Research Design – Types, Examples, and Selection Guide

Research design is the structured plan that connects a research question to the evidence needed to answer it. It specifies what or whom the researcher will study, what data will be collected, how variables or concepts will be examined, when and where the research will occur, and how the evidence will be analysed and interpreted.

Introduction

Every credible research project needs more than an interesting topic and a collection method. It needs a logical plan showing how the evidence will answer the research question.

That plan is the research design.

A research design determines whether a study will explore an unfamiliar issue, describe a population, examine relationships, evaluate a programme, explain experiences, test an intervention, or combine several forms of evidence. It also links the research question to the sample, measurements, data-collection procedures, analytical strategy, ethical safeguards, and conclusions.

This guide explains:

• What research design means.
• How it differs from methodology and methods.
• How research designs are classified.
• The main quantitative, qualitative, and mixed-methods designs.
• How to select and justify an appropriate design.
• How to write a research-design section.
• How validity, ethics, digital tools, open science, and artificial intelligence affect modern study planning.

Key Takeaways

• Research design is the overall logical structure of a study, not simply its data-collection method.
• A study can have several compatible design labels, such as “quantitative, descriptive, cross-sectional survey.”
• The research question should determine the design; researchers should not select a favourite method first and force the question to fit it.
• Experimental designs are strongest for estimating causal effects when random assignment is ethical and feasible, but they are not appropriate for every question.
• Qualitative quality depends on coherence, credibility, reflexivity, transparency, and contextual depth rather than statistical generalisation.
• Mixed-methods research requires meaningful integration of qualitative and quantitative components.

What Is Research Design?

Research design is the overall framework used to answer a research question with appropriate and credible evidence. It coordinates the study’s purpose, methodological approach, participants or data sources, measurements, procedures, time frame, comparisons, analysis, and quality safeguards.

Creswell and Creswell describe research design as part of a connected set of decisions involving philosophical assumptions, research approaches, and specific methods (Creswell & Creswell, 2022). The exact terminology varies among disciplines, but the central idea is consistent: design explains how the study will move from a question to a defensible conclusion.

A research design normally answers questions such as:

• What is the study trying to discover, describe, explain, predict, or evaluate?
• What kind of evidence is needed?
• Who or what is the unit of analysis?
• How will cases or participants be selected?
• Will the researcher manipulate an intervention or observe existing conditions?
• Will data be collected once or repeatedly?
• How will concepts or variables be measured?
• What comparisons will be made?
• How will the data be analysed?
• What alternative explanations or sources of bias must be addressed?
• What ethical and practical constraints apply?

Simple research-design example

Suppose a researcher asks:

Is participation in a weekly peer-mentoring programme associated with first-year university students’ academic confidence?

The design might be described as:

• Approach: Quantitative.
• Purpose: Explanatory or associational.
• Intervention status: Observational if students choose whether to participate.
• Time frame: Longitudinal if confidence is measured at the beginning and end of the semester.
• Strategy: Cohort or panel study.
• Data collection: Standardised academic-confidence scale.
• Analysis: Regression or change-score analysis with adjustment for relevant baseline differences.

Calling the project merely “a questionnaire study” would be incomplete because the questionnaire is only the collection instrument.

Why Is Research Design Important?

Research design matters because it determines whether the evidence collected can actually answer the research question. A strong design improves coherence, reduces avoidable bias, anticipates practical problems, and clarifies what conclusions the evidence can and cannot support.

A research design helps researchers:

1. Align the research question with the evidence.
2. Identify the correct participants, cases, texts, records, or datasets.
3. Select suitable measurements and collection procedures.
4. Plan valid comparisons.
5. Anticipate bias, confounding, missing data, and researcher influence.
6. Match the analysis to the type and structure of the data.
7. Address ethics before recruitment or collection begins.
8. Estimate the necessary time, budget, skills, and sample.
9. Document decisions transparently.
10. Avoid making stronger claims than the design permits.

A sophisticated statistical analysis cannot fully repair data that were collected from the wrong population, at the wrong time, with an inappropriate comparison group or invalid measure. Important design decisions therefore need to be made before evidence is collected.

