Customer Problems and Product Discovery Research

Introduction

Product management begins before feature definition. It begins with understanding a meaningful customer problem, the circumstances in which it occurs, the outcomes customers seek, and the constraints that shape their behaviour. Product discovery is the disciplined process of reducing uncertainty about these matters before substantial resources are committed to building and scaling a product.

Discovery is not simply asking customers what features they want. Customers are often better at describing their experiences and workarounds than at designing solutions. The product manager therefore combines several forms of evidence: interviews, observation, surveys, behavioural data, market research, personas, Jobs-to-be-Done (JTBD), outcome-driven thinking and journey mapping. The goal is not to collect opinions but to develop defensible explanations of customer behaviour and identify underserved outcomes.

A useful distinction is between a problem and a solution. “Users need a mobile app” is a proposed solution. “Users cannot reliably know whether a service request has been received and when it will be resolved” is a problem statement. Strong discovery maintains this distinction long enough to test whether the problem is real, important and worth solving.

Learning Objectives

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

1. Distinguish customer needs, pains, desired outcomes, problems and solution ideas.
2. Plan and conduct ethical, reliable customer interviews and observations.
3. Use surveys appropriately for measurement and validation rather than premature discovery.
4. Construct useful personas, JTBD statements and outcome-oriented opportunity statements.
5. Map customer journeys and identify moments of friction, failure and opportunity.
6. Triangulate qualitative and quantitative evidence and recognise research bias.
7. Translate discovery findings into product decisions, assumptions and testable hypotheses.

1. The Nature of Customer Problems

A customer problem is a difficulty, unmet need, risk or constraint experienced by a defined group in a particular context. Problems have at least four dimensions:

• Actor: who experiences the situation?
• Context: when and where does it occur?
• Current behaviour: what does the person currently do?
• Consequence: what functional, emotional or social outcome is affected?

For example, “small-business owners dislike accounting software” is vague. A more useful formulation is: “When preparing quarterly tax information, owners who do not have specialist accounting knowledge spend substantial time reconciling transactions across systems and remain uncertain whether their submission is complete.” This statement identifies context, behaviour and consequence without prescribing a product.

Customer needs can be classified as:

A pain point is an undesirable aspect of an experience, such as delay, cost, uncertainty, effort, risk or embarrassment. A pain point is not automatically a valuable product opportunity. Its frequency, severity, affected population, willingness to change and strategic fit must also be considered.

A problem is especially promising when it is frequent or consequential, poorly served by existing alternatives, experienced by an accessible and economically viable segment, and aligned with the organisation’s capabilities and strategy.

2. Product Discovery as an Evidence Process

Discovery is an iterative learning system rather than a one-time research phase. A typical cycle is:

1. State assumptions about customers, situations and outcomes.
2. Select the smallest appropriate research method.
3. Gather evidence from relevant participants or behavioural sources.
4. Analyse patterns while preserving contradictory evidence.
5. Reformulate the problem and rank opportunities.
6. Test the riskiest assumptions with further research or experiments.

A product team should distinguish three levels of evidence:

• Claim: an interpretation, such as “customers find onboarding confusing.”
• Evidence: observations, quotations, behaviour or measured data supporting the claim.
• Decision: a proposed action, such as simplifying identity verification.

Confusing these levels creates false certainty. One customer quotation is evidence of that customer’s experience, not proof that an entire market behaves identically. Conversely, a large survey may show that a problem is common but not explain why it occurs.

Exploratory and evaluative research

Exploratory research seeks to discover behaviours, meanings and problems. Interviews, observation and diary studies are common. Evaluative research assesses a proposed concept, prototype or experience. Usability testing, concept tests and controlled experiments are examples. A survey can serve either purpose, but its design must match the purpose.

Discovery research should be explicit about uncertainty. A research brief can state:

• What decision will this research inform?
• What do we currently believe?
• Which assumptions are most risky?
• Which participants experience the situation?
• What evidence would change our decision?

3. Customer Interviews

A customer interview is a structured conversation designed to understand a participant’s experiences, behaviours, motivations and constraints. It is not a sales call, focus group, requirements meeting or opportunity to persuade the participant.

Planning the interview

Begin with a research question rather than a feature question. “How do people currently recover from a failed delivery?” is more useful than “Would you use delivery alerts?” The former investigates behaviour; the latter invites speculation.

