Consumer Insight Systems and Research Design

Introduction

Marketing decisions are often presented as requests for information: “How large is the market?”, “Why are customers leaving?”, or “Will a lower price increase demand?” Yet these requests are not automatically research problems. A research problem is a decision-relevant uncertainty that can be translated into answerable questions, observable constructs, defensible methods, and explicit decision rules.

Consumer psychology and insight mining therefore require more than collecting opinions or producing dashboards. They require a system that connects managerial decisions with theory, evidence, interpretation, and action. Such a system must distinguish symptoms from causes, describe consumers without stereotyping them, identify relationships without overstating causality, and make uncertainty visible.

This chapter develops that system. It explains how to frame decision problems, prepare hypotheses and insight briefs, select among exploratory, descriptive, causal, and predictive designs, combine primary and secondary evidence, evaluate qualitative and quantitative methods, and establish validity, reliability, sampling quality, and triangulation. The central principle is that methodological sophistication is valuable only when it improves the quality of a decision.

Learning Objectives

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

1. Translate a managerial problem into a researchable decision question.
2. Construct an insight brief containing objectives, constructs, hypotheses, methods, and decision criteria.
3. Compare exploratory, descriptive, causal, and predictive research designs.
4. Evaluate primary, secondary, qualitative, quantitative, and mixed-method evidence.
5. Assess measurement validity, reliability, sampling adequacy, and research quality.
6. Use triangulation to integrate different forms of evidence.
7. Identify ethical, governance, privacy, and artificial-intelligence risks in consumer research.

1. From Managerial Problems to Research Problems

A managerial problem concerns what an organization should do. A research problem concerns what must be learned to make that decision responsibly. The distinction is foundational.

For example, “sales have declined” is a symptom, not a research question. Possible underlying explanations include reduced awareness, distribution gaps, competitor switching, dissatisfaction, seasonal effects, price sensitivity, or measurement error. Each explanation implies different data and different action.

A useful translation sequence is:

Decision context → decision problem → information need → research question → constructs and measures → method → decision rule.

A decision context identifies the situation, stakeholders, constraints, and time horizon. The decision problem specifies the choice facing management. The information need identifies uncertainty that could change that choice. A research question states what evidence is required. Constructs are the abstract concepts to be studied, such as trust, perceived value, involvement, or switching intention. Measures operationalize those constructs. The method determines how evidence will be generated, and the decision rule specifies how findings will be used.

Example: Reframing a problem

Suppose a subscription service has experienced a rise in cancellations. The initial request is: “Find out why churn is increasing.” A stronger formulation is:

• Decision: Should the firm prioritize onboarding, service recovery, pricing, or product improvements?
• Research problem: Which identifiable experiences most strongly predict cancellation among recently acquired customers?
• Research questions:
• How do customers describe the first 30 days of use?
• Which service failures are most common and consequential?
• Does improved onboarding reduce cancellation relative to standard onboarding?
• Which customer segments show the greatest responsiveness to intervention?
• Decision rule: Implement an intervention if it produces a practically meaningful reduction in cancellation at an acceptable cost and does not damage customer satisfaction or equity.

This formulation avoids assuming that the answer is already known. It also separates discovery, measurement, causal testing, and prediction.

Good research questions

A strong question is:

• Relevant: its answer could influence a decision;
• Specific: its population, context, constructs, and time frame are clear;
• Answerable: suitable evidence can be obtained;
• Neutral: it does not presuppose a preferred result;
• Feasible: it fits budget, access, time, and ethical constraints;
• Consequential: alternative findings lead to different actions.

Questions such as “Do consumers like our brand?” are usually too broad. “Among category buyers in the target market, how does perceived reliability compare with the two leading alternatives, and which perceptions are associated with stated consideration?” is more researchable, although it still requires careful measurement and sampling.

2. Insight Briefs and Hypothesis Development

An insight brief is a compact research charter. It aligns executives, researchers, analysts, agencies, and data owners before data collection begins. A brief should not be a decorative project summary; it should expose assumptions and establish what evidence will count.