Research Design vs Methodology, Approach, Strategy, and Methods

These terms overlap in ordinary academic writing, and some disciplines use them differently. However, the following distinctions are useful.

Concept	Meaning	Example
Research philosophy or paradigm	Assumptions about knowledge, reality, values, and how inquiry should be conducted	Postpositivism, constructivism, pragmatism, critical theory
Research approach	Broad orientation to the form of evidence and reasoning	Quantitative, qualitative, mixed methods; inductive, deductive, abductive
Research methodology	The rationale and principles guiding how the inquiry is conducted	A qualitative interpretive methodology or a quantitative hypothesis-testing methodology
Research design	The logical structure connecting the question to the evidence and conclusion	Cross-sectional correlational design, randomised experiment, phenomenological design
Research strategy or tradition	A recognisable way of organising inquiry	Survey, case study, ethnography, grounded theory, experiment
Research method	A specific procedure used to collect or analyse data	Questionnaire, interview, observation, regression, thematic analysis
Research instrument	The actual tool used to obtain or record data	Scale, test, interview guide, observation checklist, sensor

Is a survey a research design?

A survey can refer to a broader research strategy or to a collection procedure. It does not provide a complete design description by itself.

For example:

• A descriptive cross-sectional survey estimates current attitudes.
• A longitudinal panel survey tracks changes among the same participants.
• A survey experiment randomly assigns respondents to different messages.
• A mixed-methods survey combines structured responses with a qualitative component that is separately analysed and integrated.

Is qualitative research a design?

Qualitative research is usually best understood as a broad approach. A complete description should identify the qualitative design or tradition, such as phenomenology, ethnography, grounded theory, narrative inquiry, qualitative case study, or participatory action research.

Elements of Research Design

A complete research design generally includes the following interconnected elements.

1. Research problem and rationale

The design begins with a clearly defined problem. The problem should identify what is unknown, uncertain, contested, ineffective, or insufficiently understood.

2. Aim, objectives, and research questions

The aim states the overall purpose. Objectives divide that purpose into achievable tasks. Research questions specify exactly what the study will answer.

Questions should precede major design decisions. A question asking “How do patients experience remote consultation?” requires a different design from “Does remote consultation reduce waiting time?”

3. Conceptual or theoretical framework

A framework identifies the main concepts, expected relationships, interpretive assumptions, or theoretical perspective. It helps explain why particular variables, cases, or experiences are being examined.

Not every exploratory study begins with a formal theory, but every design rests on assumptions that should be recognised.

4. Methodological approach

The researcher determines whether qualitative, quantitative, mixed-methods, or another specialised approach best fits the question.

5. Unit of analysis

The unit of analysis is the entity about which conclusions will be made. It may be:

• Individuals.
• Households.
• Schools.
• Organisations.
• Countries.
• Events.
• Documents.
• Online communities.
• Policies.
• Conversations.
• Published studies.

Confusing the unit of observation with the unit of analysis can produce invalid conclusions. For example, individual students may provide responses, while the primary comparison is between schools.

6. Population, setting, and sample

The design identifies:

• The target population.
• The accessible population.
• The setting.
• Inclusion and exclusion criteria.
• Sampling strategy.
• Expected sample size.
• Recruitment process.

Quantitative studies may require a power or precision calculation. Qualitative sampling is usually justified through information needs, diversity, theoretical relevance, or depth rather than statistical representativeness.

7. Constructs, variables, or phenomena

Quantitative designs define exposures, interventions, outcomes, predictors, confounders, mediators, moderators, and control variables.

Qualitative designs identify the experience, process, culture, case, narrative, discourse, or phenomenon being investigated.

8. Operationalisation and measurement

Operationalisation explains how an abstract concept will be observed or measured.

For example, “academic engagement” might be represented through:

• Attendance records.
• Learning-platform activity.
• A validated engagement scale.
• Classroom observations.
• Student interviews.

These measures capture different aspects of engagement and should not be treated as automatically interchangeable.

9. Data-collection procedures

The design specifies:

• What will be collected.
• By whom.
• From where.
• In what sequence.
• Under what conditions.
• Using which instruments.
• With what quality controls.