Recruit participants based on relevant characteristics and behaviours, not merely demographic convenience. A sampling plan may include current users, non-users, recent switchers, dissatisfied customers and people who use competing alternatives. Include exclusion criteria when conflicts of interest or safety concerns exist.

A semi-structured guide normally contains:

1. Introduction, consent and confidentiality explanation.
2. Context and background.
3. A recent specific episode.
4. Current process and alternatives.
5. Difficulties, workarounds and consequences.
6. Decision criteria and constraints.
7. Closing reflection and opportunity for questions.

The most productive questions are behavioural and temporal: “Tell me about the last time…”, “What happened next?”, “How did you decide?”, and “What did you do when that failed?” Follow-up probes should clarify rather than lead.

Weak question: “Would an automated reminder save you time?”

Stronger question: “Think about the last time you missed or nearly missed this task. What did you use to remember it, and what happened?”

Conducting and analysing interviews

Interviewers should listen more than they speak, tolerate pauses, avoid correcting participants and record exact language. With permission, audio recording can improve analysis; otherwise, detailed notes should distinguish quotation from interpretation.

Analysis begins with familiarisation and coding. Codes may describe behaviour (“checks status by phone”), context (“during shift change”), pain (“uncertain ownership”) or desired outcome (“know whether action is complete”). Researchers then cluster related codes into themes and compare them across segments.

A thematic finding should include scope and evidence: “Across seven of ten dispatch coordinators, unresolved orders were checked manually at shift change; participants described this as both time-consuming and a source of handover errors.” This is stronger than “users hate the dashboard.”

Interview evidence is vulnerable to recall error, social desirability, interviewer influence and recruitment bias. Asking about recent behaviour, seeking disconfirming cases and comparing interviews with observed behaviour reduces these risks.

4. Observation and Contextual Inquiry

Observation reveals the difference between reported behaviour and actual behaviour. People may omit routine workarounds, fail to remember interruptions or describe an idealised process. Contextual inquiry combines observation with brief questions in the setting where work occurs.

An observation plan should specify:

• setting and time period;
• actors, tools and handoffs;
• events to observe;
• interruptions and exceptions;
• privacy and safety boundaries;
• method of recording evidence.

The researcher should avoid changing the workflow unnecessarily. Notes should record what happened, not only what the observer inferred. For example, “the operator opened three systems and copied an order number twice” is an observation; “the operator is frustrated by poor integration” is an interpretation requiring supporting evidence.

Observation is particularly valuable for complex services, operational products, accessibility research and situations involving tacit knowledge. It may be impractical where behaviour is rare, private or highly distributed. Diary studies, remote screen recordings or artefact reviews can supplement direct observation.

5. Surveys: Measurement with Discipline

Surveys are useful when the product team needs comparable responses from a larger population. They can estimate prevalence, segment respondents, measure attitudes or prioritise already-defined problems. They are weak substitutes for early discovery because respondents may not recognise a problem, may answer hypothetically or may select an option that sounds socially desirable.

A sound survey requires a defined population, an appropriate sampling strategy, neutral wording and a response format suited to the construct. Avoid double-barrelled questions such as “How fast and reliable is the service?” and avoid leading questions such as “How valuable would our convenient new feature be?” Ask about behaviour before attitudes where possible.

Response rates alone do not establish representativeness. Non-response bias occurs when people who respond differ systematically from those who do not. Convenience samples can be useful for directional learning but should not be presented as population estimates.

For a proportion estimate from a simple random sample, the approximate standard error is:

Here, p̂ is the observed sample proportion and n is the sample size. This approximation assumes independent observations and a sample design close to simple random sampling. It does not correct for coverage error, non-response bias or poorly worded questions. For example, if 60 of 200 respondents report a behaviour, p̂ = 0.30. The estimated standard error is approximately √(0.30 × 0.70 / 200) ≈ 0.032. This describes sampling variability, not total survey error.

In product discovery, statistical precision should not disguise weak construct validity. Measuring the wrong question very precisely remains strategically unhelpful.

6. Personas: Useful Abstractions, Not Stereotypes

A persona is a research-based representation of a meaningful user pattern. It combines relevant goals, behaviours, contexts, constraints and motivations. Personas help teams maintain empathy and make design trade-offs, but they should not be fictional biographies detached from evidence.

A practical persona includes:

• role and relevant context;
• primary tasks and desired outcomes;
• behaviours and tools;
• barriers and anxieties;
• accessibility or environmental constraints;
• evidence base and confidence level.

