Research Philosophy, Theory and Design

Introduction

Market research is not simply the collection of consumer opinions or the production of statistical reports. It is a disciplined process of deciding what counts as knowledge, how knowledge can be generated, which claims are justified by evidence, and how those claims should guide managerial action. In brand management, these decisions influence positioning, identity, customer experience, innovation, communication, pricing and portfolio strategy.

A study of brand meaning, for example, may require interpretation of consumers’ narratives, while a study of advertising effectiveness may require measurement and causal inference. Neither approach is automatically superior. The appropriate choice depends on the research question, the assumptions made about reality and knowledge, the resources available, and the consequences of the decision.

This chapter examines three influential philosophical perspectives—positivism, interpretivism and pragmatism—and connects them to theory, methodology and research design. It evaluates exploratory, descriptive, diagnostic, causal and predictive designs, and explains how validity, reliability, triangulation and limitations should be assessed in professional market research.

Learning Objectives

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

1. Distinguish ontology, epistemology, methodology and methods.
2. Compare positivist, interpretivist and pragmatic perspectives.
3. Explain how theory informs research questions, constructs and hypotheses.
4. Select and evaluate exploratory, descriptive, diagnostic, causal and predictive designs.
5. Assess validity, reliability, credibility, dependability and transferability.
6. Use triangulation appropriately rather than treating it as automatic proof.
7. Identify ethical, practical and technological limitations in brand research.

1. The Architecture of Research Thinking

Research philosophy concerns the assumptions that underpin inquiry. Four related concepts are particularly important.

Ontology concerns the nature of reality. A realist may assume that markets, attitudes and behaviours exist independently of the researcher, although they may be difficult to observe. A constructionist may emphasise that meanings such as “premium,” “authentic” or “sustainable” are produced through social interaction and culture.

Epistemology concerns what can be known and how knowledge can be justified. A positivist tradition privileges systematic observation, measurement and testing. An interpretivist tradition asks how people understand their worlds and how meanings are formed. A pragmatic tradition evaluates knowledge partly by its usefulness in addressing a practical problem.

Methodology is the logic that connects philosophical assumptions to a research strategy. A survey, experiment, ethnography or mixed-methods design is not merely a technical choice; each embodies assumptions about evidence and explanation.

Methods are the specific procedures used to collect and analyse data, such as interviews, observation, questionnaires, experiments, social-listening analysis or regression modelling.

A useful distinction is between methodological fit and methodological fashion. A method is suitable when it can answer the research question with an appropriate level of confidence, not merely because it is popular or technologically advanced.

2. Positivism, Interpretivism and Pragmatism

Positivism

Positivism treats social phenomena as sufficiently regular to be observed, measured and analysed. In market research, it is commonly associated with structured questionnaires, probability sampling, experiments, behavioural data and statistical hypothesis testing.

A positivist researcher may conceptualise brand consideration as a measurable construct, operationalise it through survey items, and test whether perceived relevance predicts purchase intention. The approach values replicability, standardisation and control of alternative explanations.

Its strengths include comparability across segments, estimation of prevalence, formal testing and support for decisions involving scale. Its limitations include the risk of reducing complex meanings to crude indicators, assuming that respondents interpret questions identically, and overlooking context, emotion and power relationships.

Positivism does not mean that every quantitative study is perfectly objective. Choices about constructs, measures, samples, exclusions and models involve judgement. Quantitative results are therefore shaped by design decisions as well as by numerical calculation.

Interpretivism

Interpretivism views social reality as meaning-laden and often multiple. Consumers do not merely respond to brands; they interpret, negotiate and sometimes resist brand meanings. The same advertisement may signify empowerment to one group, opportunism to another and irrelevance to a third.

Interpretivist research commonly uses depth interviews, focus groups, ethnography, participant observation, diary studies, netnography and semiotic analysis. The researcher seeks to understand participants’ perspectives in context rather than impose a predetermined measurement structure.

Quality is assessed through concepts such as credibility, reflexivity, transparency, dependability and transferability. A strong interpretive account explains how data were generated, how interpretations developed, which voices were included or excluded, and why the conclusions are plausible.

Interpretivism can reveal symbolic meanings, contradictions and unexpected consumer practices. However, small or purposive samples do not normally support statistical estimates of population prevalence. Interpretation can also be affected by researcher assumptions, participant performance and the relationship between researcher and participant.