10. Time horizon

The study may be:

• Cross-sectional.
• Repeated cross-sectional.
• Longitudinal.
• Prospective.
• Retrospective.
• Concurrent.
• Sequential.

11. Comparison and control

Analytical studies need a defensible comparison. This might involve:

• Experimental and control groups.
• Exposed and unexposed participants.
• Cases and controls.
• Before-and-after measurements.
• Different sites.
• Contrasting cases.
• Alternative interpretations.
• Qualitative negative cases.

12. Analysis plan

The analysis should be planned with the question and data structure in mind. It may include:

• Descriptive statistics.
• Hypothesis tests.
• Regression.
• Multilevel modelling.
• Time-series analysis.
• Thematic analysis.
• Framework analysis.
• Narrative analysis.
• Discourse analysis.
• Grounded-theory coding.
• Mixed-methods integration.

13. Quality safeguards

The design should address validity, reliability, trustworthiness, bias, confounding, reflexivity, missing data, measurement error, and analytical transparency.

14. Ethics and data governance

Ethical planning may include:

• Informed consent.
• Privacy and confidentiality.
• Risk minimisation.
• Fair participant selection.
• Safeguarding.
• Data security.
• Withdrawal procedures.
• Use of sensitive or secondary data.
• Institutional ethics review.
• Responsible use of AI and online platforms.

15. Feasibility

An ideal design that cannot be implemented is not a useful design. Time, access, budget, sample availability, researcher skills, equipment, software, and institutional requirements must be considered.

How Research Designs Are Classified

There is no single universally accepted list of research designs. Different classifications answer different questions.

The clearest approach is to treat research design as a stack of compatible decisions.

Classification by Research Purpose

Exploratory research design

Exploratory research investigates a poorly understood problem in order to generate insights, concepts, questions, or hypotheses.

It is appropriate when:

• The problem is not clearly defined.
• Existing theory is limited.
• Relevant variables are uncertain.
• Researchers need to understand language, context, or possible mechanisms.

Common methods include literature exploration, interviews, focus groups, observation, document analysis, and small pilot studies.

Example: Interviewing first-generation university students to identify previously undocumented barriers to using academic-support services.

Exploratory findings should not usually be presented as definitive population estimates or causal conclusions.

Descriptive research design

Descriptive research systematically documents the characteristics, distribution, frequency, or development of a phenomenon.

It answers questions such as:

• What is happening?
• Who is affected?
• How common is it?
• Where does it occur?
• How does it vary?
• How has it changed?

Example: Estimating the percentage of postgraduate students who use generative AI tools for literature searching.

Descriptive research may use surveys, observations, records, content analysis, or qualitative description.

Correlational or associational design

Correlational research examines whether variables vary together without assigning the exposure or intervention.

Example: Examining whether weekly study time is associated with examination performance.

Correlation alone does not establish causation. Alternative explanations may include confounding, reverse causality, measurement error, and selection bias.

Explanatory research design

Explanatory research investigates why or how a pattern, process, outcome, or relationship occurs.

It may use:

• Statistical modelling.
• Longitudinal evidence.
• Process tracing.
• Comparative case studies.
• Mechanism-focused interviews.
• Experiments or quasi-experiments.

Explanatory research is broader than experimental research. Explanation may concern meanings, processes, structures, mechanisms, or causal effects.

Causal research design

Causal research estimates whether a change in one factor produces a change in an outcome under specified conditions.

A credible causal design must consider:

• Temporal order.
• Counterfactual comparison.
• Confounding.
• Selection.
• Measurement.
• Alternative explanations.

Randomised experiments are powerful causal designs because random assignment can balance observed and unobserved participant characteristics on average. Quasi-experimental designs seek credible causal comparisons without random assignment (Shadish et al., 2002).

Evaluative research design

Evaluative research judges the design, implementation, effectiveness, efficiency, value, or consequences of a policy, intervention, service, or programme.

Evaluation may address:

• Whether the programme works.
• For whom it works.
• Under what conditions.
• How it is implemented.
• Whether benefits justify costs.
• What unintended effects occur.

Predictive research design

Predictive research develops or tests models that estimate future or unknown outcomes.