Avoid demographic decoration that does not influence the product decision. “Alex, 34, likes travel” is not useful unless age or travel behaviour changes the problem. Personas should also show variation within a segment. A buyer, administrator and end user may have different outcomes and decision authority.

Personas should be revisited when evidence changes. They are communication instruments, not permanent truths or grounds for excluding real people.

7. Jobs-to-be-Done and Outcome-Driven Thinking

The Jobs-to-be-Done perspective focuses on the progress a person is trying to make in a particular situation. A job is not necessarily a task performed with a product. It may include functional, emotional and social dimensions.

A useful job statement is:

When [situation], I want to [motivation or progress], so I can [desired outcome].

For example: “When handing over an unresolved customer issue at the end of a shift, I want to communicate its status and ownership clearly, so the next colleague can act without repeating the investigation.” The statement does not assume a dashboard, messaging system or workflow automation.

Outcome-driven thinking decomposes a job into measurable outcomes. An outcome may be expressed as “minimise the time required to determine whether a request is complete” or “minimise the likelihood of losing context during handover.” Good outcomes describe what customers are trying to improve, avoid prescribing implementation, and can be evaluated.

JTBD is valuable because it shifts attention from demographic categories and feature requests to circumstances and progress. However, it should not be treated as a universal replacement for segmentation. Markets may still require analysis of purchasing authority, regulation, accessibility, culture, economics and identity. Jobs are also interpreted through social and organisational contexts; they are not purely individual.

8. Journey Mapping

A customer journey map represents the sequence of stages through which a person moves while pursuing a goal. It may include actions, touchpoints, emotions, questions, backstage processes, pain points and desired outcomes.

A journey map should have a defined actor, goal and start and end points. A generic service journey might include:

1. Recognise a need.
2. Explore alternatives.
3. Decide and initiate.
4. Use or receive the service.
5. Resolve problems or seek support.
6. Continue, switch or recommend.

Journey mapping is not merely a workshop exercise. Each important claim should be linked to research evidence. Mapping should include exceptions and failure paths, not only the ideal happy path. High-value opportunities often occur at handoffs, delays, recovery moments and points of uncertainty.

A service blueprint extends a journey map by showing frontstage interactions and backstage people, systems and processes. This is particularly important when a customer-visible problem is caused by operational constraints rather than interface design.

Figure: A journey map links stages of a goal-directed experience to observed experience and opportunity areas.

9. Synthesis and Opportunity Prioritisation

Synthesis turns raw material into an explicit model of the problem. Teams should maintain an evidence repository containing notes, quotations, observations, survey results and metadata about source and date. Affinity mapping can group related evidence, but clustering must not erase contradictory cases.

A problem statement should identify a user, context, unmet need and consequence. An opportunity statement then describes an area for improvement without specifying a feature. For example: “How might we reduce uncertainty during the waiting period without increasing manual support workload?”

Prioritisation can consider dimensions such as:

• importance to the customer;
• frequency or incidence;
• dissatisfaction with current alternatives;
• strategic fit;
• feasibility and capability;
• reach and economic value;
• ethical and regulatory risk.

A simple scoring table can support discussion, but scores are not objective facts. Teams should document the evidence and confidence behind each score. High importance with low confidence usually calls for more discovery rather than immediate delivery.

10. Applied Case: Illustrative Online Clinic Discovery

Background

An illustrative digital-health company observes that many patients abandon online appointment booking. The initial proposed solution is a redesigned booking interface.

Problem and research plan

The product team reframes the question: “What prevents patients from completing appointment booking, and what outcome are they trying to achieve?” It recruits recent abandoners, successful bookers, people who telephone the clinic and clinic staff. It conducts interviews, observes booking sessions with consent, reviews support contacts and sends a short survey after identifying the main behavioural patterns.

Analysis

Interviews reveal that patients are not merely confused by navigation. Many are uncertain which appointment type matches their symptoms. Some fear choosing incorrectly and being rejected at the clinic. Observation shows that patients open a separate search engine to interpret medical terms. Staff report that incorrect bookings create rework. The journey map identifies a major pain point between “select reason” and “confirm appointment,” where uncertainty is high and consequences appear serious.

The team creates two evidence-based personas: a time-constrained patient seeking a quick, confident decision, and a caregiver booking on behalf of another person. A JTBD statement is: “When arranging care for an unfamiliar symptom, I want to choose an appropriate appointment type with confidence, so the clinic will be prepared and I will not need to repeat the process.”