Pragmatism

Pragmatism begins with the problem and asks which forms of evidence are most useful for addressing it. It does not require a rigid commitment to one philosophical tradition. A pragmatic project may combine qualitative exploration, quantitative measurement and experimentation in a sequence that fits the decision.

For example, a brand may first conduct interviews to discover how consumers define “ethical fashion,” then develop survey measures to estimate the size of each meaning segment, and finally run an experiment to test whether alternative claims change choice. The combination is not justified merely because “more data are better”; each stage must answer a distinct question.

Pragmatism is especially relevant to managerial research because decisions often require both understanding and prediction. Its danger is instrumentalism without intellectual discipline: researchers may select methods opportunistically, ignore incompatible assumptions, or treat business usefulness as proof of truth.

Comparative view

3. Theory and Its Role in Brand Research

Theory is an organised set of concepts and propositions that explains relationships, processes or meanings. It is not simply a literature review, a list of variables or a managerial slogan.

A construct is an abstract concept, such as brand trust, perceived authenticity or involvement. An indicator is an observable representation of a construct, such as a questionnaire item, behavioural measure or coded statement. Operationalisation specifies how a construct will be observed or measured.

A conceptual framework maps the expected relationships among constructs. A hypothesis is a testable proposition, often expressed as an expected relationship between variables. For example: higher perceived brand credibility is associated with stronger purchase intention, after accounting for category involvement. An interpretive study may instead use sensitising concepts—ideas that guide attention without imposing fixed variables.

Theory can be used deductively, inductively or abductively. Deduction begins with theory and tests implications. Induction develops concepts or explanations from patterns in data. Abduction moves iteratively between surprising observations and plausible explanations. In applied brand research, abduction is often valuable because market evidence may challenge existing assumptions.

A sound research question should identify the decision or phenomenon, the relevant population or context, and the type of knowledge required. “What do consumers think of our brand?” is too broad. “How do first-time buyers interpret the brand’s repairability claim, and which interpretation affects willingness to recommend?” is more analytically useful.

4. Research Designs by Purpose

Research purposes overlap, but the distinctions below help researchers select an appropriate design.

Exploratory design

Exploratory research investigates a poorly understood problem, discovers relevant concepts and generates propositions. It is appropriate when managers do not yet know which attributes matter or why a campaign is being rejected.

Methods include secondary research, expert interviews, focus groups, depth interviews, social-media observation and pilot studies. Sampling is usually purposive or theoretical rather than designed to estimate population percentages. Analysis focuses on themes, contrasts, mechanisms and emerging explanations.

Exploration should not be presented as proof of prevalence. If eight interviewees mention convenience, the result suggests an issue for further investigation; it does not establish that a particular percentage of the market values convenience.

Descriptive design

Descriptive research estimates characteristics, distributions or patterns. It can answer questions such as: What proportion of category buyers recognise the brand? Which segments report the highest awareness? How frequently do customers use the service?

Structured surveys, panels, audits, transaction records and observational counts are common. Sampling quality is central. A large convenience sample may be less informative than a smaller probability sample if the objective is population estimation.

A proportion estimate can be represented as:

Here, p̂ is the estimated proportion, x is the number of sampled cases with the attribute, and n is the total number of valid sampled cases. The notation is appropriate when cases are defined consistently and the sample design is understood. It does not, by itself, correct for non-coverage, non-response or measurement error.

Diagnostic design

Diagnostic research asks why a pattern exists. A fall in brand preference is descriptive; identifying whether the fall is associated with distribution problems, perceived price unfairness, competitor innovation or negative publicity is diagnostic.

Diagnostic analysis can combine qualitative accounts, subgroup comparisons, journey analysis, observational evidence and statistical modelling. Correlation can identify associations, but association alone does not establish cause. For example, loyal customers may report higher satisfaction because loyalty influences satisfaction reporting, because satisfaction creates loyalty, or because a third factor affects both.

Causal design

Causal research evaluates whether changing one factor produces a change in an outcome. Experiments are the strongest design for causal inference when random assignment, treatment control and valid outcome measurement are feasible.

In a simple randomised experiment, participants are assigned to treatment and control conditions. The average treatment effect can be expressed as:

Here, Y(1) is the potential outcome if a unit receives treatment, Y(0) is the potential outcome if it does not, and E denotes the population average. The fundamental problem is that the same unit cannot generally be observed simultaneously under both conditions. Randomisation makes the groups comparable in expectation, allowing the difference in observed group means to estimate the causal effect under assumptions such as adequate compliance, no interference that distorts treatment effects, and valid outcome measurement.

