Research Design and Customer Evidence

Introduction

Product discovery is the disciplined process of reducing uncertainty about a customer problem, its context, the value of possible solutions, and the conditions under which customers may adopt them. Discovery research is not simply the collection of opinions. It is the design of a defensible evidence system that connects research questions to appropriate participants, methods, analysis and product decisions.

For product managers, research quality is a strategic capability. Weak evidence can cause teams to solve an unimportant problem, misunderstand a customer segment, overestimate demand or design for an imagined user. Strong evidence does not eliminate uncertainty, but makes uncertainty visible and enables better decisions about whether to continue, modify, pause or stop an opportunity.

This chapter examines interviews, observation, contextual inquiry, diary studies, surveys and secondary research. It also addresses sampling, triangulation, bias control, research ethics, analysis and the translation of evidence into product decisions.

Learning Objectives

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

1. Formulate answerable discovery research questions.
2. Select and combine qualitative and quantitative research methods.
3. Design interviews, observations, contextual inquiries, diary studies and surveys.
4. Develop defensible sampling and recruitment strategies.
5. Identify and control common sources of research bias.
6. Use triangulation to assess the strength and limits of customer evidence.
7. Apply ethical principles to product research involving people and their data.
8. Translate findings into validated assumptions, opportunity statements and product decisions.

1. Evidence as a Product-Discovery System

A useful distinction is between data, finding, insight and decision.

• Data are observations, statements, behaviours, responses or documents collected during research.
• A finding is a patterned or significant observation derived from data.
• An insight explains why a finding matters in relation to a customer need, behaviour or constraint.
• A decision specifies what the product team will do because of the evidence.

For example, the statement “eight of ten participants mentioned exporting reports” is data summarised as a finding. The insight may be that customers export reports because they must submit evidence to a manager using a separate system. A resulting decision could be to prioritise shareable compliance summaries rather than merely improving the in-product dashboard.

Research should begin with uncertainty, not with a preferred solution. A product team might ask:

• Which users experience the problem most frequently?
• What triggers the problem?
• How do users currently cope?
• What consequences follow from the problem?
• What alternatives do users use, and why?
• Which assumptions about value, feasibility or adoption are most dangerous?

A research question should be narrow enough to investigate and broad enough not to predetermine the answer. “Would customers like our proposed feature?” is usually weak because it invites hypothetical approval. “How do operations managers currently reconcile delayed orders, and what consequences result?” is more useful because it investigates existing behaviour and context.

Evidence quality

Evidence quality has several dimensions:

No method is automatically rigorous. An interview can be insightful but biased; a survey can be numerically precise but measure the wrong construct. Rigour results from alignment among the question, method, sample, analysis and decision.

2. Research Paradigms and Method Selection

Discovery research commonly draws on two complementary traditions.

Interpretivist and qualitative research seeks to understand meanings, practices, experiences and contexts. It is particularly useful when the problem is poorly understood, language is uncertain, behaviour is complex or the team needs to discover unanticipated needs.

Positivist and quantitative research seeks to measure variables, estimate prevalence, compare groups or test relationships. It is useful when the team has defined constructs and needs to assess magnitude, distribution or confidence in a pattern.

These traditions should not be treated as competing ideologies. A product team may use qualitative interviews to discover the dimensions of a problem, observation to understand actual behaviour, and a survey to estimate how common the identified pattern is.

Method selection matrix

A practical sequence is often exploratory qualitative work, focused measurement and then targeted validation. However, sequence depends on the decision. When a regulatory deadline or market-size estimate is central, secondary and quantitative research may come first.

3. Designing the Research Plan

A research plan is a decision document, not merely a schedule. It should contain:

1. Decision context: What product decision is pending?
2. Research objectives: What must be learned?
3. Research questions: What specific questions will guide collection?
4. Target population: Whose experience is relevant?
5. Method and rationale: Why is each method suitable?
6. Sampling and recruitment: Who will participate and how?
7. Protocol: What will researchers ask, observe or record?
8. Analysis plan: How will data be organised and interpreted?
9. Ethics and data management: What consent, privacy and security controls apply?
10. Decision rules: What evidence would support continuation, revision or abandonment?

Assumptions and hypotheses

Product teams should distinguish an assumption from a hypothesis. An assumption is an unverified belief, such as “small retailers need real-time inventory alerts.” A hypothesis makes the belief researchable by specifying a population, context and expected pattern: “Among independent retailers with more than one sales channel, stock discrepancies create weekly operational work and lead managers to use manual reconciliation.”