Approach and outcome

Instead of immediately redesigning every screen, the team tests a small intervention: plain-language explanations, a guided question path and transparent escalation to staff when uncertainty remains. It measures completion, incorrect appointment rate, support contacts and patient-reported confidence. The product decision is based on the outcome sought—appropriate and confident booking—rather than on the original feature request.

Learning points

The case illustrates that abandonment may be a symptom rather than the underlying problem. Interviews exposed meaning and fear; observation exposed workarounds; operational data exposed consequences; and the survey helped estimate how widespread the pattern was. In healthcare, consent, privacy, accessibility, clinical safety and non-diagnostic communication are essential constraints. A product manager must not treat convenience as more important than safe care.

11. Ethics, Governance and Responsible Use of Technology

Research participants should understand the purpose of the study, what participation involves, how data will be used, and that participation is voluntary. Collect only data necessary for the research question, protect identifying information, set retention limits and restrict access. Sensitive domains may require institutional review, sector-specific regulation or specialist consent procedures.

Researchers should avoid coercive incentives, deceptive recruitment and unnecessary collection of health, financial, biometric or location data. Accessibility must be considered in recruitment, consent, interview format and research materials. Findings should not systematically exclude people with disabilities, limited connectivity, language differences or low digital confidence.

AI tools can assist with transcription, translation, clustering and summarisation, but they introduce risks of confidentiality breaches, hallucinated quotations, dialect bias and false thematic certainty. Teams should obtain permission before processing recordings, use approved systems, inspect original evidence, label machine-generated output and retain human responsibility for interpretation. AI may accelerate analysis; it does not replace contextual understanding or ethical judgement.

12. Common Failure Modes and Best Practices

Common failures include interviewing only enthusiastic existing users, asking leading questions, treating feature requests as requirements, relying on hypothetical intent, presenting personas as facts, ignoring backstage operations and stopping after a single attractive pattern.

Best practices are to:

• begin with decisions and assumptions;
• recruit for relevant experiences and edge cases;
• ask about recent behaviour and alternatives;
• triangulate methods without pretending they measure the same thing;
• separate evidence, interpretation and decision;
• seek disconfirming evidence;
• record confidence and unresolved questions;
• include accessibility, safety and operational stakeholders;
• revisit findings as the product and market change.

Chapter Summary

Customer discovery is the disciplined investigation of customer situations, behaviours, needs, pains and desired outcomes before committing to solutions. Interviews reveal meaning and recent experience; observation reveals context and workarounds; surveys support measurement when sampling and question design are appropriate. Personas provide evidence-based representations of behavioural patterns, while Jobs-to-be-Done frames the progress customers seek in particular circumstances. Outcome-driven thinking focuses on improvements customers value without prescribing implementation. Journey maps connect stages, touchpoints, emotions, failures and opportunities, and service blueprints extend this view into backstage operations.

Effective product managers triangulate evidence, distinguish claims from observations, test risky assumptions and preserve uncertainty where evidence is weak. Discovery must also respect consent, privacy, accessibility, safety and regulatory obligations. The central professional discipline is to move from “What should we build?” to “What problem is important, for whom, in what context, and what evidence would justify acting?”

Key Terms

Customer problem: A difficulty, unmet need, risk or constraint experienced in a defined context.

Pain point: An undesirable aspect of an experience, such as delay, effort, uncertainty or risk.

Product discovery: Iterative research and reasoning used to reduce uncertainty before or during product development.

Qualitative research: Research focused on meanings, experiences, behaviours and context.

Quantitative research: Research using numerical data to estimate, compare or measure patterns.

Customer interview: A structured conversation used to investigate experience, behaviour and motivation.

Contextual inquiry: Observation combined with questions in the setting where work occurs.

Survey: A structured instrument for collecting comparable responses from participants.

Persona: An evidence-based representation of a meaningful user pattern.

Jobs-to-be-Done: A perspective focusing on the progress a person seeks in a particular situation.

Outcome: The result or improvement a customer seeks to achieve or avoid.

Journey map: A visual or textual representation of stages in a goal-directed experience.

Service blueprint: A journey representation that also shows backstage processes and systems.

Triangulation: Comparing evidence from different sources or methods.

Research bias: A systematic influence that distorts recruitment, data collection, analysis or interpretation.

Problem statement: A concise description of actor, context, unmet need and consequence.

Opportunity statement: A non-prescriptive area for improving a customer outcome.