Brand experiments may test message framing, price presentation, packaging, landing-page design or promotional incentives. Internal validity can be weakened by treatment contamination, demand effects, attrition, poor randomisation or outcomes that do not represent real behaviour. External validity may be limited when an artificial test environment differs from the market.

Predictive design

Predictive research estimates likely future or unobserved outcomes, such as churn, response to an offer or probability of conversion. Prediction and explanation should not be confused. A variable can improve prediction without being a cause; a causal variable may have weak predictive value if its effect is small or difficult to measure.

A common classification measure is accuracy:

TP, TN, FP and FN denote true positives, true negatives, false positives and false negatives. Accuracy is appropriate only when the decision context and class balance make it meaningful. In a rare-event problem, a model that predicts “no purchase” for everyone may have high accuracy but little managerial value. Precision, recall, calibration, expected profit and fairness may be more relevant.

Predictive models require separation of training and evaluation data, careful handling of missing data and protection against leakage. A model that performs well on historical data may fail when consumer behaviour, media conditions or competitors change.

Design selection matrix

In practice, a project may use a sequence: exploratory interviews, descriptive survey, causal test and predictive deployment. The sequence should be driven by uncertainty and decision risk, not by an assumption that every project must contain every method.

5. Validity, Reliability and Quality

Validity

Validity concerns whether a study supports the claims it makes.

Construct validity asks whether an operational measure represents the intended construct. A measure of “authenticity” that only asks whether a brand is old may have inadequate construct coverage.

Internal validity concerns whether an observed effect in the study can reasonably be attributed to the proposed cause rather than confounding, selection, maturation or measurement problems.

External validity concerns the extent to which findings apply across people, settings, times or stimuli. Generalisation is not automatic; it depends on sampling, context and similarity between the study and target market.

Ecological validity concerns realism of the setting and task. A laboratory choice may not reflect a purchase involving budget constraints, social influence and competing products.

In qualitative research, analogous concerns are often expressed through credibility, transferability, dependability and confirmability rather than through statistical validity alone.

Reliability

Reliability concerns consistency and dependability of measurement or procedure. A scale may show internal consistency yet fail to measure the intended construct. Conversely, a qualitative study can be dependable when its sampling, interview process, coding decisions and analytic trail are transparent, even though another researcher may not produce identical interpretations.

For a scale with items measuring one construct, internal consistency is sometimes summarised using Cronbach’s alpha:

Here, k is the number of items, σᵢ² is the variance of item i, and σₜ² is the variance of the total score. The statistic assumes that items are meaningfully related and is not a universal proof of unidimensionality or validity. Very high values can even indicate redundant items. Researchers should examine item content, dimensionality and the purpose of the scale rather than apply a single cutoff mechanically.

Reliability can be improved through clear procedures, interviewer training, pilot testing, consistent coding rules, attention checks where appropriate, test–retest assessment and documentation of deviations.

6. Triangulation and Mixed Evidence

Triangulation compares evidence from different sources, methods, researchers, theories or settings. Its purpose is not simply to make a conclusion appear stronger. Different methods may reveal different aspects of a phenomenon because they access different forms of evidence.

Suppose survey respondents rate a brand as highly sustainable, interviews reveal scepticism about its claims, and purchase data show little behavioural response. The apparent contradiction is analytically valuable. It may indicate social desirability, weak behavioural relevance, a measurement problem or a gap between attitude and action.

Common forms include:

• Data triangulation: different samples, times, locations or data sources.
• Method triangulation: interviews, surveys, observation or experiments.
• Investigator triangulation: more than one analyst or field researcher.
• Theory triangulation: competing conceptual explanations.
• Mixed-methods integration: deliberate connection of qualitative and quantitative stages.

Integration should be planned. Researchers may connect stages by using qualitative findings to develop survey items, merge results in a joint display, or embed interviews within an experiment to explain reactions. Mere accumulation of data does not constitute triangulation.

7. Illustrative Case: Repositioning an Urban Mobility Brand

The following is a fictional illustrative case.

Background

MetroLoop, a subscription-based urban mobility brand, experienced declining renewal rates among customers aged 25–40. Management initially suspected that price was the main problem.

Problem

The decision was whether to reduce price, improve the service experience, or reposition the brand around reliability and sustainability.

Analysis and approach

The research team adopted a pragmatic sequential design. First, 24 purposively selected customers participated in interviews and two weeks of travel diaries. The analysis showed that “reliability” meant more than vehicle availability: customers valued predictable pick-up, transparent billing and confidence that support would respond quickly. Sustainability was appreciated but rarely decisive at the moment of renewal.