A discovery hypothesis need not predict a causal effect. It may simply state a pattern to investigate. Teams should record confidence and consequence. A low-confidence, high-consequence assumption deserves early research.

Interview guide design

A semi-structured guide combines consistency with flexibility. A strong guide usually moves from broad experience to specific episodes:

• “Tell me about the last time you encountered this situation.”
• “What happened next?”
• “How did you decide what to do?”
• “What alternatives did you consider?”
• “What was difficult or costly?”
• “Who else was involved?”
• “How do you handle this when the usual approach fails?”

Questions should be neutral, singular and behaviour-focused. Avoid leading questions (“How useful would our automation be?”), double-barrelled questions (“Was it fast and accurate?”), vague abstractions (“What do you value?”) and hypothetical commitment (“Would you buy this?”). Probes should clarify rather than persuade.

A useful interview structure is:

The interviewer should listen more than speak. Silence, reflective summaries and non-directive probes often produce better evidence than rapid questioning.

4. Observation and Contextual Inquiry

Interviews reveal what people remember, believe and choose to report. Observation reveals what they do, including tacit routines and workarounds they may not consider noteworthy. The difference between reported and observed behaviour is not necessarily dishonesty; people forget, normalise effort and reconstruct events after the fact.

Observation may be structured, with a predefined coding scheme, or unstructured, with open field notes. The researcher records actions, tools, sequence, interruptions, physical or digital environment, social interactions and exceptions.

Contextual inquiry is a partnership in which the researcher observes work in its natural setting while asking occasional questions about what is happening. It combines observation with interpretation, but the researcher must avoid taking over the task or turning the session into a conventional interview.

A contextual inquiry protocol should specify:

• the activity and setting to be observed;
• what may be recorded;
• how interruptions will be handled;
• privacy boundaries and sensitive information;
• questions for clarification;
• how the participant can pause or stop the session.

Field notes should separate observation from interpretation. For example:

• Observation: “The coordinator opened three browser tabs, copied an order number and switched between systems four times.”
• Interpretation: “The workflow may involve costly cross-system reconciliation.”
• Follow-up question: “What information are you trying to confirm across these systems?”

This separation reduces premature theorising.

5. Diary Studies and Longitudinal Evidence

A diary study asks participants to record experiences close to the time they occur, often over several days or weeks. It is suitable when behaviour is episodic, distributed over time or affected by changing conditions. Examples include recording delivery failures, medication-management difficulties, commuting disruptions or decisions made during a procurement cycle.

Diary prompts should be brief, specific and timed appropriately. A prompt might ask:

1. What happened?
2. What were you trying to accomplish?
3. What did you do?
4. What made the situation easier or harder?
5. What was the consequence?

Researchers should pilot the diary, limit participant burden, provide reminders and offer multiple accessible formats where appropriate. Diary entries are not automatically objective: participants choose what to record, may alter behaviour because they are being studied and may discontinue participation. Their value lies in capturing episodes and variation that a single interview may miss.

6. Survey Research

Surveys are useful for measuring frequency, distribution, satisfaction, perceived importance, segment differences and associations. They are weak tools for discovering entirely unknown problems or predicting actual purchase behaviour from stated intention alone.

Construct and question design

A construct is an abstract concept being measured, such as perceived effort, trust or frequency of interruption. Researchers should define it operationally before writing questions. A question should measure one construct at a time and provide response options that are exhaustive and mutually exclusive where possible.

Poor question: “How satisfied are you with the speed and reliability of the service?”

Better questions:

• “How satisfied are you with the time required to complete the task?”
• “How often does the service fail during the task?”

Response scales need balanced wording, clear anchors and a meaningful recall period. “In the past 30 days” is more interpretable than “usually.” Questions about sensitive behaviour should minimise judgement and explain why the information is requested.

Sampling and survey precision

A probability sample gives each eligible population member a known, non-zero selection probability. It supports statistical generalisation when coverage and response conditions are adequate. A convenience sample is easier but may be systematically different from the target population.

For a proportion estimated from a simple random sample, an approximate margin of error at confidence level associated with critical value z is:

Here, E is the margin of error, p is the estimated population proportion, n is the sample size and z is the relevant standard-normal critical value. This approximation assumes a probability sample, independent observations and an adequately large sample for the normal approximation. For the most conservative planning value, researchers often use p = 0.5, because it produces the largest variance. It does not correct for non-response bias or a flawed sampling frame.