A robust insight brief normally includes:

Hypotheses

A hypothesis is a logically derived, empirically testable proposition. It is not merely a prediction or a statement of managerial preference. A useful hypothesis identifies the expected relationship, population, direction where justified, and relevant conditions.

For instance: “Among first-time users, perceived onboarding clarity is negatively associated with cancellation within 90 days.” This proposition identifies a population, a predictor, an outcome, and a direction. It does not by itself establish causality; an association may reflect satisfaction, customer quality, or other confounding factors.

Hypotheses can be:

• Descriptive: “A majority of category users recognize the brand.”
• Relational: “Perceived fairness is positively associated with trust.”
• Causal: “A transparent fee explanation increases completed applications.”
• Moderating: “The effect of transparency is stronger among low-trust consumers.”
• Mediating: “Service recovery increases retention through restored trust.”
• Predictive: “Usage frequency and unresolved complaints improve prediction of renewal.”

Hypotheses should be derived from consumer theory, prior evidence, observed patterns, or a clearly stated exploratory phase. Researchers should distinguish preregistered hypotheses from post hoc explanations developed after inspecting data. Both may be useful, but they have different evidential status.

3. Four Purposes of Marketing Research

The labels exploratory, descriptive, causal, and predictive refer to dominant purposes. A single project may use more than one, but the purposes should not be confused.

Exploratory research

Exploratory research investigates poorly understood meanings, motivations, practices, and problem structures. It is particularly useful when management does not know which variables matter or when existing categories do not reflect consumers’ lived experience.

Common methods include depth interviews, focus groups, ethnography, diary studies, social listening, open-ended surveys, and analysis of customer-service records. Its outputs include concepts, language, hypotheses, journey maps, and candidate mechanisms.

Exploration is not a weak version of quantitative research. Its standards are different: theoretical sensitivity, depth, reflexivity, systematic coding, and evidence of interpretive credibility. Small qualitative samples can be appropriate for discovering meaning, but they cannot estimate population prevalence without additional evidence.

Descriptive research

Descriptive research estimates distributions, profiles, frequencies, attitudes, behaviors, and trends. It answers questions such as who buys, how often they buy, which attributes they associate with brands, and how awareness varies by segment.

Surveys, panels, transactional data, observation, and structured content analysis are common sources. Descriptive research depends strongly on sampling, question wording, measurement quality, weighting, and time design. A precise estimate from a biased sample remains misleading.

Causal research

Causal research examines whether changing one factor produces a change in an outcome. Randomized experiments are the strongest design for causal identification because random assignment tends to balance observed and unobserved characteristics across conditions.

A basic between-groups experiment compares an intervention group with a control group. For a binary outcome, a simple difference in proportions is:

Here, p₁ is the outcome proportion in the treatment group and p₀ is the outcome proportion in the control group. The estimate is interpretable as a causal effect only when assignment is appropriately randomized, treatment conditions are implemented as intended, outcomes are measured consistently, and interference, attrition, and other design threats are adequately controlled.

If 18% of treated users renew and 15% of control users renew, the estimated absolute difference is 0.03, or 3 percentage points. This is not the same as a 3% relative increase. The relative increase is:

which equals 0.03 ÷ 0.15 = 0.20, or 20%, under this example. Both figures should be reported with context, sample sizes, uncertainty intervals where appropriate, and business relevance.

Quasi-experiments, such as interrupted time series, difference-in-differences, regression discontinuity, or matched comparisons, may be useful when randomization is infeasible. Their assumptions must be defended rather than implied. Correlation, before-and-after change, or model importance does not automatically establish causality.

Predictive research

Predictive research estimates the likelihood or level of a future or unobserved outcome. Examples include predicting churn, response, conversion, lifetime value, or category entry. Prediction focuses on out-of-sample performance, not necessarily on explaining why an outcome occurs.

A predictive model should be evaluated using data not used to fit it, ideally through a preplanned holdout set or cross-validation. Relevant metrics depend on the task. For classification, accuracy can be misleading with imbalanced outcomes. Precision, recall, specificity, sensitivity, area under the receiver operating characteristic curve, calibration, and expected business value may be more informative. For continuous outcomes, mean absolute error and root mean squared error are common, but interpretation must reflect the decision context.

