives customers use when evaluating a product. A product may compete with a different category if customers regard it as a substitute for the same job.

Porter's Five Forces can structure an industry-level assessment: rivalry, threat of new entrants, supplier power, buyer power and threat of substitutes. It is a strategic framework, not a numerical market-size model. A product manager should combine it with direct customer evidence and current competitor observations.

Competitor intelligence must be collected ethically. Public prices, published product information, lawful mystery shopping and customer interviews may be appropriate. Obtaining confidential information through deception, unauthorised access or improper inducement is not.

9. Market Sizing: TAM, SAM and SOM

Market sizing clarifies the scale of an opportunity and the assumptions behind a business case.

• TAM (Total Addressable Market): the total potential demand if the product could serve every relevant customer under the defined conditions.
• SAM (Serviceable Available Market): the portion of TAM that fits the product's geography, category, regulatory scope and capabilities.
• SOM (Serviceable Obtainable Market): the realistically attainable portion of SAM within a stated time period, considering competition, capacity and execution.

A top-down estimate begins with an external market total and applies defensible filters. A bottom-up estimate starts with customer units, adoption, frequency and price. The bottom-up approach is often more useful for product planning because its assumptions map directly to operational drivers.

A basic annual revenue estimate is:

where N is the number of potential customers or units, A is the proportion adopting during the period, F is purchase frequency or annual quantity, and P is average price. All variables must refer to compatible populations and periods.

Worked market-sizing example

An illustrative software product serves 50,000 eligible small firms in a defined region. Research suggests that 20% have the required operational need, 30% of those are reachable through the firm's channels, and the expected annual subscription is 240 currency units.

The serviceable customer base is:

50,000 × 0.20 × 0.30 = 3,000 firms.

The corresponding serviceable annual revenue opportunity is:

3,000 × 240 = 720,000 currency units.

This is not automatically SOM. If the company can acquire and support only 600 customers in the first three years, a simple attainable revenue ceiling at full subscription price would be 600 × 240 = 144,000 per year, before churn, discounts, costs and competitor reaction. The estimate should be presented as a range when the assumptions are uncertain.

10. Integrating Analytics into Product Decisions

Analytics becomes useful when linked to explicit decisions. A product research brief should state:

1. the decision and deadline;
2. the customers and market boundary;
3. the hypotheses to test;
4. the data required and its quality limits;
5. the decision rule or success threshold;
6. the owner of the decision;
7. what will happen if evidence is inconclusive.

A practical evidence portfolio may combine interviews for problem discovery, a survey for prevalence, behavioural data for actual usage, concept testing for appeal, conjoint analysis for trade-offs, pricing research for willingness to pay and a market model for scale. No single method answers every question.

Illustrative Case: FlowDesk for Independent Retailers

Background

FlowDesk is an illustrative cloud application intended to help independent retailers manage stock and supplier reordering. The founding team initially assumed that all small retailers wanted an advanced forecasting dashboard.

Problem

Early interviews revealed that many owners did not describe forecasting as their primary need. They wanted to avoid stockouts on fast-selling items, reduce time spent checking spreadsheets and receive simple reorder recommendations. The team also discovered two potential segments: technologically confident multi-outlet retailers and time-constrained single-store owners.

Analysis

The team conducted 20 semi-structured interviews, analysed anonymised pilot usage data from a spreadsheet-import prototype and surveyed 300 eligible retailers. The evidence suggested that single-store owners experienced the problem more frequently, but multi-outlet retailers had higher willingness to pay and stronger ability to adopt software. A journey map showed severe friction during data import and onboarding.

A concept test compared three offers: a forecasting dashboard, automated reorder alerts and a combined package. The reorder-alert concept generated clearer comprehension and stronger trial interest. A conjoint exercise then varied alert frequency, integration options, support level and monthly price. Results indicated that reliable integration and setup support mattered more than an extensive dashboard, while price sensitivity increased among single-store firms.

Approach and solution

The team selected multi-outlet retailers as its initial target segment. It repositioned the product around “preventing stockouts with dependable reorder guidance,” reduced the first release to core alerts and integrations, and introduced assisted onboarding. Pricing research supported a tiered plan based on the number of outlets, with a time-limited pilot used to measure actual conversion rather than relying only on stated intent.