For example, with p = 0.5, n = 400 and z ≈ 1.96, the approximate margin of error is:

Thus the margin of error is approximately 4.9 percentage points under the stated assumptions. This is not a guarantee that the estimate is within 4.9 points if the sample is biased or the response rate is poor.

Product discovery surveys frequently use non-probability samples. In that case, numerical percentages should be described as results among respondents, not automatically as estimates of the entire market. Sample size cannot repair systematic coverage error.

7. Secondary Research

Secondary research uses existing material, including academic studies, government statistics, industry reports, company records, support tickets, product analytics, reviews, regulatory documents and competitor information. It is often the fastest way to understand market structure and avoid repeating prior work.

A source should be assessed for:

• provenance and author expertise;
• date and relevance to the current context;
• definitions and measurement method;
• population and sampling design;
• incentives or conflicts of interest;
• transparency of limitations;
• consistency with other sources.

Internal data can reveal behavioural scale but rarely explains motivation. App-store reviews may reveal recurring pain points but overrepresent highly satisfied or dissatisfied users. Industry reports may provide useful market context but use commercial definitions or undisclosed methods. Triangulation is therefore essential.

8. Sampling and Recruitment

Sampling is a substantive design decision because the people included determine which realities become visible. Define the population using relevant characteristics: role, usage, experience, geography, organisation size, accessibility needs or exposure to the problem.

Common strategies include:

• Purposive sampling: selecting people with relevant experience.
• Criterion sampling: requiring all participants to meet defined criteria.
• Maximum-variation sampling: seeking contrasting contexts to explore range.
• Homogeneous sampling: focusing on a narrowly comparable group.
• Quota sampling: filling predefined category counts without necessarily using random selection.
• Stratified random sampling: randomly sampling within defined population strata.
• Snowball sampling: recruiting through participant referrals, useful for hard-to-reach populations but vulnerable to network bias.

Qualitative sample adequacy is not determined by a universal number. Researchers should consider information power: the narrower the aim, more specific the sample, stronger the interview dialogue and more established the theory, the fewer participants may be needed. Researchers should monitor whether new sessions continue to add materially different insights. “Saturation” should not be used casually; it should specify whether the goal is thematic saturation, meaning saturation, code saturation or understanding of variation.

Recruitment incentives should compensate time without becoming coercive. Exclusion criteria should be justified rather than used merely for convenience. Recruiting only existing enthusiastic users produces a narrow account of need and may hide barriers faced by non-users or former users.

9. Bias, Validity and Researcher Reflexivity

Bias is systematic distortion, not simply random error. Important forms include:

• Selection bias: some relevant people are more likely to participate.
• Non-response bias: those who do not respond differ from respondents.
• Recall bias: participants inaccurately remember past events.
• Social-desirability bias: participants report what appears acceptable.
• Confirmation bias: researchers seek evidence supporting an existing belief.
• Leading-question bias: wording encourages a particular answer.
• Hawthorne effect: behaviour changes because people know they are observed.
• Survivorship bias: research includes successful users but excludes those who abandoned or failed.
• Availability bias: vivid or recent examples receive excessive weight.
• Sponsor bias: funding or organisational expectations influence collection or interpretation.

Bias control begins before fieldwork. Use neutral guides, pilot instruments, recruit varied participants, include disconfirming cases, record deviations and separate raw evidence from interpretation. During analysis, maintain a reflexive note describing the researcher’s role, assumptions and relationship to the product. Where feasible, have a second researcher review a sample of transcripts or coded evidence. Disagreement is not merely a problem; it can reveal ambiguous concepts or unsupported interpretations.

Validity should be discussed precisely. Construct validity asks whether the method measures the intended concept. Internal validity concerns the credibility of an explanation within the studied setting. External validity or transferability concerns whether learning applies elsewhere. A discovery study may have strong contextual insight without supporting population-wide prevalence claims.

10. Triangulation and Evidence Synthesis

Triangulation compares evidence across methods, sources, researchers or settings. It does not mean that all methods must produce identical results. Apparent disagreement may reflect different aspects of a phenomenon.