Prediction and causal explanation are different goals. A variable may improve prediction without being a cause, and a causal factor may contribute little additional predictive accuracy.

Classification by Methodological Approach

Quantitative research design

Quantitative designs use numerical evidence to estimate quantities, compare groups, test hypotheses, model associations, evaluate interventions, or make predictions.

They usually involve:

• Predefined variables.
• Structured measurement.
• Numerical analysis.
• Explicit sampling and comparison rules.
• Statistical uncertainty.

Qualitative research design

Qualitative designs investigate meanings, experiences, processes, interactions, cases, cultures, narratives, or contexts through non-numerical evidence.

They typically emphasise:

• Depth.
• Context.
• Participant perspectives.
• Interpretation.
• Reflexivity.
• Flexible but systematic inquiry.

Mixed-methods research design

Mixed-methods research combines qualitative and quantitative components and deliberately integrates them to answer a shared or connected set of questions.

Integration may occur by:

• Connecting one phase to another.
• Building an instrument from qualitative findings.
• Comparing findings.
• Embedding one component within another.
• Merging datasets or interpretations.
• Producing joint conclusions or meta-inferences.

Merely collecting two forms of data is not sufficient if the components remain unrelated (Fetters et al., 2013).

Classification by Researcher Intervention

Experimental design

In an experiment, the researcher assigns or manipulates an intervention and observes its effect.

A true experiment normally includes:

• Manipulation of an independent variable.
• A comparison condition.
• Random assignment.
• Standardised outcome measurement.

Blinding, allocation concealment, adherence monitoring, and attrition management may also be important.

Quasi-experimental design

A quasi-experiment evaluates an intervention without random assignment.

Common quasi-experimental designs include:

• Non-equivalent control-group design.
• Interrupted time series.
• Regression discontinuity.
• Difference-in-differences.
• Controlled before-and-after study.
• Natural experiment.

These designs can support causal inference when their assumptions are plausible and alternative explanations are carefully addressed.

Observational design

In an observational study, the researcher does not assign the exposure or intervention.

Examples include:

• Cross-sectional studies.
• Cohort studies.
• Case-control studies.
• Ecological studies.
• Observational longitudinal studies.

Observational designs are essential when manipulation is unethical, impossible, or irrelevant. Their main challenges often include confounding, selection bias, information bias, and uncertainty about temporal order.

Classification by Time

Cross-sectional design

A cross-sectional study examines a population or sample at one point or short period in time.

It is useful for:

• Estimating prevalence.
• Describing current characteristics.
• Examining simultaneous associations.
• Comparing groups at a particular time.

Its main limitation is weak evidence about temporal order.

Longitudinal design

A longitudinal study examines change over time.

Forms include:

• Panel study: Repeatedly studies the same participants.
• Cohort study: Follows a group sharing a characteristic or exposure.
• Repeated cross-sectional study: Draws different samples from the same population over time.
• Time-series design: Uses repeated measurements across many time points.

Longitudinal research can clarify sequence and change but may face attrition, measurement drift, long duration, and high cost.

Prospective design

A prospective study defines the sample and follows events forward from the beginning of the study.

Retrospective design

A retrospective study uses events, exposures, or records that already exist when the study is designed.

Retrospective research may be efficient but is constrained by the quality, completeness, and original purpose of existing records.

Classification by Data Source

Primary research

The researcher collects original data specifically for the study.

Secondary-data research

The researcher analyses data previously collected for another purpose, such as administrative records, surveys, archives, repositories, social-media data, or public datasets.

Evidence synthesis

Evidence-synthesis designs analyse existing studies rather than individual participants. They include:

• Systematic reviews.
• Scoping reviews.
• Meta-analyses.
• Qualitative evidence syntheses.
• Umbrella reviews.
• Rapid reviews.
• Realist reviews.

The review question should determine the review design. PRISMA 2020 provides reporting guidance for systematic reviews but is not a substitute for a complete review methodology (Page et al., 2021).

Major Quantitative Research Designs

Descriptive survey design

A descriptive survey collects structured responses to estimate characteristics, attitudes, behaviours, or conditions in a defined population.

Example: Measuring the prevalence of research anxiety among doctoral students.

