3. Marketing Research and Customer Insight

Introduction and Context

Insurance marketing decisions are made in conditions of uncertainty. Customers may not fully understand the risks they face, the products available to them, or the implications of policy exclusions. Insurers, meanwhile, must design products before claims experience is fully known, communicate intangible benefits, select suitable distribution channels, and maintain trust over a relationship that may last for many years. Marketing research provides the evidence required to make these decisions more intelligently.

Marketing research in insurance is not limited to measuring satisfaction or asking customers whether they like an advertisement. It is a systematic process of defining a decision problem, collecting relevant information, analysing evidence and converting findings into action. Customer insight goes a step further. It involves explaining why customers behave as they do, what unmet needs or anxieties shape their decisions, and how the organisation can respond profitably and responsibly.

The distinction is important. Data may show that policy renewal has declined among young urban customers. Research investigates whether the cause is price, poor digital experience, inadequate product relevance, low trust, competitor activity or a combination of factors. Insight connects this explanation to a managerial response, such as redesigning communication, simplifying renewal, offering flexible coverage or improving service recovery.

This chapter examines qualitative and quantitative research, market surveys, voice-of-customer programmes, sampling, competitive intelligence, data interpretation and insight validation. It also considers ethical and regulatory responsibilities, digital research tools and the use of analytics and artificial intelligence in insurance marketing.

Learning Objectives

After studying this chapter, students should be able to:

1. Explain the role of marketing research and customer insight in insurance decision-making.

2. Distinguish exploratory, descriptive and causal research designs.

3. Design appropriate qualitative and quantitative research studies.

4. Apply sampling principles and assess the quality of survey evidence.

5. Develop voice-of-customer and competitive-intelligence programmes.

6. Interpret customer data without confusing correlation with causation.

7. Validate insights through triangulation, experimentation and managerial judgement.

8. Recognise ethical, privacy, governance and operational risks in insurance research.

From Market Information to Actionable Insight

Marketing research is a structured investigation undertaken to support a marketing decision. Its outputs may include estimates of market size, customer segments, purchase drivers, channel preferences, brand perceptions, satisfaction levels or responses to a proposed product. Customer insight is the meaningful interpretation of such evidence in relation to customer behaviour and organisational action.

A useful distinction can be made among four levels of information:

• Data are recorded observations, such as claim counts, survey responses or website visits.
• Information is organised data, such as the renewal rate for customers aged between 25 and 35.
• Finding is an observed pattern, such as lower renewal among customers who received delayed service.
• Insight is an explanation with managerial significance, such as the conclusion that renewal declines because customers interpret delayed service as evidence that the insurer will be difficult to deal with during a claim.

Insight is therefore not simply an attractive statement. It should be evidence-based, relevant to a decision, sufficiently specific to guide action and capable of being tested. A strong insight often combines an observed behaviour, an underlying motivation and an implication for the organisation.

Insurance research must also recognise the special nature of the product. Insurance is intangible, purchased partly for future protection, and often evaluated only when a claim occurs. Customers may have limited financial literacy and may use intermediaries rather than dealing directly with the insurer. Research must therefore investigate not only buying behaviour but also comprehension, trust, perceived fairness, advice quality, claims experience and willingness to disclose information.

Defining the Research Problem

The quality of a research project depends heavily on the quality of the initial problem definition. A management problem is expressed in terms of action: for example, “How can motor policy renewal be improved?” A research problem translates this into questions that evidence can answer: “What factors explain non-renewal, how important is each factor, and do these factors differ by customer segment and channel?”

A practical problem-definition sequence is:

1. Identify the managerial decision and the alternatives under consideration.

2. Clarify what is already known from internal data and previous research.

3. Identify information gaps and competing explanations.

4. Define the population, time period and business context.

5. Formulate research objectives and, where appropriate, hypotheses.

6. Specify how findings will influence a decision.

Research objectives should be precise. “Understand customer needs” is too broad. A stronger objective would be: “To identify the main barriers preventing first-time buyers from purchasing health insurance through digital channels and to estimate the relative importance of those barriers.”

Research Designs and Methodological Choices

Exploratory research

Exploratory research is appropriate when the problem is unclear or when the organisation lacks a sufficiently developed understanding of customer motivations. It is commonly used before designing a questionnaire or developing a new product. Methods include literature review, analysis of complaints, interviews with employees and intermediaries, focus groups, observation, social-listening analysis and review of call-centre transcripts.