For example, interviews may show that users describe reporting as frustrating, observation may show repeated manual copying, and usage data may show frequent export activity. Convergence strengthens the case that the workflow is important. If a survey reports low concern but observation shows severe workarounds, the team should investigate wording, sample composition, frequency and the difference between stated importance and tolerated behaviour.

A practical evidence matrix can be structured as follows:

Confidence should reflect evidence quality and limitations, not the number of quotations. An attractive narrative is not necessarily a reliable conclusion.

From evidence to opportunity

Researchers should preserve a traceable chain:

Raw evidence → coded excerpt → pattern → insight → opportunity → decision.

A good opportunity statement identifies a population, situation, unmet need and consequence without prescribing a solution. For example: “When independent retailers receive stock updates from multiple channels, they need a dependable way to identify discrepancies before replenishment decisions, because manual reconciliation consumes staff time and creates avoidable stock-outs.”

This formulation supports several possible solutions and keeps the problem distinct from the product concept.

11. Research Ethics and Data Governance

Ethical research treats participants as people with rights, not merely sources of product input. Core principles include respect for autonomy, informed consent, minimisation of harm, fairness and responsible stewardship of data.

Participants should understand the purpose of the study, what participation involves, recording arrangements, foreseeable risks, voluntary nature of participation, compensation and withdrawal process. Consent must be understandable and appropriate to the population. A participant should not be required to disclose sensitive information unrelated to the research question.

Privacy requires more than removing names. Researchers should consider indirect identifiers, small groups, screen contents, voice recordings, location, employment information and combinations of seemingly harmless data. Collect only what is necessary, restrict access, define retention periods and securely delete data when no longer needed. If research involves children, employees in dependent relationships, patients or other vulnerable groups, additional safeguards and institutional review may be required.

Product managers should distinguish research from marketing. A research invitation should not conceal a sales purpose. Participants should not be pressured to approve a concept, and negative evidence must be reported fairly. When AI tools transcribe, summarise or code research data, teams must assess vendor terms, confidentiality, data residency, human review, hallucination risk and whether identifiable data may be used for model training. AI output is an aid to analysis, not a substitute for researcher judgement or participant consent.

Applicable laws vary by jurisdiction. Privacy and data-protection requirements may include lawful processing, purpose limitation, data minimisation, access rights and breach controls. Organisational policies and sector-specific rules may impose stricter standards. Legal review is appropriate when handling health, financial, biometric, children’s or employee data.

12. Analysis and Communication

Qualitative analysis commonly involves familiarisation, coding, categorisation, comparison and interpretation. Codes should remain close to the data initially, then develop into higher-level themes. Researchers should retain negative cases and distinguish frequency from importance. A rare event may be strategically critical, while a frequently mentioned minor irritation may not justify product investment.

Quantitative analysis should begin with data quality checks: missing responses, inconsistent values, duplicate submissions, outliers and invalid skip patterns. Descriptive statistics should match the measurement scale. Percentages should state their denominator; subgroup comparisons should disclose sample sizes. Correlation should not be presented as causation without an appropriate design and assumptions.

A research readout should state:

1. What was studied and why.
2. Who participated and who was not represented.
3. What was observed or measured.
4. How the evidence was analysed.
5. What is known, uncertain and contradicted.
6. What decision or experiment follows.

Use direct quotations sparingly and with context. A quotation illustrates a theme; it does not prove prevalence. Avoid presenting a “customer voice” as if all customers share one view.

13. Applied Case: Improving a Small-Business Cash-Flow Product

The following is an illustrative case, created for educational purposes.

Background and problem

A fintech team believed that small-business owners needed an AI cash-flow forecasting feature. The proposed feature had received enthusiastic reactions in internal workshops, but the team lacked evidence about the underlying problem, current workarounds and willingness to use automated forecasts.

Research design

The team defined the decision as: “Should we invest in automated forecasting, and for which small-business segment?” It used four methods over four weeks:

• 18 semi-structured interviews with owners and bookkeepers;
• contextual inquiry with 8 participants during weekly finance administration;
• a 14-day diary study with 10 participants who managed irregular income;
• a survey of 312 respondents recruited through an accounting association and product mailing list.

The sample deliberately included current users, former users and non-users of the company’s product. The team also reviewed support tickets, public reviews and relevant small-business finance research.