Predictive performance can deteriorate because of data drift, policy changes, missingness, feedback loops, or leakage. A variable recorded after the outcome or influenced by the intervention cannot legitimately be used to claim prospective prediction.

4. Research Philosophies and Design Logic

Research design reflects assumptions about knowledge. A postpositivist orientation emphasizes measurement, comparison, uncertainty, and attempts to reduce bias. An interpretivist orientation emphasizes meaning, context, language, and the situated nature of consumer experience. Pragmatism asks which combination of methods best answers the decision problem.

These orientations need not be treated as mutually exclusive in applied marketing. A firm may first interpret how consumers understand a new service, then measure the prevalence of those meanings, and finally test an intervention. The key is methodological fit: the claims made must match the evidence produced.

A research design specifies more than a method. It includes:

• unit of analysis, such as person, household, transaction, review, or store;
• temporal structure, such as cross-sectional, longitudinal, cohort, or panel;
• setting, such as naturalistic, laboratory, digital, or field;
• intervention and comparison conditions, if causal;
• sampling and recruitment procedures;
• measurement and operational definitions;
• analysis plan and decision criteria.

A common error is to select a familiar method first, then retrofit the question. Graduate-level practice reverses that sequence.

5. Primary, Secondary, Qualitative, Quantitative, and Mixed Methods

Primary and secondary research

Primary research is collected for the current decision: interviews, surveys, experiments, observations, or new behavioral tracking. It can be tailored but is often expensive and slow.

Secondary research uses existing material, including government statistics, syndicated studies, prior internal research, CRM records, transaction logs, reviews, search data, and academic literature. It is efficient but may have mismatched definitions, unknown sampling procedures, outdated collection periods, or restricted access to metadata.

Secondary data should be audited for:

1. Provenance: Who collected it and for what purpose?
2. Population: Who is included or excluded?
3. Definitions: What exactly does each variable mean?
4. Time and context: When and under what conditions was it generated?
5. Missingness and errors: What is absent, duplicated, or inconsistently recorded?
6. Permission and rights: Is use lawful, ethical, and contractually permitted?

Qualitative research

Qualitative research is suited to meaning, process, identity, emotion, interaction, and context. Interviews can examine personal histories and sensitive experiences. Focus groups reveal social norms and disagreement, although dominant participants and group conformity can distort expression. Ethnography observes behavior in context. Diary and mobile methods capture experience over time. Online communities and review analysis can reveal naturally occurring language but may overrepresent highly motivated voices.

Quality requires purposive sampling logic, a transparent interview guide, reflexive field notes, systematic coding, attention to negative cases, and an account of how interpretations were developed. Saturation is not a magic sample-size number. It refers to the point at which additional data no longer generate sufficiently new insights for the analytic purpose; researchers should state what kind of saturation is claimed.

Quantitative research

Quantitative research supports estimation, comparison, modeling, and formal uncertainty assessment. Surveys require carefully designed constructs, response scales, screening, routing, pretesting, and procedures for nonresponse. Behavioral data can be highly relevant but are not self-explanatory: an observed click may reflect curiosity, accidental exposure, or interface design rather than preference.

Quantitative analysis should distinguish statistical significance from practical significance. A very large sample can make a trivial difference statistically detectable, while a meaningful effect may be uncertain in a small sample. Confidence intervals, effect sizes, assumptions, and sensitivity analyses are therefore important.

Mixed methods

Mixed methods intentionally integrates qualitative and quantitative evidence rather than placing two unrelated studies side by side. Common designs include:

• Exploratory sequential: qualitative discovery followed by quantitative measurement;
• Explanatory sequential: quantitative pattern followed by qualitative explanation;
• Convergent: qualitative and quantitative evidence collected in parallel and compared;
• Embedded: one method supports a primary design, such as interviews within an experiment.

Integration may occur through connecting samples, building an instrument from themes, merging results in a joint display, or using one dataset to explain another. Disagreement is not automatically a failure. It may reveal segment differences, context effects, measurement problems, or distinct layers of consumer experience.