Exploratory research is flexible and rich in meaning, but it normally does not provide statistically representative estimates. A few interviews can reveal an important issue, but they cannot establish how common that issue is in the total market.

Descriptive research

Descriptive research measures characteristics of a market or population. It may estimate awareness, consideration, satisfaction, channel usage, price sensitivity, policy ownership or intention to renew. Structured surveys, customer databases and panel studies are common sources.

Descriptive research requires clear operational definitions. For example, “customer satisfaction” could mean overall satisfaction, satisfaction with the latest interaction, satisfaction with claims settlement or willingness to recommend. These are related but not identical constructs.

Causal research

Causal research examines whether a change in one factor produces a change in another. In insurance marketing, an insurer may test whether a simplified renewal message increases completion, whether a claims-status notification reduces inbound calls, or whether a revised premium presentation improves conversion.

Randomised controlled experiments provide stronger causal evidence than simple comparisons. If customers are randomly assigned to a control group and a treatment group, differences in outcomes can more confidently be attributed to the intervention, assuming the experiment is properly implemented. Quasi-experimental designs, such as difference-in-differences or matched comparisons, may be used where randomisation is impractical.

Mixed-method research

Mixed-method research combines qualitative and quantitative approaches. A common sequence is qualitative exploration, quantitative measurement and a final qualitative or experimental stage for validation. For example, interviews may identify confusion about health-policy exclusions; a survey may measure its prevalence; and a message experiment may test whether clearer explanation improves purchase confidence.

Qualitative Methods for Insurance Insight

Qualitative research seeks depth, context and meaning. It is especially useful when customers find it difficult to articulate needs directly, when emotional factors are important, or when the organisation needs to understand the customer journey.

In-depth interviews allow a researcher to explore personal experiences, fears, financial priorities and interpretations of insurance language. Interviews are useful for sensitive subjects such as health, death, debt and past claims. The interviewer should use a discussion guide without turning the conversation into a rigid questionnaire. Probing questions such as “What happened next?” or “What did that phrase mean to you?” often reveal more than direct preference questions.

Focus groups generate interaction among participants. They can be useful for testing advertising concepts, exploring language and comparing reactions to product ideas. However, group dynamics may suppress minority views, encourage conformity or allow articulate participants to dominate. Sensitive financial subjects may therefore require individual interviews instead.

Ethnographic and observational methods examine behaviour in context. Observing how an intermediary explains a policy, how a customer completes a digital proposal form or how a family discusses health protection may reveal difficulties that respondents do not mention in a survey.

Customer-journey research follows stages such as need recognition, information search, comparison, purchase, onboarding, policy servicing, claim and renewal. It identifies moments of truth, effort, anxiety and abandonment. Journey maps should be based on evidence rather than assumptions and should distinguish the customer’s experience from the organisation’s internal process.

Qualitative analysis usually involves coding transcripts, identifying recurring themes, comparing segments and developing explanatory interpretations. Researchers should preserve contradictory evidence rather than selecting only quotations that support a preferred conclusion.

Quantitative Research and Survey Design

Quantitative research converts responses or observations into numerical evidence. Its quality depends on construct definition, measurement design, sampling, fieldwork and analysis.

A questionnaire should have a clear sequence. It normally begins with screening questions, proceeds to behaviour and experience, then attitudes and evaluations, and ends with demographic or classification questions. Questions should be concise, neutral and relevant. Double-barrelled questions such as “How satisfied are you with the price and service?” should be separated because a respondent may hold different opinions about price and service.

Response scales must match the construct. A five-point agreement scale can measure attitudes, while a frequency scale is more suitable for behaviour. Balanced scales should include both positive and negative options where appropriate. “Not applicable” and “Don’t know” options may be necessary, particularly when customers lack knowledge of policy terms.

Pretesting is essential. Cognitive interviewing asks respondents to explain how they understood a question and selected an answer. Pilot testing can identify excessive completion time, confusing instructions, missing response options and technical problems.

Important measures in insurance research include:

• Awareness: whether the customer recognises a brand or product.
• Consideration: whether the customer would include the brand in a purchase decision.
• Conversion: the proportion of qualified prospects who purchase.
• Customer effort: the perceived difficulty of completing a task.
• Satisfaction: the evaluation of an experience against expectations.
• Retention or renewal: continued ownership or continuation of a policy.
• Advocacy: willingness to recommend the insurer.