Important decisions include:

• Sampling frame.
• Response rate.
• Non-response bias.
• Instrument validity.
• Weighting.
• Mode of administration.

Correlational design

A correlational design measures two or more variables and examines their statistical association.

Example: Examining the relationship between sleep duration and academic performance.

Researchers should avoid describing an association as an “effect” unless the design and assumptions support causal inference.

Randomised experimental design

Participants or units are randomly assigned to conditions.

Example: Randomly assigning classes to receive either a new instructional method or standard teaching, then comparing learning outcomes.

Variations include:

• Parallel-group design.
• Cluster-randomised design.
• Crossover design.
• Factorial design.
• Stepped-wedge design.

CONSORT 2025 should be consulted when reporting randomised trials.

Quasi-experimental design

A quasi-experimental design estimates intervention effects without random assignment.

Example: Comparing changes in attendance before and after a new policy in one district against changes in a similar district without the policy.

Credibility depends on the comparison strategy and assumptions, not merely on applying an advanced statistical test.

Cohort design

A cohort design follows exposed and unexposed groups to compare later outcomes.

Example: Following employees with high and low levels of remote-working exposure to examine changes in job satisfaction.

Cohorts may be prospective or retrospective.

Case-control design

A case-control study begins with an outcome.

Researchers compare people or units with the outcome to those without it and examine prior exposures.

Example: Comparing previous workplace conditions among employees who did and did not experience a specified occupational injury.

This design is efficient for rare outcomes but may be vulnerable to selection and recall bias.

Cross-sectional analytical design

A cross-sectional analytical study measures possible exposures and outcomes at approximately the same time.

It can identify associations but often cannot establish which came first.

Longitudinal panel design

The same participants provide data repeatedly.

Example: Surveying the same students every semester to study changes in academic belonging.

Repeated observations are statistically dependent, so the analysis must account for the data structure.

Major Qualitative Research Designs

Qualitative case study

A case study investigates a bounded contemporary case in depth and in context.

The case might be:

• One school.
• A programme.
• An organisation.
• A community.
• An event.
• A policy implementation.
• Several contrasting cases.

Case studies may use interviews, observations, documents, records, images, and quantitative evidence.

The researcher must define the boundaries of the case: what is included, excluded, where, and during which period.

Phenomenological design

Phenomenology investigates how people experience and make sense of a phenomenon.

Example: Exploring the lived experience of doctoral candidates returning to study after a long career break.

Researchers seek depth across participants who have directly experienced the phenomenon.

Grounded-theory design

Grounded theory develops an explanatory theory or process model from systematically collected and analysed data.

Example: Developing a theory of how novice teachers learn to manage classroom uncertainty.

Data collection and analysis proceed iteratively. Sampling may become increasingly theoretical as emerging categories are refined.

Grounded theory should not be claimed merely because themes emerged from interviews.

Ethnographic design

Ethnography investigates the practices, meanings, interactions, and culture of a group or social setting.

Example: Studying how members of an online open-source community negotiate authority and expertise.

Ethnography usually requires sustained engagement, observation, reflexive field notes, and attention to context.

Narrative research

Narrative research studies how individuals or groups construct and communicate experience through stories.

It may examine:

• Life histories.
• Career narratives.
• Illness narratives.
• Institutional stories.
• Identity development.

Analysis considers both what is narrated and how the narrative is structured and situated.

Qualitative descriptive design

Qualitative description provides a close, accessible account of participants’ views or experiences without necessarily adopting a more specialised tradition.

It is useful for practice-oriented questions where the goal is a clear summary rather than theory generation or phenomenological interpretation.

Action research

Action research combines inquiry with cycles of planning, action, observation, reflection, and improvement.

It is commonly used by teachers, practitioners, organisations, and communities.

Participatory and community-based research

Participatory designs involve affected individuals or communities in decisions about the question, design, data, interpretation, and use of findings.

Participation should involve meaningful influence rather than symbolic consultation.

Mixed-Methods Research Designs

Convergent design

Qualitative and quantitative data are collected during a similar phase, analysed separately, and then compared or merged.