Latent need: An important need that customers may not explicitly articulate.

Non-response bias: Distortion caused when respondents differ systematically from non-respondents.

Review Questions

Short Answer Questions

1. What distinguishes a customer problem from a proposed solution?
2. Why are recent behavioural questions generally more useful than hypothetical feature questions?
3. State the main components of a useful persona.
4. What is the purpose of a journey map?
5. Why should survey results not automatically be treated as representative?

Long Answer Questions

1. Explain how interviews, observation and surveys can be combined in a product discovery cycle. Discuss the distinct contribution and limitation of each method.
2. Describe Jobs-to-be-Done and outcome-driven thinking, and explain how they prevent premature commitment to features.
3. Explain how to conduct a rigorous customer interview, including recruitment, questioning, recording and analysis.
4. Discuss the relationship between personas, journey maps and service blueprints in understanding a service problem.
5. Evaluate the ethical and governance responsibilities involved in using customer research data and AI-assisted analysis.

Analytical / Application Questions

1. A banking team hears that customers want “a faster app.” Design a discovery plan that distinguishes speed from underlying functional, emotional and risk-related outcomes.
2. A survey finds that 78% of respondents say they would use a proposed feature, but support records show little evidence of the underlying behaviour. Analyse possible explanations and recommend next research steps.
3. Using the online clinic case, identify the initial solution assumption, the deeper customer problem, the relevant evidence sources and two measures that would indicate whether the intervention improved the outcome.

Multiple Choice Questions with Answers

Question 1

Which statement is a problem rather than a proposed solution?

A. Users need a mobile app.

B. Users need dashboard automation.

C. Users cannot reliably determine whether their service request has been received.

D. Users need push notifications.

Correct Answer: C. Users cannot reliably determine whether their service request has been received.

Question 2

What is the primary strength of contextual inquiry?

A. It estimates market prevalence with statistical precision.

B. It reveals behaviour and workarounds in the setting where work occurs.

C. It eliminates the need for participant consent.

D. It directly determines the final product design.

Correct Answer: B. It reveals behaviour and workarounds in the setting where work occurs.

Question 3

Which interview question is most behaviourally grounded?

A. Would you like an automated reminder?

B. Do you think this feature is valuable?

C. How satisfied are you with technology generally?

D. Tell me about the last time you missed this task and what you did next.

Correct Answer: D. Tell me about the last time you missed this task and what you did next.

Question 4

What is a principal limitation of a convenience sample in a survey?

A. It cannot contain numerical responses.

B. It may not represent the wider population.

C. It always produces a zero response rate.

D. It cannot measure attitudes.

Correct Answer: B. It may not represent the wider population.

Question 5

Which statement best reflects the Jobs-to-be-Done perspective?

A. Customers should be segmented only by age.

B. A job is always a feature requested by a user.

C. The focus is on progress sought in a particular situation.

D. Jobs-to-be-Done removes the need for market analysis.

Correct Answer: C. The focus is on progress sought in a particular situation.

Question 6

What should a journey map include to support credible product decisions?

A. Only the ideal happy path.

B. Stages, touchpoints, experience and evidence-based pain points.

C. Only the organisation’s internal processes.

D. A list of competitors’ features.

Correct Answer: B. Stages, touchpoints, experience and evidence-based pain points.

Question 7

In the survey formula SE(p̂) = √(p̂(1 − p̂) / n), what does n represent?

A. The observed proportion

B. The number of survey questions

C. The sample size

D. The non-response rate

Correct Answer: C. The sample size

Question 8

Which practice best reduces interview recall error?

A. Ask participants to predict all future behaviour.

B. Ask about a recent, specific episode.

C. Explain the preferred product before questioning.

D. Recruit only satisfied customers.

Correct Answer: B. Ask about a recent, specific episode.

Question 9

What is an opportunity statement intended to do?

A. Specify the final technical architecture.

B. Describe an area for improvement without prescribing a feature.

C. Replace all customer evidence with intuition.

D. Guarantee product-market fit.

Correct Answer: B. Describe an area for improvement without prescribing a feature.

Question 10

Which is an appropriate governance practice when using AI to analyse interview recordings?

A. Upload recordings to any public tool without consent.

B. Accept generated quotations without checking source material.

C. Use approved systems, protect confidentiality and verify outputs against evidence.

D. Allow AI output to determine decisions without human review.

Correct Answer: C. Use approved systems, protect confidentiality and verify outputs against evidence.