These measures should not be treated as interchangeable. A customer may be satisfied with a claim outcome but unwilling to recommend the insurer because premiums are high. Similarly, stated purchase intention may not translate into actual purchase.

Sampling: Representativeness and Practical Judgement

A population is the complete group about which the organisation wishes to draw conclusions. A sampling frame is the operational list or source from which participants are selected. The sample is the subset that provides data.

Probability sampling gives members of the population a known chance of selection. Simple random sampling, systematic sampling, stratified sampling and cluster sampling are examples. Stratification is especially useful when important groups, such as product types, regions or distribution channels, differ substantially and must be represented.

Non-probability methods include convenience, purposive, quota and snowball sampling. They may be suitable for exploratory research or hard-to-reach groups, but their findings should not automatically be generalised to the full market.

Sample size is influenced by the required precision, population variability, study design, subgroup analysis and available resources. A larger sample does not correct a biased sampling frame. A very precise estimate from respondents who systematically exclude digitally inactive or low-literacy customers may still be misleading.

Researchers must consider non-response bias, coverage error, self-selection, interviewer effects and weighting. Weighting can improve alignment with known population characteristics, but it cannot fully repair poor measurement or severe undercoverage. Findings should therefore report not only sample size but also recruitment method, response rate, field dates, screening criteria and important limitations.

Voice of the Customer as an Operating System

A voice-of-customer (VoC) programme is a continuing process for collecting, integrating and acting on customer feedback. It should connect customer evidence with operational ownership rather than operate as an isolated survey exercise.

A mature VoC programme combines solicited feedback, such as post-interaction surveys, with unsolicited feedback, such as complaints, reviews, emails, call recordings, intermediary comments and social-media conversations. Transactional feedback captures a specific event, while relationship research assesses the broader relationship with the insurer.

The programme should define closed-loop procedures. A negative claim experience, for example, may trigger immediate contact by a service team, investigation of the underlying process and escalation when vulnerability or potential misconduct is involved. At aggregate level, recurring themes should be assigned to business owners, prioritised by customer impact and tracked until improvement is demonstrated.

VoC measures are useful only when interpreted with operational data. A fall in satisfaction alongside a rise in claim settlement time is more informative than either measure alone. Text analytics can classify complaints, but human review remains necessary for ambiguity, irony, vulnerability and emerging issues.

Competitive Intelligence in Insurance

Competitive intelligence is the systematic collection and interpretation of information about competitors, substitutes, market structures and external forces. It is not industrial espionage. Legitimate sources include public product documents, price observations, regulatory filings, advertisements, websites, customer reviews, intermediary feedback, recruitment information, analyst reports and mystery shopping conducted ethically.

A competitive-intelligence process begins by identifying decisions requiring external evidence. These may concern product features, exclusions, pricing presentation, digital journeys, claims promises, partnerships or distribution models. Information should then be assessed for source reliability, recency, comparability and possible bias.

A competitor comparison matrix can map dimensions such as coverage, exclusions, service access, digital functionality, advice, claims communication and perceived trust. The purpose is not to copy competitors but to identify points of parity, points of difference and areas where the market is failing customers.

Competitive intelligence can be weakened by superficial benchmarking. A competitor’s apparently low premium may reflect narrower coverage, higher deductibles or different underwriting assumptions. Comparisons must therefore use equivalent coverage and clearly state assumptions.

Interpreting Evidence and Validating Insight

Research interpretation requires more than reporting percentages. Analysts should examine distributions, subgroup differences, confidence intervals where appropriate, effect sizes and the practical importance of findings. Statistical significance does not necessarily imply managerial significance, particularly in large samples.

Correlation does not establish causation. Customers who use digital channels may show higher retention, but this may be because they are younger, more financially confident or less likely to make claims. Multivariate analysis, controlled experiments and careful segmentation can help separate competing explanations.

Segmentation should be based on meaningful differences in needs or behaviour, not merely on attractive descriptive labels. A useful segment is identifiable, sufficiently substantial, reachable through a suitable channel, distinct in response and actionable for the insurer.