Example: Combining a staff survey with interviews to examine employee experiences of hybrid work.

The design must explain how agreement, complementarity, and contradiction will be interpreted.

Explanatory sequential design

Quantitative research occurs first. Qualitative research then helps explain important, unexpected, or unclear numerical findings.

Example:

1. Survey students about online-learning satisfaction.
2. Identify unusually high- and low-satisfaction groups.
3. Interview selected students.
4. Use the interviews to explain the survey patterns.

Exploratory sequential design

Qualitative research occurs first. Its findings inform a later quantitative phase.

Example:

1. Interview patients about barriers to using a health portal.
2. Develop a questionnaire from the qualitative categories.
3. Test the questionnaire in a larger sample.
4. Estimate how common the barriers are.

Embedded design

One methodological component is placed inside a larger design.

For example, interviews may be embedded within a randomised trial to examine implementation, acceptability, or participant experiences.

Multiphase design

Several connected qualitative, quantitative, or mixed-methods studies are conducted across a programme of research.

This may be appropriate for developing, testing, implementing, and evaluating a complex intervention.

Research-Design Comparison Table

Design	Main purpose	Typical question	Major strength	Major limitation
Exploratory	Generate understanding or hypotheses	What factors may be involved?	Flexible and discovery-oriented	Does not normally provide definitive estimates
Descriptive	Document characteristics or frequency	What is happening and how common is it?	Produces a systematic profile	Does not by itself explain causation
Correlational	Examine associations	Are X and Y related?	Studies naturally occurring variation	Confounding and reverse causality
Randomised experiment	Estimate causal effects	Does intervention X change outcome Y?	Strong counterfactual comparison	May be unethical, costly, or artificial
Quasi-experiment	Estimate effects without random assignment	Did a policy produce a change?	Useful in real-world settings	Relies on design-specific assumptions
Cross-sectional	Describe one time point	What is the current prevalence?	Efficient and relatively fast	Weak temporal evidence
Longitudinal	Examine change and sequence	How does Y change over time?	Establishes temporal ordering more clearly	Attrition, time, and cost
Case study	Understand a bounded case	How and why did this occur in this case?	Rich contextual understanding	Limited basis for statistical generalisation
Phenomenology	Understand lived experience	What is it like to experience X?	Deep experiential insight	Requires careful interpretive discipline
Grounded theory	Develop theory from data	What process explains X?	Produces an empirically grounded model	Resource-intensive and frequently misapplied
Ethnography	Understand culture and practice	How does this group create meaning?	Captures behaviour in context	Demands access and sustained engagement
Mixed methods	Integrate complementary evidence	What is happening, and why?	Combines breadth and depth	Requires integration expertise and resources
Systematic review	Synthesise existing studies	What does the total evidence show?	Transparent evidence synthesis	Limited by included evidence and review quality

How to Choose a Research Design

Step 1: Start with the research question

Identify the central verb or purpose.

• Explore: Exploratory qualitative or mixed-methods design.
• Describe: Descriptive survey, observational study, qualitative description, or case study.
• Compare: Comparative observational, experimental, quasi-experimental, or comparative-case design.
• Test an intervention: Experimental or quasi-experimental design.
• Understand experience: Phenomenology, narrative research, or qualitative description.
• Develop theory: Grounded theory.
• Study culture or practice: Ethnography.
• Examine change: Longitudinal, cohort, panel, or time-series design.
• Synthesise evidence: Systematic, scoping, realist, or other review design.

Step 2: Determine the type of conclusion needed

Ask whether the project needs:

• Description.
• Interpretation.
• Association.
• Prediction.
• Causal estimation.
• Theory development.
• Evaluation.
• Evidence synthesis.

Do not use causal language when the design can support only association.

Step 3: Assess existing knowledge

When little is known, exploration may be necessary before structured measurement.

When concepts and measures are well established, a confirmatory quantitative design may be appropriate.

When numerical patterns are known but poorly understood, explanatory sequential mixed methods may help.

Step 4: Decide whether intervention is possible

Ask:

• Can the exposure be manipulated?
• Would manipulation be ethical?
• Can participants or sites be randomly assigned?
• Is a comparison group available?
• Is a natural policy change creating a quasi-experimental opportunity?

