Top of page
Skip to main content

Research design: aligning purpose, methods, strategy and time horizon

The central design problem is rarely choosing a survey instead of a case study. It is asking “why” when the design can only describe “how many”, or collecting data once and then claiming to explain change over time.

Research design is the overall plan for answering your question. It aligns purpose, methodological choice, strategy and time horizon with your philosophy and objectives. No option is inherently best: the decisions should fit together and be judged against appropriate quality criteria.

Start reading

1. Research design and coherence

In Saunders et al.'s research onion, philosophy and approaches to theory development form the outer layers. Methodological choice, strategy and time horizon translate the question, aim and objectives into an executable project; sampling, collection and analysis are the tactics that implement it (2023, pp. 176–177).

A design should identify data sources, collection and analysis, ethical issues, and constraints involving access, time, location and funding. Explain why each decision is appropriate to your question and philosophy, even if the plan later changes (Saunders et al., 2023, pp. 178–179).

Planning a journey

The aim is the destination, objectives are criteria for choosing a route, and the design is the route itself. You may have to change route, but the revised route must still reach the intended destination (Saunders et al., 2023, p. 178).

2. Research purpose

The overarching question indicates the purpose. Saunders et al. (2023, pp. 179–181) distinguish four purposes that can be combined.

Table 1. Four research purposes, synthesised from Saunders et al. (2023, pp. 179–181).
PurposeTaskTypical questionsDesign consideration
ExploratoryClarify understanding of a problem or phenomenonWhat? How?Flexible; focus narrows as understanding develops
DescriptiveProvide an accurate account of events, people or situationsWho? What? Where? When? How?Define the phenomenon before collecting evidence
ExplanatoryExplain relationships, including proposed causal mechanismsWhy? How?Statistical or qualitative explanation requires suitable evidence
EvaluativeAssess a strategy, policy or programmeHow well? To what extent?Explaining why it works can support a theoretical contribution

Description has a legitimate role, but a dissertation may need to go beyond describing a situation to explaining it. Descripto-explanatory studies use description as a foundation for explanation. Many quantitative SEM dissertations describe their samples and variables before evaluating relationships between constructs (Saunders et al., 2023, pp. 180–181).

3. Quantitative, qualitative or mixed methods?

These labels concern numerical data, non-numerical data such as words, images, sound and video, or both. A questionnaire analysed statistically is mono-method quantitative research. Adding structured observation analysed statistically can make it multi-method quantitative research. Mixed methods involve both quantitative and qualitative collection and analysis (Saunders et al., 2023, pp. 181–182).

Table 2. Typical contrasts adapted from Saunders et al. (2023, p. 184, Table 5.1). These are broad patterns, not compulsory rules for every study.
FeatureQuantitativeQualitative
Researcher–participant relationshipSeeks independenceResearcher involved
Common termRespondentsParticipants or informants
FocusVariables and relationshipsMeanings and associated concepts
Sampling emphasisProbability sampling where statistical generalisation is intendedUsually non-probability sampling
GeneralisationStatisticalTheoretical
CollectionHighly structuredUnstructured or semi-structured
AnalysisStatistics and graphsConceptualisation
Meaning drawn fromNumbersWords, texts and images

Quantitative designs often accompany positivism and deduction, but can also fit realism or pragmatism and can generate hypotheses. Qualitative designs often accompany interpretivism and induction, yet many use abduction, moving between inductive ideas and deductive checks (Saunders et al., 2023, pp. 183–185).

Changing hypotheses after seeing results

HARKing means hypothesising after the results are known. Discuss necessary changes with your supervisor and distinguish exploratory findings from original hypotheses; do not silently rewrite the hypotheses to fit model results (Saunders et al., 2023, p. 184).

4. Mixed-methods designs

Mixed methods integrate quantitative and qualitative collection and analysis within one project, often alongside pragmatism or critical realism (Saunders et al., 2023, pp. 186–187).

Table 3. Mixed-methods designs synthesised from Saunders et al. (2023, pp. 187–190).
DesignSequenceUseful when
Concurrent triangulationQuantitative and qualitative work in the same phaseComparing whether evidence converges
Concurrent embeddedOne approach is embedded in another, such as open questions within a surveyOne form of evidence has a supporting role
Sequential exploratoryQualitative, then quantitativeExploring concepts or developing measures before broader testing
Sequential explanatoryQuantitative, then qualitativeExplaining unexpected findings or unclear mechanisms
Sequential multi-phaseThree or more linked phasesEach phase informs the next; substantial time is available

Sequential designs take time because later phases depend on earlier findings. The rationale for integration should also be specific. The book, drawing on Bryman (2006) and others, discusses initiation, facilitation, complementarity, interpretation, generalisability, diversity, problem solving, focus, triangulation and confidence in conclusions. Saying only “to improve reliability” does not show how the phases serve each other (Saunders et al., 2023, pp. 188–190).