Insight validation can use triangulation: comparing interviews, surveys, behavioural data, complaints, claims records and experiments. If several independent sources point to the same explanation, confidence increases. Contradictions should be investigated rather than averaged away. A pilot intervention can test whether the proposed insight leads to the expected behavioural change.

A disciplined insight statement may follow this form: “For [customer group], [context or tension] leads to [behaviour] because [underlying reason]. Therefore, the insurer should [action], which can be evaluated through [measure].” This format links evidence, interpretation, action and evaluation.

Technology, Analytics and Artificial Intelligence

Digital tools have expanded the speed and scale of insurance research. Customer data platforms can combine policy, interaction and campaign information; dashboards can monitor journeys; web analytics can identify abandonment points; and text-mining tools can classify large volumes of complaints and reviews.

Predictive analytics can estimate renewal probability, propensity to respond or likelihood of service escalation. However, prediction is not the same as explanation. A model may identify customers likely to lapse without revealing whether price, service or changing life circumstances is responsible. Research remains necessary to understand motives and design appropriate interventions.

Generative artificial intelligence can assist with questionnaire drafting, transcript summarisation, coding suggestions and the generation of alternative communication concepts. Outputs require human review because models may omit minority voices, reproduce bias, fabricate interpretations or mishandle confidential information. Sensitive personal data should not be entered into unauthorised systems.

Responsible analytics requires data minimisation, access controls, audit trails, explainability proportionate to risk, bias testing and documented human accountability. In insurance, research and marketing practices must also comply with applicable data-protection, consumer-protection, insurance and communications requirements.

Ethical and Professional Responsibilities

Participants should understand the purpose of research, what participation involves, how information will be used and whether participation is voluntary. Consent must not be obtained through misleading claims or excessive pressure. Researchers should collect only information necessary for the stated purpose and protect confidentiality.

Special care is required when research involves health information, financial hardship, vulnerable customers, minors or people making decisions under stress. Incentives should be reasonable and should not compromise voluntary participation. Personal data used for marketing research should be separated, where appropriate, from decisions that could unfairly affect underwriting or claims.

Researchers must avoid leading questions, fabricated respondents, selective reporting and misuse of customer quotations. Governance should define retention periods, data access, vendor responsibilities, quality checks and procedures for correcting or deleting information where required. Trust is not merely an ethical ideal in insurance; it is a commercial asset that can be damaged by intrusive or unfair research.

Illustrative Case: Improving Health-Insurance Renewal

Background

A fictional insurer, Meridian Health Assurance, observed that renewal rates among customers who purchased individual health policies through its digital channel were lower than those among customers assisted by agents. Management initially suspected that digital customers were more price-sensitive.

Situation and Problem

Internal data showed that non-renewal was concentrated among customers who had not made a claim. The insurer planned to offer a discount but wanted evidence before changing prices. The research question was: “Why do digitally acquired customers fail to renew, and what intervention is most likely to improve continuation without encouraging unsuitable sales?”

Analysis

The research team first analysed policy, interaction and renewal data. It found that non-renewing customers had higher rates of incomplete profile updates and more visits to pages explaining exclusions. This was an association, not proof of cause.

Researchers conducted twenty-four individual interviews across renewing and non-renewing customers, including customers with and without claims. Participants often described the policy as difficult to interpret. Several believed that a policy with no claim had provided no value. Others were uncertain about waiting periods and assumed that renewal would automatically repeat previous coverage.

A survey of 1,200 eligible customers then measured the prevalence of these themes. Lack of understanding of coverage and perceived lack of value were stronger predictors of non-renewal than stated price dissatisfaction. The survey also showed that many customers preferred a short annual coverage review rather than a discount alone.

Approach and Solution

Meridian designed a controlled test. One group received the existing renewal notice. A second received a clearer notice showing covered risks, key exclusions, waiting-period reminders and a link to a personal coverage review. A third received the clearer notice plus an optional adviser call. The groups were randomly assigned, and renewal completion, call volume, complaints and cancellation after renewal were monitored.

The test was accompanied by compliance review to ensure that the message was accurate, balanced and not misleading. Customers were not told that all medical expenses would be covered, and the communication clearly distinguished information from advice.

Outcome

The illustrative test showed that clearer explanation improved renewal completion and reduced avoidable clarification calls. The adviser option was valued by customers with complex circumstances, but it was not necessary for every customer. The insurer therefore adopted a tiered approach: clear standard communication for all customers and human assistance for customers showing uncertainty or requesting advice.