Outcome

In the pilot, the team measured activation as importing a catalogue, connecting at least one supplier and acting on a reorder alert. This definition exposed that many trial users registered but never reached useful product value. Improving import guidance increased activation, while removing low-value dashboard features reduced development effort. The team did not treat the pilot as proof of long-term product-market fit; it scheduled further research on retention, support cost and competitor response.

Learning points

1. A broad market label concealed different needs and economics.
2. The customer's job was more actionable than the requested feature.
3. Concept clarity and onboarding were as important as feature attractiveness.
4. Stated interest needed to be checked against behaviour.
5. Targeting requires balancing need, value, reachability and organisational capability.

11. Limitations, Ethics and Governance

Market and customer analytics can cause harm when data is collected or used irresponsibly. Researchers should obtain appropriate consent, explain intended use, minimise data collection, protect confidentiality and comply with applicable privacy and consumer-protection law. Sensitive characteristics should not be used to unfairly exclude customers or manipulate vulnerable groups.

Important governance practices include:

• documenting data sources, definitions and retention periods;
• restricting access according to role;
• testing datasets for missingness and biased coverage;
• distinguishing correlation from causation;
• auditing automated segmentation and recommendation systems;
• providing human review for consequential decisions;
• avoiding dark patterns and deceptive price presentation;
• allowing customers to understand meaningful data uses.

AI can assist with interview transcription, thematic coding, forecasting, competitor monitoring and persona synthesis. However, generated summaries may omit minority views, reproduce stereotypes or present uncertain claims as facts. Human researchers must verify source material, protect confidential information and retain an auditable link between evidence and decision.

Common analytical errors

• treating a convenience sample as representative;
• asking leading or double-barrelled survey questions;
• confusing intention with behaviour;
• using market reports with incompatible definitions;
• counting competitors without analysing substitutes;
• presenting a single market-size number without assumptions;
• interpreting correlation as a causal effect;
• optimising conversion while ignoring retention or customer welfare;
• selecting a segment only because it is large;
• allowing dashboards to determine strategy without a decision context.

Best-Practice Checklist

Before approving a product decision, ask:

• Is the target customer clearly defined?
• Is the customer problem supported by more than one evidence source?
• Are segment boundaries and inclusion criteria explicit?
• Does the persona represent a meaningful behavioural group?
• Are journey pain points connected to measurable outcomes?
• Was the concept tested with realistic stimuli and comparisons?
• Are price and demand assumptions stated?
• Does the market estimate show its arithmetic and uncertainty?
• Have substitutes and likely competitor responses been considered?
• Are privacy, fairness and data-quality risks addressed?
• What evidence would change the decision?

Chapter Summary

Market, customer and product analytics connect research evidence to product decisions. Segmentation divides markets into groups that differ in needs or responses; targeting selects segments according to attractiveness, fit and ability to serve; and positioning defines the intended value in the customer's mind. Personas and jobs-to-be-done provide practical representations of customer context, while journey maps reveal friction across discovery, purchase, use and renewal.

Concept testing evaluates proposed solutions, but stated intention should be supplemented with behavioural evidence. Conjoint analysis estimates trade-offs among product attributes and price under explicit modelling assumptions. Pricing research may use perceptual, stated-choice and experimental approaches, while demand models help examine the relationship between price, quantity and revenue. Competitor analysis must include substitutes and the option of doing nothing. TAM, SAM and SOM clarify market boundaries and attainable opportunity, provided that assumptions are transparent.

The professional product manager integrates multiple methods, recognises uncertainty, protects customer interests and converts findings into explicit decisions, metrics and learning plans.

Key Terms

Segmentation: Division of a market into groups with relevant similarities.

Targeting: Selection of market segments for focused organisational action.

Positioning: Intended meaning and distinctive value of a product relative to alternatives.

Persona: Research-based representation of a meaningful customer type.

Jobs to be done: Functional, emotional or social progress a customer seeks in a situation.

Customer journey map: Structured representation of customer stages, actions, touchpoints and friction.

Concept test: Research evaluation of a proposed product or offer before full investment.

Control group: Comparable group not receiving the experimental treatment.

Conjoint analysis: Method for estimating preferences for combinations of attributes and levels.

Part-worth utility: Estimated contribution of an attribute level to preference or choice.

Willingness to pay: Maximum amount a customer is prepared to exchange for an offer under specified conditions.

Demand: Quantity customers are willing and able to purchase at specified conditions.