Analysis

Interviews suggested that participants spoke about “forecasting,” but their practices differed. Some wanted to anticipate tax obligations; others wanted to identify whether payroll could be met. Observation showed that many did not build formal forecasts. Instead, they checked bank balances, outstanding invoices and a spreadsheet maintained by an accountant. Diary entries revealed that uncertainty became acute after delayed customer payments, not continuously.

Survey results showed that 61% of respondents had experienced at least one cash shortfall concern in the preceding six months. Because the sample was non-probability and partly drawn from an existing mailing list, the team reported this as “61% of respondents,” not as a population estimate. Among respondents with employees, the concern was more frequent than among sole traders, but the team treated this as an exploratory subgroup pattern rather than a definitive market estimate.

Triangulation supported a narrower opportunity: owners with employees and irregular invoice payment needed early warnings linked to payroll and tax dates. Evidence did not support the assumption that every small business needed a continuously updated AI forecast.

Approach and outcome

The team changed the concept from a general forecasting dashboard to a scenario-based cash-commitment alert. It designed a prototype that allowed users to enter expected invoice dates, payroll commitments and tax obligations, then displayed uncertainty and requested confirmation rather than presenting a falsely precise prediction.

A subsequent usability test identified confusing terminology and showed that users wanted to see the assumptions behind an alert. The team postponed model automation until it could establish data quality and user trust.

Learning points

1. Customers’ solution language may conceal different underlying jobs.
2. Contextual inquiry exposed workarounds that interviews alone did not reveal.
3. A survey helped estimate the pattern within respondents but did not remove sampling limitations.
4. Segmentation changed the opportunity more than the original feature idea did.
5. Trust required transparent assumptions, not merely better prediction claims.

14. Best-Practice Checklist

Before fieldwork, confirm that the team has a decision, researchable questions, a justified sample, a tested protocol, consent materials and a data-management plan. During fieldwork, protect neutrality, record context, seek disconfirming evidence and monitor sample gaps. During analysis, distinguish observation from interpretation, preserve contradictory cases and document the path from evidence to decision. After reporting, state limitations, assign ownership for follow-up and store data according to approved retention rules.

The most mature product teams do not ask whether research “proved” an idea. They ask what uncertainty has been reduced, what remains unknown, whose experience is missing and what low-cost action will produce the next useful evidence.

Chapter Summary

Research design turns product uncertainty into a structured programme of inquiry. Interviews reveal experiences, meanings and decision processes; observation and contextual inquiry reveal situated behaviour; diary studies capture events over time; surveys measure patterns when constructs and populations are sufficiently defined; and secondary research supplies existing market, regulatory and organisational evidence.

Rigorous discovery aligns research questions, methods, sampling, analysis and product decisions. Sampling must reflect the population relevant to the decision, while qualitative adequacy depends on information power and variation rather than a universal participant count. Bias can arise from recruitment, memory, question wording, researcher expectations and organisational incentives. Neutral protocols, reflexivity, disconfirming cases and transparent limitations improve credibility.

Triangulation compares evidence across methods and sources. It strengthens confidence when evidence converges and creates valuable questions when it conflicts. Ethical research requires informed consent, privacy protection, data minimisation, fair treatment and appropriate governance of recordings, sensitive information and AI-assisted analysis. The ultimate purpose of research is not to produce attractive reports but to create a traceable path from customer evidence to better product decisions.

Key Terms

Contextual inquiry: Research conducted in a participant’s natural setting while combining observation with limited clarification.

Construct: An abstract concept, such as trust or effort, that must be operationally defined for measurement.

Diary study: A longitudinal method in which participants record experiences near the time they occur.

External validity: The extent to which findings can reasonably apply beyond the studied participants or setting.

Finding: A meaningful pattern or observation derived from collected data.

Information power: The idea that sample adequacy depends on the specificity and quality of information collected, not only sample size.

Insight: An explanatory interpretation of why a finding matters.

Interview guide: A structured but flexible set of topics and questions used to conduct interviews consistently.

Non-response bias: Distortion caused when people who do not participate differ systematically from respondents.

Observation: Systematic attention to what people do in a relevant setting.

Purposive sampling: Selecting participants because they possess characteristics or experience relevant to the research question.

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

Quantitative research: Research focused on measurement, distributions, comparisons or relationships among variables.

Research hypothesis: A specific, researchable proposition about a population, context or expected pattern.

Sampling frame: The operational list or source from which a sample is selected.

Secondary research: Analysis of evidence originally collected or published for another purpose.