Second, a stratified survey of 1,200 active and lapsed users measured the prevalence of these concerns. The sample was weighted to align with known customer distributions, although weighting could not eliminate non-response bias. Reliability perceptions were strongly associated with renewal intention, but the survey was diagnostic rather than causal.

Third, MetroLoop ran a four-week field experiment across comparable service zones. Some zones received improved real-time availability information and proactive service alerts; other zones continued with standard communication. Renewal rates and complaint rates were tracked. Randomisation at the zone level reduced contamination but created a relatively small number of experimental units, limiting statistical precision.

Finally, follow-up interviews explored how customers interpreted the intervention. Participants valued accurate information but reacted negatively when alerts promised certainty that the service could not deliver.

Outcome

The company did not implement a blanket price reduction. It improved availability information, redesigned billing explanations and changed communications from absolute promises to transparent service-status updates. The experiment indicated improved renewal in treatment zones, but management treated the estimate as context-specific and continued monitoring after rollout.

Learning points

1. The original managerial assumption confused a possible cause with an established cause.
2. Exploratory research clarified the meaning of reliability.
3. Descriptive research assessed the distribution of concerns.
4. The field experiment provided stronger causal evidence than survey association alone.
5. Qualitative follow-up identified a risk—overpromising—that aggregate outcomes might conceal.
6. Pragmatism required integration, not indiscriminate use of methods.

8. Ethics, Governance and Contemporary Practice

Market research must respect informed consent, voluntary participation, privacy, confidentiality and the right to withdraw. Researchers should collect only data necessary for the stated purpose, explain foreseeable uses, protect sensitive information and avoid deceptive or coercive recruitment. Legal obligations vary by jurisdiction and may include data-protection, consumer-protection and sector-specific rules.

Brand researchers must be especially careful with children, vulnerable groups, location data, inferred attributes and online communities. Publicly accessible content is not automatically ethically unrestricted; context, reasonable expectations and potential harm matter.

Artificial intelligence can assist transcription, coding, sentiment classification, survey design and predictive modelling. However, automated outputs may reproduce training-data bias, erase ambiguity, expose confidential material or create fabricated interpretations. Human review, provenance records, secure processing, validation against a sample of manually assessed cases and disclosure of material AI use are prudent practices. Generative systems should support, not replace, the researcher’s responsibility for sampling, interpretation and claims.

Contemporary research also faces declining survey response, platform-dependent digital traces, synthetic respondents, privacy-enhancing technologies and rapid changes in media environments. These developments increase the importance of methodological transparency. Researchers should state who was studied, how data were generated, what was excluded, which assumptions were made and where uncertainty remains.

9. Best-Practice Checklist

Before approving a research design, ask:

1. What decision will the evidence inform?
2. Is the question exploratory, descriptive, diagnostic, causal, predictive or a justified combination?
3. Which philosophical assumptions are relevant, and are they consistent with the claims?
4. Are constructs clearly defined and operationalised?
5. Does the sampling strategy match the intended generalisation?
6. Are alternative explanations and sources of bias considered?
7. Are measurement reliability and validity assessed appropriately?
8. If multiple methods are used, how will they be integrated?
9. Are privacy, consent, fairness and data governance addressed?
10. What limitations must be communicated to decision-makers?

Chapter Summary

Research philosophy shapes what researchers regard as reality, evidence and justified knowledge. Positivism emphasises measurement, systematic observation and testing; interpretivism emphasises meaning, context and participant perspectives; pragmatism selects and integrates methods according to the practical problem while retaining methodological discipline. Theory connects abstract constructs to research questions, indicators and explanations.

Exploratory designs discover concepts, descriptive designs estimate patterns, diagnostic designs investigate reasons, causal designs test effects, and predictive designs estimate future outcomes. These purposes are distinct even though one project may combine them. Validity concerns the defensibility of claims, while reliability concerns consistency and dependability. Qualitative quality is also assessed through credibility, reflexivity, dependability and transferability. Triangulation can reveal convergence, complementarity or contradiction, but it is valuable only when integration is planned and interpreted critically.

For brand managers, the central lesson is that research quality is not determined by data volume or technological sophistication. It depends on fit between the decision, the philosophical stance, the design, the evidence, the analysis and the limits of the conclusion.

Key Terms

Ontology: Assumptions about the nature of reality.

Epistemology: Assumptions about knowledge and how it can be justified.

Methodology: The logic linking research philosophy, design and methods.