Price elasticity: Percentage response in quantity associated with a percentage change in price.

Competitor: A provider or alternative that addresses the same or a related customer job.

Substitute: Different solution that customers may use instead of the focal product.

TAM: Total potential market under an unrestricted service assumption.

SAM: Portion of TAM within the product's serviceable boundaries.

SOM: Realistically obtainable portion of SAM over a stated period.

Triangulation: Use of multiple evidence sources or methods to improve confidence.

Activation: Achievement of an early product-use event indicating initial value realisation.

Review Questions

5 Short Answer Questions

1. What characteristics make a market segment actionable?
2. How does a persona differ from a demographic profile?
3. What is the purpose of a customer journey map?
4. Why should stated purchase intention be treated cautiously?
5. Distinguish TAM, SAM and SOM.

5 Long Answer Questions

1. Explain how segmentation, targeting and positioning work together in product management.
2. Discuss the strengths and limitations of conjoint analysis for product decisions.
3. Compare major approaches to pricing research.
4. Explain how competitor analysis should include substitutes and non-consumption.
5. Describe how a product team can integrate qualitative, quantitative and behavioural evidence.

3 Analytical / Application Questions

1. A product team identifies three customer groups that differ in need intensity, willingness to pay and reachability. Design a weighted evaluation framework for selecting a target segment, and explain the risks of relying only on segment size.
2. Using the demand function Q = 12,000 − 300P, calculate quantity and revenue at prices of 20 and 30. Explain why the higher-revenue option may not necessarily be the better product decision.
3. Design a research programme for a new subscription product covering segmentation, concept testing, pricing, competitor analysis and market sizing. Specify the evidence required at each stage and the decision it would inform.

Multiple Choice Questions with Answers

Question 1

Which characteristic is essential for an actionable market segment?

A. It must be the largest group in the market.

B. It must be reachable and serviceable by the organisation.

C. It must contain customers of exactly the same age.

D. It must have no competitors.

Correct Answer: B. It must be reachable and serviceable by the organisation.

Question 2

What is the primary purpose of a customer journey map?

A. To calculate total industry revenue

B. To replace all quantitative research

C. To show customer stages, touchpoints and friction

D. To determine accounting profit

Correct Answer: C. To show customer stages, touchpoints and friction

Question 3

In a concept test, why is actual behaviour generally stronger evidence than stated intention?

A. Behaviour is always free from sampling error.

B. Behaviour reflects action under more realistic conditions.

C. Intention cannot be measured in a survey.

D. Behaviour eliminates the need for a control group.

Correct Answer: B. Behaviour reflects action under more realistic conditions.

Question 4

What does conjoint analysis primarily estimate?

A. Employee productivity

B. Preferences for combinations of attributes and levels

C. The legal status of a competitor

D. The exact future market share of a product

Correct Answer: B. Preferences for combinations of attributes and levels

Question 5

If price increases by 10% and quantity demanded decreases by 20%, the price elasticity of demand is:

A. −0.5

B. 0.5

C. −2.0

D. 2.0

Correct Answer: C. −2.0

Question 6

Under the demand function Q = 10,000 − 200P, what quantity is demanded when P = 20?

A. 2,000 units

B. 4,000 units

C. 6,000 units

D. 8,000 units

Correct Answer: C. 6,000 units

Question 7

Which statement best describes SOM?

A. The entire theoretical demand worldwide

B. The portion of the market matching a product's category only

C. The realistically obtainable portion of the serviceable market

D. The market occupied by all competitors combined

Correct Answer: C. The realistically obtainable portion of the serviceable market

Question 8

Which is an ethical practice in competitor analysis?

A. Obtaining a rival's confidential files without permission

B. Using publicly available product and pricing information

C. Misrepresenting identity to obtain restricted information

D. Accessing a competitor's systems without authorisation

Correct Answer: B. Using publicly available product and pricing information

Question 9

In the revenue identity R = P × Q, what does Q represent?

A. Quality score

B. Quantity sold in the relevant period

C. Customer satisfaction

D. Fixed cost

Correct Answer: B. Quantity sold in the relevant period

Question 10

Why should market-size estimates show their assumptions?

A. To make the estimate appear larger

B. To allow evaluation, revision and sensitivity analysis

C. To avoid defining the target customer

D. To ensure the estimate is always exact

Correct Answer: B. To allow evaluation, revision and sensitivity analysis