6. Measurement: Constructs, Scales, Validity, and Reliability

A construct is an abstract concept that cannot be observed directly. “Trust,” for example, may include competence, integrity, benevolence, and predictability. Operationalization defines how the construct will be represented by indicators or questions.

Researchers should distinguish formative and reflective measurement. In a reflective model, indicators are viewed as manifestations of an underlying construct; changes in the construct are expected to influence responses. In a formative model, indicators combine to form the construct, and removing one may change its meaning. The choice affects scale development and interpretation.

Validity

Validity concerns whether evidence supports the intended interpretation and use of a measure or design.

• Face validity: Does the measure appear relevant and understandable? Useful, but weak on its own.
• Content validity: Do items adequately cover the construct domain? Requires conceptual definition and expert review.
• Construct validity: Does the measure behave as theory predicts? Convergent and discriminant evidence are relevant.
• Criterion-related validity: Does the measure relate to an appropriate external criterion, concurrently or predictively?
• Internal validity: In a causal design, can observed change be attributed to the intervention rather than alternative explanations?
• External validity: To what populations, settings, and times can findings reasonably generalize?
• Ecological validity: Does the research context resemble consequential real-world experience?

Validity is not a permanent property of an instrument. A scale validated in one population, language, channel, or decision context may not function identically elsewhere.

Reliability

Reliability concerns consistency or precision. Test–retest reliability assesses stability over time when the construct should remain stable. Inter-rater reliability assesses agreement among observers or coders. Internal consistency examines whether items function coherently, but a high coefficient does not prove unidimensionality or validity.

For a set of k items, coefficient alpha is often written as:

Here, σᵢ² is the variance of item i, σₜ² is the variance of the total score, and k is the number of items. The formula assumes a conventional additive scale and is most interpretable when items are reasonably unidimensional. Alpha can increase simply because many similar items are included; it should not be treated as a universal pass/fail test. Researchers should also inspect item content, factor structure, alternative reliability estimates, and measurement invariance across groups.

Measurement invariance asks whether a construct has comparable meaning across groups or occasions. Without adequate invariance, observed group differences may reflect different interpretation of the instrument rather than genuine psychological differences.

7. Sampling and Generalization

Sampling determines whose experiences can influence the conclusion. A target population is the conceptual group about which the study intends to speak. A sampling frame is the operational list or mechanism from which units are selected. The sample is the set actually observed.

Probability sampling gives units a known, non-zero chance of selection and supports design-based inference when implemented correctly. Simple random, systematic, stratified, and cluster sampling are common forms. Stratification can improve representation of important subgroups; clustering can reduce cost but often increases sampling variance because units within clusters resemble one another.

Nonprobability sampling includes convenience, quota, purposive, snowball, and opt-in online samples. It may be appropriate for exploratory or hard-to-reach populations, but conventional margins of error do not automatically apply. Researchers should describe recruitment, compare sample composition with relevant benchmarks, use calibration or weighting cautiously, and state the limits of generalization.

Sampling error is not the only threat. Coverage error excludes people from the frame; nonresponse error arises when participation relates to the outcome; measurement error results from poor questions or instruments; processing error arises during coding, cleaning, or linkage. Increasing sample size does not cure systematic bias.

For a proportion estimated from a simple random sample, an approximate 95% margin of error is:

where p is the estimated proportion and n is the sample size. This approximation assumes an appropriate probability sample, independent observations, and sufficiently large expected counts. It does not account automatically for design effects, weighting, finite populations, nonresponse bias, or model-based adjustments. Thus, it should not be presented as a guarantee of accuracy for an opt-in panel.

8. Triangulation and Evidence Integration

Triangulation strengthens insight by examining a question through multiple evidence paths. It can involve:

• Data triangulation: different sources, times, locations, or segments;
• Method triangulation: interviews, surveys, experiments, and behavioral records;
• Analyst triangulation: independent coding or interpretation;
• Theory triangulation: competing explanations;
• Investigator or stakeholder triangulation: different perspectives on meaning and action.

Triangulation is not simply counting how many sources agree. Three datasets may share the same underlying bias. Conversely, disagreement may be informative. For example, customers may report high satisfaction while usage data show declining engagement. The discrepancy could reflect social desirability, a poorly worded survey, habitual usage, or a difference between affective evaluation and behavior.