Positivism: A perspective emphasising systematic observation, measurement and testing.

Interpretivism: A perspective emphasising socially constructed meaning and context.

Pragmatism: A problem-centred perspective concerned with useful and defensible inquiry.

Theory: An organised set of concepts and propositions explaining relationships or processes.

Construct: An abstract concept represented through observable indicators.

Operationalisation: Specification of how a construct will be observed or measured.

Exploratory research: Inquiry used to discover concepts, issues and propositions.

Descriptive research: Research estimating characteristics, distributions or patterns.

Diagnostic research: Research investigating why an observed pattern occurs.

Causal research: Research examining whether a change produces an outcome change.

Predictive research: Research estimating likely future or unobserved outcomes.

Validity: The extent to which evidence supports the intended claims.

Reliability: Consistency and dependability of measurement or procedure.

Triangulation: Comparison and integration of evidence from multiple sources or methods.

Reflexivity: Critical examination of the researcher’s influence and assumptions.

Internal validity: Confidence that an observed effect is attributable to the proposed cause.

External validity: Extent to which findings apply beyond the study conditions.

Review Questions

Short Answer Questions

1. Distinguish ontology from epistemology.
2. What is the principal purpose of exploratory research?
3. Why does correlation not necessarily establish causation?
4. Define construct validity.
5. Identify two forms of triangulation.

Long Answer Questions

1. Compare positivism, interpretivism and pragmatism in relation to brand research.
2. Explain how theory guides constructs, operationalisation and hypotheses.
3. Evaluate the differences among exploratory, descriptive and diagnostic designs.
4. Discuss internal, external and ecological validity in a brand experiment.
5. Explain why reliability alone cannot establish that a measure is valid.

Analytical / Application Questions

1. A beverage brand has declining consideration but does not know why. Design a sequential research programme and justify each stage.
2. A survey finds that consumers value sustainability, but sales data show no change in choice. Use triangulation to propose at least three explanations and a method for testing them.
3. A predictive churn model is highly accurate because 95% of customers do not churn. Evaluate the model and recommend more appropriate performance and governance criteria.

Multiple Choice Questions with Answers

Question 1

Which perspective places greatest emphasis on socially constructed meaning and context?

A. Positivism

B. Interpretivism

C. Pragmatism

D. Functionalism

Correct Answer: B. Interpretivism

Question 2

A study designed to estimate the percentage of consumers recognising a brand is primarily:

A. Exploratory

B. Descriptive

C. Causal

D. Ethnographic

Correct Answer: B. Descriptive

Question 3

Which design is most directly concerned with whether changing an intervention produces a change in an outcome?

A. Descriptive

B. Exploratory

C. Causal

D. Diagnostic

Correct Answer: C. Causal

Question 4

In the expression ATE = E[Y(1)] − E[Y(0)], Y(1) represents:

A. The observed outcome for the control group

B. The potential outcome under treatment

C. The average treatment error

D. The sample size under treatment

Correct Answer: B. The potential outcome under treatment

Question 5

Which statement about predictive accuracy is correct?

A. It always establishes causation

B. It is unaffected by class imbalance

C. It is calculated using true and false classifications

D. It makes external validation unnecessary

Correct Answer: C. It is calculated using true and false classifications

Question 6

Which form of validity concerns whether a measure represents the intended construct?

A. Construct validity

B. External validity

C. Ecological validity

D. Statistical conclusion validity

Correct Answer: A. Construct validity

Question 7

What is the principal purpose of triangulation?

A. To guarantee that all methods produce identical findings

B. To compare and integrate different sources or forms of evidence

C. To replace sampling with technology

D. To eliminate the need for theory

Correct Answer: B. To compare and integrate different sources or forms of evidence

Question 8

A very high Cronbach’s alpha may indicate:

A. That the scale is automatically valid

B. That items may be redundant

C. That causal effects have been proven

D. That sampling bias is impossible

Correct Answer: B. That items may be redundant

Question 9

Which approach is most consistent with pragmatism?

A. Using only experiments in every study

B. Rejecting numerical evidence as subjective

C. Selecting and integrating methods according to the decision problem

D. Avoiding theory in applied research

Correct Answer: C. Selecting and integrating methods according to the decision problem

Question 10

A large convenience sample is problematic for population estimation mainly because:

A. It always contains too few observations

B. Its selection may not represent the target population

C. It cannot contain quantitative data

D. It prevents descriptive analysis

Correct Answer: B. Its selection may not represent the target population