Step 5: Choose the time structure

Determine whether one observation is enough or whether change, sequence, development, or delayed outcomes require repeated measurement.

Step 6: Define the population, cases, and sampling logic

The sampling strategy should match the intended conclusion.

• Population estimates require a defensible sampling frame.
• Causal studies require credible comparison and sufficient sample size.
• Qualitative studies require information-rich, relevant participants or cases.
• Multilevel questions may require sampling at more than one level.

Step 7: Match data collection and analysis

Plan collection and analysis together.

For example:

• A comparison of average scores may require numerical measurement and group-comparison analysis.
• A process question may require repeated interviews, observations, or documents.
• A nested school dataset may require multilevel analysis.
• A mixed-methods design requires an explicit integration procedure.

Step 8: Evaluate validity, ethics, and feasibility

Before finalising the design, ask:

• What are the most serious threats to the intended conclusion?
• Can they be prevented or measured?
• Is the design ethically acceptable?
• Is recruitment realistic?
• Are the instruments available and appropriate?
• Does the team have the necessary analytical expertise?
• Can the project be completed with available time and resources?

Worked Research-Design Examples

Example 1: Undergraduate education study

Question: What proportion of final-year students experience high research anxiety, and which academic factors are associated with it?

Appropriate design:

• Quantitative.
• Descriptive and correlational.
• Cross-sectional.
• Survey strategy.
• Primary data.

Participants: Final-year students selected through stratified sampling across departments.

Data: Validated research-anxiety scale, supervision frequency, prior methods training, academic discipline, and demographic variables.

Analysis: Prevalence estimates, group comparisons, and multivariable regression.

Limitation: Because exposure and outcome are measured at the same time, temporal direction may remain uncertain.

Example 2: Qualitative health study

Question: How do adults newly diagnosed with tinnitus experience the first six months after diagnosis?

Appropriate design:

• Qualitative.
• Longitudinal.
• Phenomenological or experience-focused design.

Participants: Purposively selected adults with recent diagnoses.

Data: Interviews shortly after diagnosis and again after three and six months, supported by participant diaries.

Analysis: Interpretive thematic or phenomenological analysis of experience and change over time.

Quality considerations: Reflexive notes, transparent coding, attention to negative cases, and a clear account of researcher positioning.

Example 3: Policy evaluation

Question: Did introducing free school breakfast improve attendance?

Appropriate design:

• Quantitative.
• Evaluative and causal.
• Quasi-experimental.
• Difference-in-differences or controlled interrupted time-series design.

Data: Attendance records before and after implementation from participating and suitable comparison schools.

Analysis: Estimate whether attendance changed more in intervention schools than in comparison schools.

Critical assumption: In the absence of the programme, the groups would have followed sufficiently similar trends.

Example 4: Mixed-methods technology study

Question: Does an AI-supported feedback tool improve student writing, and how do students experience its feedback?

Appropriate design:

• Mixed methods.
• Experimental or quasi-experimental quantitative component.
• Embedded or explanatory qualitative component.

Quantitative phase: Compare writing improvement between students using the tool and those receiving standard feedback.

Qualitative phase: Interview selected students about trust, usefulness, confusion, dependence, and revision decisions.

Integration: Compare performance outcomes with participant explanations and examine whether experiences help explain variation in improvement.

How to Write the Research-Design Section

A research-design section should identify the design, explain why it fits the question, and show how its components work together.

Research-design template

This study will use a [qualitative/quantitative/mixed-methods] [specific design] to investigate [research problem or question]. The design is appropriate because the study aims to [explore/describe/compare/explain/evaluate/test/develop]. The unit of analysis will be [unit], and the study will be conducted in [setting] over [time frame]. Participants or cases will be selected using [sampling strategy]. Data will be collected through [methods and instruments] and analysed using [analysis]. The design addresses [key validity or trustworthiness concerns] through [safeguards]. Ethical issues involving [consent, privacy, risk, access, or data governance] will be managed by [procedures].