Learning Points

The case demonstrates that a management assumption about price may conceal a comprehension and value-perception problem. It also illustrates the value of mixed methods, the distinction between association and causation, controlled testing, ethical communication and the need to evaluate both commercial and customer outcomes.

Challenges and Common Mistakes

Several errors recur in insurance research. Organisations may begin with a preferred solution and design research merely to justify it. They may ask broad questions without identifying the decision to be made, rely on convenience samples, use jargon, treat stated intention as behaviour, or report an average that hides important segment differences.

Other risks include survey fatigue, low response rates, overinterpretation of small qualitative samples, poorly designed incentives, inconsistent definitions across departments and dashboards that measure activity rather than outcomes. Customer feedback can also become operationally overwhelming when every comment is treated as equally urgent.

Good research manages these risks through clear objectives, transparent methods, pilot testing, appropriate sampling, triangulation, independent challenge and explicit decision rules. Research should be proportionate: not every decision requires a large national survey, but important strategic decisions deserve more than anecdote.

Current Developments and Best Practice

Insurance research is increasingly continuous rather than periodic. Organisations combine real-time behavioural signals with scheduled qualitative and quantitative studies. Journey analytics, digital experimentation and automated text classification support faster learning, while privacy-preserving methods seek to reduce unnecessary exposure of personal data.

The strongest practice is not simply technologically advanced. It is organised around a learning cycle: define the decision, gather diverse evidence, interpret it in context, test the proposed response, monitor outcomes and update the understanding. Customer insight should be shared across marketing, product, underwriting, claims, service, compliance and distribution. When insight remains within the research department, it rarely changes the customer experience.

Chapter Summary

Marketing research supplies systematic evidence for insurance decisions, while customer insight explains the motivations and tensions behind observed behaviour. Effective research begins with a well-defined management problem and may use exploratory, descriptive, causal or mixed-method designs. Qualitative methods provide depth; quantitative methods estimate prevalence and relationships; experiments provide stronger evidence about causal effects.

Sampling quality depends on population definition, frame coverage, selection method, non-response management and transparent reporting. Voice-of-customer programmes integrate surveys, complaints, interactions and unsolicited feedback into closed-loop improvement. Competitive intelligence uses legitimate external sources to understand market alternatives without confusing superficial comparison with strategic insight.

Interpretation requires attention to measurement quality, segmentation, effect size, causality and triangulation. Digital analytics and artificial intelligence can increase scale and speed but require human judgement, privacy protection, bias controls and governance. Ethical research protects participants, respects vulnerable customers and reports evidence honestly. The ultimate value of research lies in validated action that improves both customer outcomes and organisational performance.

Key Terms

Marketing research: A systematic process of defining a marketing problem, collecting and analysing evidence, and supporting decisions.

Customer insight: An evidence-based explanation of customer behaviour and motivation that guides action.

Exploratory research: Flexible research used to clarify an unclear problem and generate possible explanations.

Descriptive research: Research that measures characteristics, behaviours or attitudes within a population.

Causal research: Research designed to determine whether one factor produces a change in another.

Mixed-method research: A design combining qualitative and quantitative evidence.

Voice of the customer: A structured programme for collecting, integrating and acting on customer feedback.

Competitive intelligence: Systematic analysis of legitimate information about competitors and market forces.

Population: The complete group about which research conclusions are intended.

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

Probability sampling: Sampling in which members have a known chance of selection.

Non-probability sampling: Sampling in which selection probabilities are unknown or unequal.

Non-response bias: Bias caused when respondents differ systematically from those who do not respond.

Triangulation: Comparison of evidence from multiple methods or sources to strengthen interpretation.

Operational definition: A precise specification of how a concept will be measured.

Customer journey: The sequence of stages and interactions through which a customer experiences an insurer.

Effect size: The magnitude of a difference or relationship, considered independently of statistical significance.

Closed-loop feedback: A process in which customer feedback leads to individual response and organisational improvement.

Insight validation: Testing whether an interpretation is supported by evidence and leads to the expected outcome.