Selection bias: Systematic distortion caused by unequal inclusion of relevant people or cases.

Triangulation: Comparison of evidence across methods, sources, researchers or settings.

Validity: The credibility and appropriateness of an inference or measurement.

Workaround: An improvised behaviour or tool used to overcome a limitation in an existing process or product.

Review Questions

Short Answer Questions

1. Distinguish data, finding, insight and decision in product discovery.
2. What is the principal difference between observation and contextual inquiry?
3. Why should a product team distinguish a non-probability survey result from a population estimate?
4. Name four common forms of research bias.
5. What are the central ethical requirements for recording a research interview?

Long Answer Questions

1. Explain how a product manager should move from a strategic decision to a complete research plan.
2. Compare interviews, diary studies and surveys in terms of their purposes, strengths and limitations.
3. Discuss how sampling affects the credibility and transferability of discovery research.
4. Explain triangulation and show how apparently conflicting evidence can improve product understanding.
5. Evaluate the ethical and governance issues involved in using AI to transcribe and analyse customer research.

Analytical / Application Questions

1. A team claims that 72% of customers need a new feature after surveying 150 people recruited from its most active users. Analyse the claim’s limitations and propose a stronger research design.
2. Interviews suggest that users want “faster reporting,” while observation shows that they spend most of their time collecting data from multiple systems. Formulate an opportunity statement and identify two follow-up research questions.
3. Design a mixed-method discovery study for a service used irregularly during high-stress events. Justify the sample, methods, bias controls, ethical safeguards and decision rules.

Multiple Choice Questions with Answers

Question 1

Which sequence best represents the movement from raw research material to a product decision?

A. Decision → insight → raw evidence → finding

B. Raw evidence → coded pattern → insight → decision

C. Insight → sample → raw evidence → decision

D. Finding → recruitment → hypothesis → raw evidence

Correct Answer: B. Raw evidence → coded pattern → insight → decision

Question 2

Which method is most appropriate for understanding how a participant actually performs work across several systems in a natural setting?

A. Contextual inquiry

B. Market-size survey

C. Concept preference poll

D. Desk research only

Correct Answer: A. Contextual inquiry

Question 3

A survey uses a convenience sample of existing enthusiastic users. What is the main methodological concern?

A. The sample is necessarily too large

B. The sample may exclude dissatisfied or non-adopting users

C. The survey cannot contain closed questions

D. Convenience samples always have zero useful information

Correct Answer: B. The sample may exclude dissatisfied or non-adopting users

Question 4

In the margin-of-error formula E = z √(p(1 − p)/n), what does n represent?

A. The confidence level

B. The estimated population proportion

C. The sample size

D. The critical value

Correct Answer: C. The sample size

Question 5

Which interview question is most behaviour-focused?

A. Would you like an automated solution?

B. How valuable is innovation to you?

C. Tell me about the last time you handled this situation.

D. Do you agree that the current process is inefficient?

Correct Answer: C. Tell me about the last time you handled this situation.

Question 6

What is a principal advantage of a diary study?

A. It eliminates all participant bias

B. It captures experiences close to when they occur over time

C. It guarantees a probability sample

D. It requires no participant effort

Correct Answer: B. It captures experiences close to when they occur over time

Question 7

What does triangulation primarily involve?

A. Repeating one leading question many times

B. Comparing evidence across methods or sources

C. Replacing qualitative research with a large survey

D. Selecting only evidence that supports the product concept

Correct Answer: B. Comparing evidence across methods or sources

Question 8

Which practice best reduces confirmation bias during analysis?

A. Ignoring contradictory cases

B. Using only quotations from senior customers

C. Searching deliberately for disconfirming evidence

D. Allowing the sponsor to rewrite all findings

Correct Answer: C. Searching deliberately for disconfirming evidence

Question 9

If p = 0.5, n = 400 and z = 1.96, the approximate margin of error is:

A. 0.49 percentage points

B. 4.9 percentage points

C. 19.6 percentage points

D. 49 percentage points

Correct Answer: B. 4.9 percentage points

Question 10

Which statement reflects responsible research data governance?

A. Collect all available data in case it becomes useful

B. Share identifiable recordings freely within the organisation

C. Collect necessary data, restrict access and define retention limits

D. Use AI summaries as a substitute for checking source evidence

Correct Answer: C. Collect necessary data, restrict access and define retention limits