Expert consultation before a survey

Interviews followed by a survey can be sequential exploratory research when the interviews are analysed and genuinely inform the instrument. Conversations used only to improve wording may instead be pretesting within a mono-method quantitative study. Name the design according to the work performed.

5. Research strategy

Strategy links philosophy with collection and analysis. Strategies are not inherently superior or mutually exclusive: a survey can be embedded in a case study. Choose according to coherence, existing understanding, time, resources and access (Saunders et al., 2023, p. 191).

Table 4. Research strategies and methodological choices, based on Saunders et al. (2023, p. 191, Table 5.3).
StrategyCommon methodological choices
ExperimentMono-method or multi-method quantitative
SurveyMono-method or multi-method quantitative
EthnographyMono-method or multi-method qualitative
Grounded theoryMono-method or multi-method qualitative
Narrative inquiryMono-method or multi-method qualitative
Archival and documentary researchQuantitative, qualitative or mixed
Case studyMulti-method qualitative, multi-method quantitative or mixed
Action researchQuantitative, qualitative or mixed

Survey

Surveys often accompany deduction and address what, who, where and how much. They can support descriptive, exploratory and explanatory work and help evaluate relationships. Standardised evidence from many respondents is economical and comparable; probability sampling can support population inference. Researchers must still design and test instruments, secure an appropriate sample, encourage response and manage questionnaire length. Careless implementation is a major weakness (Saunders et al., 2023, pp. 194–196).

Experiment

Experiments examine whether changing an independent variable changes an outcome. A classical experiment randomly assigns participants to intervention and control conditions. Randomisation balances groups in expectation and reduces alternative explanations, rather than guaranteeing identical groups in every realised sample. Quasi-experiments and within-subject designs provide other options (Saunders et al., 2023, pp. 192–193).

Table 5. Variable roles adapted from Saunders et al. (2023, p. 193, Table 5.4).
RoleMeaning
Independent variableManipulated or varied to examine an outcome
Dependent variableOutcome that may change with the independent variable
MediatorProposed mechanism transmitting a relationship
ModeratorChanges the nature of a relationship
Control variableMeasured factor held constant or adjusted for when justified
ConfounderExtraneous factor that may distort inference about a relationship

Case studies and qualitative strategies

A case study examines a phenomenon in depth within its real context. A case may be a person, group, organisation, process or event; defining its boundaries is essential. It is particularly useful when phenomenon and context are difficult to separate. Ethnography, grounded theory, narrative inquiry and action research offer other qualitative routes; their purposes should be distinguished (Saunders et al., 2023, pp. 191, 206–207).

6. Cross-sectional or longitudinal?

Is the study a snapshot or a diary of events over time? Cross-sectional and longitudinal horizons can accompany quantitative, qualitative and mixed methods (Saunders et al., 2023, p. 212). Course-based projects are often cross-sectional because time is limited. Longitudinal designs examine change and development; repeated secondary data can offer a longitudinal element even when new collection is impractical (pp. 212–213).

Cross-sectional surveys and causal claims

Measuring predictors and outcomes from the same respondents at the same time can create common-method concerns. Procedural responses include separating measurement occasions or sources (Podsakoff et al., 2003). Spector (2019) argues that cross-sectional designs should be used deliberately rather than dismissed automatically. Without a suitable causal identification argument, report associations or findings consistent with hypotheses, not proof that one variable caused change over time.

7. Evaluating design quality

“How do I know?” is the quality question. Research cannot eliminate all uncertainty, but it can reduce the chance of an incorrect answer. Consult your institution's criteria before choosing how to discuss reliability, validity or alternative standards (Saunders et al., 2023, p. 214).

Reliability and validity

Reliability concerns consistent collection and interpretation, including agreement within a project and whether procedures can be repeated. Measurement validity concerns measuring the intended concept; internal validity concerns the credibility of relationships and explanations; external validity concerns the scope of generalisation. Reliability is necessary but insufficient for valid conclusions (Saunders et al., 2023, pp. 215–216).