A practical integration matrix should record each finding, its source, population, method, limitations, degree of convergence, and managerial implication. Conclusions should be graded: established, supported but conditional, plausible, or unresolved.

9. A Complete Research Workflow

A disciplined workflow can be organized as follows:

1. Diagnose the decision: Identify the choice, alternatives, constraints, and consequences.
2. Audit existing evidence: Review internal data, prior studies, market information, and theory.
3. Map uncertainty: Separate known facts, assumptions, contradictions, and missing information.
4. Define constructs and population: Establish conceptual and operational boundaries.
5. Prepare the insight brief: State questions, hypotheses, methods, risks, and decision rules.
6. Choose a design: Match exploratory, descriptive, causal, predictive, or mixed methods to the claim.
7. Pilot and pretest: Check comprehension, timing, routing, instrumentation, and data capture.
8. Collect and document data: Preserve protocols, deviations, consent records, and metadata.
9. Analyze against the plan: Distinguish confirmatory from exploratory analyses.
10. Integrate and challenge findings: Seek negative cases, alternative explanations, and subgroup differences.
11. Translate insight into action: Specify what should change, for whom, why, and under what conditions.
12. Monitor outcomes: Treat implementation as a further learning cycle rather than the end of research.

Illustrative Case: Repositioning a Sustainable Household Product

Background

A fictional household-products company, GreenNest, sells a refillable cleaning product. Awareness is high, but trial is low and repeat purchase is weaker than management expected. Executives believe that consumers either distrust product effectiveness or find the refill process inconvenient.

Problem formulation

The initial request is “Make the product more appealing.” The research team reframes the decision: Should GreenNest prioritize proof of cleaning performance, simplify the refill experience, adjust the price architecture, or target a different segment?

The team prepares three hypotheses:

• H1: Perceived cleaning effectiveness is positively associated with trial intention.
• H2: Perceived refill effort is negatively associated with repeat intention.
• H3: A demonstration of performance increases trial more among consumers with low prior category knowledge than among highly knowledgeable consumers.

Approach

First, twelve purposively selected interviews and two online communities explore consumer language, routines, and perceived barriers. The team discovers that “sustainable” is interpreted in competing ways: some participants focus on waste reduction, while others demand evidence about chemical safety and cleaning power.

Second, a probability-based survey of category buyers measures awareness, perceived effectiveness, perceived effort, trust, price perceptions, and trial intention. Items are pretested for comprehension, and demographic and category-use quotas are monitored. The survey estimates the prevalence of barriers but cannot by itself establish that changing a barrier will cause trial.

Third, a randomized digital experiment assigns eligible consumers to one of three product pages: standard information, performance demonstration, or refill-simplicity demonstration. Trial-button selection is measured, with randomization and exposure quality documented. A follow-up study examines whether effects persist beyond immediate clicks.

Analysis and interpretation

Interview findings explain meanings and suggest constructs. Survey results identify the scale and distribution of those constructs. The experiment estimates whether alternative messages change behavior under the tested conditions. Suppose the performance demonstration increases trial-button selection, but only modestly, while the refill-simplicity demonstration increases repeat-plan selection. The appropriate conclusion is not “performance does not matter.” Rather, performance may be necessary for credibility, while convenience is the more actionable lever for repeat behavior.

Outcome

GreenNest retains performance evidence, redesigns refill instructions, and pilots a simplified package in selected stores. It monitors actual repeat purchase, customer complaints, and return rates rather than relying only on digital engagement. The company also documents that findings apply to category buyers exposed to the tested messages, not automatically to all households.

Learning points

The case demonstrates sequential mixed methods, the distinction between association and causality, the value of decision rules, and the importance of linking psychological constructs to observable outcomes. It also shows why a single “consumer insight” rarely answers all parts of a managerial problem.