Review Questions

Short Answer Questions

1. Distinguish data, findings and customer insight in an insurance context.

2. What is the difference between exploratory and descriptive research?

3. Why is pretesting important in questionnaire design?

4. Define non-response bias and state one way to reduce it.

5. What is the purpose of a closed-loop voice-of-customer process?

Long Answer Questions

1. Explain how an insurer should move from a management problem to a clearly defined research problem.

2. Compare qualitative interviews, focus groups and customer-journey research for insurance applications.

3. Discuss the principles of sampling and explain why a large sample may still produce biased findings.

4. Explain how competitive intelligence can support product and distribution decisions while remaining ethical.

5. Discuss the role and limitations of analytics and artificial intelligence in customer insight programmes.

Analytical / Application Questions

1. An insurer finds that customers who receive more service calls are less likely to renew. Design a research programme to determine whether calls cause dissatisfaction or whether dissatisfied customers simply call more often.

2. A health insurer wants to test a simplified policy summary. Propose a mixed-method and experimental design, including measures of customer understanding, renewal and compliance risk.

3. A survey shows high satisfaction but complaints about delayed claims are increasing. Explain how you would investigate the apparent contradiction and decide what action to recommend.

22. Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes customer insight from raw customer data?

A. Insight is always collected through quantitative surveys

B. Insight explains behaviour and connects evidence to an actionable decision

C. Insight is a larger database containing more observations

D. Insight is limited to customer satisfaction scores

Correct Answer: B. Insight explains behaviour and connects evidence to an actionable decision

Question 2

An insurer is unsure why first-time buyers abandon its digital application. Which research approach is most appropriate initially?

A. Exploratory interviews and journey observation

B. A national pricing experiment without preliminary research

C. A renewal survey restricted to existing customers

D. A competitor ranking based only on advertising expenditure

Correct Answer: A. Exploratory interviews and journey observation

Question 3

Which situation provides the strongest evidence that a new renewal message caused an increase in renewal?

A. Renewal increased after the message was launched, with no comparison group

B. Customers who liked the message renewed at a higher rate

C. Randomly assigned treatment and control groups showed a meaningful renewal difference

D. Managers believed the message was clearer than the old one

Correct Answer: C. Randomly assigned treatment and control groups showed a meaningful renewal difference

Question 4

What is the principal purpose of stratified sampling?

A. To ensure important subgroups are represented in the sample

B. To eliminate the need for a sampling frame

C. To guarantee that every respondent gives the same answer

D. To replace data analysis with managerial judgement

Correct Answer: A. To ensure important subgroups are represented in the sample

Question 5

Which is an example of a double-barrelled survey question?

A. How often did you contact the insurer last year?

B. Which insurance products do you currently hold?

C. How satisfied are you with the clarity and speed of claims service?

D. Did you renew your motor policy this year?

Correct Answer: C. How satisfied are you with the clarity and speed of claims service?

Question 6

A competitor advertises a much lower motor premium than an insurer. What should the analyst do first?

A. Conclude that the competitor has a permanent cost advantage

B. Compare equivalent coverage, exclusions, deductibles and assumptions

C. Recommend matching the price immediately

D. Ignore the competitor because advertising is not research

Correct Answer: B. Compare equivalent coverage, exclusions, deductibles and assumptions

Question 7

Which practice best illustrates triangulation?

A. Repeating the same question three times in one survey

B. Combining interviews, complaint data, survey results and behavioural records

C. Using only the largest available sample

D. Selecting the most favourable customer quotation

Correct Answer: B. Combining interviews, complaint data, survey results and behavioural records

Question 8

Why can a predictive model identifying likely non-renewers be insufficient by itself?

A. Predictive models cannot use customer data

B. Prediction identifies likely behaviour but may not explain its underlying cause

C. All predictive models are legally prohibited

D. Non-renewal is never measurable

Correct Answer: B. Prediction identifies likely behaviour but may not explain its underlying cause

Question 9

Which action is most consistent with ethical insurance research?

A. Collecting health information unrelated to the research objective

B. Using misleading claims to increase participation

C. Explaining purpose, obtaining appropriate consent and protecting confidentiality

D. Sharing identifiable responses with all departments without restriction

Correct Answer: C. Explaining purpose, obtaining appropriate consent and protecting confidentiality

Question 10

In the illustrative health-insurance case, why was a controlled test used after interviews and a survey?

A. To replace all qualitative evidence

B. To test whether the proposed communication intervention produced the expected behaviour

C. To avoid compliance review

D. To estimate competitor market share

Correct Answer: B. To test whether the proposed communication intervention produced the expected behaviour