Table 6. Threats summarised from Saunders et al. (2023, pp. 216–217, Tables 5.6 and 5.7).
Quality concernThreatExample
ReliabilityParticipant errorRespondents rush a survey immediately before lunch
ReliabilityParticipant biasPublic interviews produce positive answers because others may overhear
ReliabilityResearcher errorFatigue or preparation problems cause misunderstandings
ReliabilityResearcher biasPersonal views shape recording and interpretation
Internal validityHistoryA product recall changes customers' quality assessments
Internal validityTesting effectsBeing studied changes behaviour
Internal validityInstrumentationDefinitions or measures change between stages
Internal validityAttritionParticipants leave after promotion or relocation
Internal validityMaturationOther experiences influence later responses
Internal validityCausal ambiguityIt is unclear whether negative appraisal attitudes cause low ratings or vice versa

Causal ambiguity is a central limitation of a single-wave survey. Claims that findings apply to other contexts commonly require further research in those settings (Saunders et al., 2023, p. 216).

Qualitative quality criteria

Rigidly imposing quantitative reliability and validity on interpretive research can create a mismatch. Researchers can adapt established concepts, use parallel criteria, or adopt alternatives. Dependability, credibility and transferability broadly parallel reliability, internal validity and external validity. Prolonged engagement, peer discussion, negative cases, participant validation and reflexive records can strengthen credibility (Saunders et al., 2023, pp. 217–219).

Plan triangulation and participant validation in the design. Triangulation draws on multiple sources or methods; it cannot be added retrospectively as a label in the conclusion (Saunders et al., 2023, pp. 218–219).

Ethics belongs in the design

Anticipate harm to participants and researchers, access problems and any covert elements before collection. Ethics is a design decision rather than a form attached at the end (Saunders et al., 2023, pp. 213–214).

8. Workflow, reporting example and checklist

Derive purpose from the question

Distinguish explanation, description, exploration and evaluation.

Choose a methodological approach

Align it with philosophy and theory development. For mixed methods, specify design and rationale.

Choose a strategy

Justify it against alternatives within available time, resources and access.

Specify the time horizon

Plan safeguards for a cross-sectional survey or measurement occasions and retention for longitudinal work.

Anticipate ethics and constraints

Explain risks, data protection, access, budget and scheduling limits.

Set quality criteria

Choose appropriate standards and address the main threats.

Illustrative methodology paragraph

The study has an explanatory purpose and evaluates the proposed relationship between [predictor] and [outcome] through [mediator]. Consistent with [philosophy] and deduction, it uses a mono-method quantitative questionnaire survey. Cross-sectional data were collected in [month/year] from [n] [participants] in [context]. To reduce common-method concerns, the study [describe the procedure actually used]. The findings are interpreted as associations consistent with the theoretical model rather than evidence of change over time.

Table 7. Common design-reporting problems synthesised from Saunders et al. (2023).
Weak statementProblemBetter approach
Mixed methods because experts edited wordingNo qualitative analysisDescribe quantitative research with expert pretesting where appropriate
A “why” question answered only with descriptive statisticsPurpose and evidence do not alignAdd appropriate explanatory analysis or revise the question
“X increases Y over time” from one waveNo observed temporal changeReport the association within the design's limits
A survey without a rationaleAlternatives not consideredConnect strategy with the question and constraints
Judging interpretive work by statistical representativenessInappropriate quality criteriaConsider credibility, transferability and dependability

Checklist

  • Purpose matches the research question.
  • Methodological choice fits philosophy and theory development.
  • Mixed methods have an explicit sequence and integration rationale.
  • Strategy is justified against the question and practical conditions.
  • Time horizon and conclusions agree.
  • Ethics, access, time and budget are addressed.
  • Quality criteria and safeguards match the design.

Conclusion

A cross-sectional PLS-SEM survey can be a sound design when the question, philosophy and methods align and conclusions remain within the evidence. Describe what was actually done and explain why it was appropriate. An impressive label cannot repair an incoherent design.

References

Main source

  • Saunders, M. N. K., Lewis, P., & Thornhill, A. (2023). Research methods for business students (9th ed.). Pearson. Pearson

Additional sources

  • Bryman, A. (2006). Integrating quantitative and qualitative research: How is it done? Qualitative Research, 6(1), 97–113. https://doi.org/10.1177/1468794106058877
  • Podsakoff, P. M., MacKenzie, S. B., Lee, J.-Y., & Podsakoff, N. P. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88(5), 879–903. https://doi.org/10.1037/0021-9010.88.5.879
  • Spector, P. E. (2019). Do not cross me: Optimizing the use of cross-sectional designs. Journal of Business and Psychology, 34(2), 125–137. https://doi.org/10.1007/s10869-018-09613-8

This overview draws on the listed academic sources. Authors discussed through Saunders et al. (2023) are secondary citations. The reporting paragraph is an illustrative template written for this article.