10. Ethics, Governance, and AI-Enabled Research

Consumer research involves power, privacy, identity, and potential behavioral influence. Ethical practice requires informed consent where appropriate, voluntary participation, protection from undue harm, data minimization, secure storage, controlled access, and honest communication about use. Sensitive data require heightened safeguards. Researchers should avoid manipulative recruitment, deceptive practices without strong justification, discriminatory targeting, and unnecessary collection of identifiable information.

Legal compliance is necessary but not sufficient. Privacy regulations may govern consent, purpose limitation, access, deletion, profiling, and cross-border transfer, but ethical review should also consider dignity, fairness, vulnerability, and reasonable consumer expectations.

Artificial intelligence can support transcription, coding assistance, semantic clustering, synthetic stimulus generation, anomaly detection, and predictive modeling. However, AI outputs may reproduce sampling bias, invent quotations, obscure analytic decisions, or infer sensitive attributes. Governance should include human review, provenance records, access controls, model documentation, bias testing, disclosure where relevant, and a prohibition on treating generated text as direct consumer evidence. Synthetic respondents or generated focus-group participants are useful for ideation only; they cannot replace evidence from real consumers.

Challenges and Limitations

No design eliminates uncertainty. Common challenges include:

• Construct ambiguity: terms such as engagement, loyalty, or value may conceal several dimensions.
• Self-report limitations: memory, identity management, and social desirability affect answers.
• Behavioral ambiguity: observed behavior does not automatically reveal motivation.
• Common-method bias: measuring predictor and outcome with the same instrument can inflate relationships.
• Selection effects: highly engaged, dissatisfied, or digitally active consumers may be overrepresented.
• Causal overreach: correlations and model coefficients are often presented as explanations.
• Generalization limits: a study’s population, context, and time frame constrain application.
• Organizational distortion: stakeholders may selectively emphasize findings that support prior preferences.
• Reproducibility and drift: consumer meanings, platforms, algorithms, and markets change.

Best practice is not to promise certainty. It is to make uncertainty explicit, reduce avoidable bias, test important assumptions, and connect evidence strength to the scale of the decision.

Chapter Summary

Consumer insight systems begin with a decision, not a data source. Managers and researchers must translate symptoms into researchable problems, define constructs and populations, and state how evidence will change action. Insight briefs align objectives, hypotheses, methods, decision rules, and ethical safeguards.

Exploratory research discovers meanings and mechanisms; descriptive research estimates distributions and patterns; causal research tests the effects of interventions; and predictive research estimates future outcomes. Primary and secondary data, qualitative and quantitative methods, and mixed-method designs each have distinctive strengths and limitations. Methodological fit determines whether a conclusion is credible.

Validity concerns the defensibility of interpretation, while reliability concerns consistency or precision. Sampling quality depends on coverage, recruitment, response, measurement, and processing—not merely sample size. Triangulation integrates different evidence sources and treats disagreement as a possible insight. Finally, privacy, fairness, transparency, and human oversight are essential when consumer data and AI systems influence marketing decisions.

Key Terms

Causal research: Research designed to estimate whether changing an intervention produces a change in an outcome.

Construct: An abstract concept, such as trust or perceived value, represented through indicators.

Descriptive research: Research that estimates characteristics, frequencies, distributions, or trends.

External validity: The defensibility of generalizing findings to other populations, settings, or times.

Exploratory research: Research used to investigate poorly understood meanings, problems, or mechanisms.

Formative measurement: Measurement in which indicators combine to form a construct.

Hypothesis: A theoretically or empirically grounded proposition that can be tested with evidence.

Insight brief: A document connecting a decision, research questions, methods, risks, and decision criteria.

Internal validity: The extent to which a causal conclusion is free from competing explanations.

Measurement invariance: Evidence that a measure has comparable meaning across groups or occasions.

Mixed methods: Intentional integration of qualitative and quantitative approaches in one inquiry.

Predictive research: Research that estimates the likelihood or level of a future or unobserved outcome.

Primary data: Data collected specifically for the current research purpose.

Reliability: The consistency or precision of a measurement or procedure.

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

Secondary data: Existing data collected for another purpose or an earlier study.

Triangulation: Examination of a question through multiple data sources, methods, analysts, or theories.

Validity: The extent to which evidence supports the intended interpretation and use of findings.

Review Questions

Short Answer Questions

1. How does a managerial problem differ from a research problem?
2. What are the main components of an insight brief?
3. Distinguish exploratory research from descriptive research.
4. Why does a large sample not automatically eliminate bias?
5. What is the purpose of triangulation?

Long Answer Questions

1. Explain how a firm could move from a symptom of declining sales to a complete researchable decision problem.
2. Compare causal and predictive research in terms of purpose, design, evidence, and managerial interpretation.
3. Discuss validity and reliability, including why a reliable measure may still be invalid.
4. Evaluate the advantages and limitations of qualitative, quantitative, and mixed-method designs.
5. Explain how privacy, fairness, and transparency should shape AI-enabled consumer research.

Analytical / Application Questions

1. A retailer observes that loyalty members purchase less frequently after a mobile-app redesign. Design an exploratory, descriptive, and causal sequence to investigate the problem.
2. A survey finds that perceived brand trust is strongly associated with purchase intention. Identify at least three alternative explanations and propose a design that would better assess causality.
3. A company’s opt-in online panel overrepresents highly educated urban consumers. Explain what claims can and cannot be made, and recommend steps to improve the credibility of the evidence.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes a managerial problem from a research problem?

A. A managerial problem concerns a decision, whereas a research problem identifies the uncertainty that informs that decision.

B. A managerial problem always uses quantitative data, whereas a research problem always uses qualitative data.

C. A managerial problem is theoretical, whereas a research problem is exclusively operational.

D. A managerial problem is broader and therefore cannot be investigated.

Correct Answer: A. A managerial problem concerns a decision, whereas a research problem identifies the uncertainty that informs that decision.

Question 2

Which research purpose is primarily concerned with discovering meanings, mechanisms, and possible hypotheses?

A. Descriptive research

B. Exploratory research

C. Causal research

D. Predictive research

Correct Answer: B. Exploratory research

Question 3

In a randomized experiment, the simple difference in proportions is represented by which expression?

A. p₁ + p₀

B. p₁ ÷ p₀

C. p₁ − p₀

D. p₁ × p₀

Correct Answer: C. p₁ − p₀

Question 4

What is a central limitation of using an opt-in online panel?

A. It necessarily prevents all subgroup analysis.

B. It automatically produces a larger margin of error than a census.

C. It may have selection bias, so conventional probability-sample margins of error may not apply.

D. It cannot collect attitudinal data.

Correct Answer: C. It may have selection bias, so conventional probability-sample margins of error may not apply.

Question 5

Which form of validity concerns whether a causal effect can be attributed to the intervention rather than competing explanations?

A. Content validity

B. Internal validity

C. Face validity

D. Ecological validity

Correct Answer: B. Internal validity

Question 6

What is the main purpose of predictive research?

A. To estimate likely future or unobserved outcomes

B. To guarantee an explanation of why behavior occurs

C. To replace all descriptive research

D. To discover consumer language without measurement

Correct Answer: A. To estimate likely future or unobserved outcomes

Question 7

Which mixed-method design typically uses qualitative findings to develop a later quantitative instrument?

A. Exploratory sequential design

B. Explanatory sequential design

C. Convergent design

D. Embedded experimental design

Correct Answer: A. Exploratory sequential design

Question 8

If 18% of a treatment group renew and 15% of a control group renew, what is the absolute difference in renewal proportions?

A. 3 percentage points

B. 15 percentage points

C. 18 percentage points

D. 20 percentage points

Correct Answer: A. 3 percentage points

Question 9

Which statement about coefficient alpha is most accurate?

A. A high value proves that a scale is valid.

B. A high value can result from many similar items and does not by itself prove unidimensionality.

C. It is used only for causal experiments.

D. It measures whether a sample represents the population.

Correct Answer: B. A high value can result from many similar items and does not by itself prove unidimensionality.

Question 10

What does triangulation involve?

A. Using only the largest available dataset

B. Selecting the method preferred by senior management

C. Examining a question through multiple evidence sources, methods, analysts, or theories

D. Removing all findings that disagree

Correct Answer: C. Examining a question through multiple evidence sources, methods, analysts, or theories