Example paragraph

This study will use a quantitative descriptive cross-sectional survey design to estimate research-methods anxiety among final-year undergraduate students. This design is appropriate because the objective is to describe the current distribution of anxiety and examine associations with prior methods training and supervision frequency rather than evaluate an intervention. Students will be sampled from four academic faculties using proportionate stratified sampling. Data will be collected through an anonymous online questionnaire containing a validated anxiety scale and analysed using descriptive statistics and multivariable regression. Possible non-response bias will be assessed by comparing early and late respondents where feasible.

A good justification explains why the design is appropriate. It should not merely state that the design is “easy,” “popular,” or “commonly used.”

Validity, Reliability, and Trustworthiness

Internal validity

Internal validity concerns whether the observed relationship can credibly be attributed to the proposed explanation within the study.

Threats include:

• Selection differences.
• Confounding.
• History.
• Maturation.
• Testing effects.
• Instrument changes.
• Regression to the mean.
• Attrition.
• Contamination.
• Researcher or participant expectations.

External validity

External validity concerns whether findings may apply beyond the studied participants, setting, intervention, or period.

Generalisability is not an all-or-nothing property. Researchers should specify the population, settings, treatments, and outcomes to which an inference is intended to apply.

Not every qualitative study aims for statistical generalisation. Qualitative researchers may instead support transferability by providing enough contextual detail for readers to judge relevance elsewhere.

Construct validity

Construct validity concerns whether the operational measures and procedures adequately represent the intended concept.

A study of “digital literacy,” for example, should not assume that frequency of internet use fully represents digital literacy.

Statistical-conclusion validity

This concerns whether the numerical analysis supports an accurate conclusion about the existence, magnitude, and uncertainty of a relationship.

Threats include:

• Insufficient sample size.
• Violated assumptions.
• Multiple unplanned analyses.
• Unreliable measures.
• Incorrect model specification.
• Selective reporting.
• Failure to account for clustering or repeated observations.

Reliability

Reliability concerns consistency in measurement or procedure.

Relevant forms include:

• Internal consistency.
• Test–retest reliability.
• Inter-rater reliability.
• Procedural consistency.

Reliability alone does not establish validity. A measure can produce consistent but systematically incorrect results.

Qualitative trustworthiness

Qualitative researchers may consider:

• Credibility: Are interpretations well supported?
• Dependability: Is the process transparent and logically documented?
• Confirmability: Are interpretations grounded in evidence rather than hidden assumptions?
• Transferability: Is context described sufficiently?
• Reflexivity: Has the researcher considered how their role and assumptions shaped the inquiry?
• Authenticity: Are different participant positions represented responsibly?

Techniques may include prolonged engagement, triangulation, negative-case analysis, peer discussion, audit trails, reflexive journaling, and careful use of participant feedback.

No single technique automatically guarantees quality. The safeguards must fit the design and epistemological approach.

Mixed-methods quality

Mixed-methods quality requires more than separate qualitative and quantitative rigour. Researchers should also evaluate:

• Whether mixed methods was necessary.
• Whether the components answer connected questions.
• Whether the sequence is justified.
• Whether sampling across components is coherent.
• Where and how integration occurs.
• How contradictions are interpreted.
• Whether the final conclusions genuinely reflect both components.

Advantages of a Clear Research Design

A clear design:

• Connects the question to the evidence.
• Makes assumptions visible.
• Guides sampling and measurement.
• Prevents avoidable analytical problems.
• Clarifies ethical responsibilities.
• Improves project management.
• Supports transparent reporting.
• Helps reviewers evaluate the credibility of conclusions.
• Makes limitations easier to recognise.
• Facilitates replication, extension, or evidence synthesis where appropriate.

Limitations of Research Design

Research design does not eliminate every problem.

• Real-world constraints may prevent the theoretically strongest design.
• Every design answers some questions better than others.
• Randomised experiments may have limited relevance outside controlled conditions.
• Observational studies may face residual confounding.
• Qualitative findings depend on context and interpretive transparency.
• Longitudinal studies face attrition and changing conditions.
• Mixed-methods studies require more time, skills, and integration.
• Existing-data studies are constrained by variables collected for other purposes.
• Adaptive design changes can complicate interpretation.
• An apparently rigorous design can still fail through poor implementation.

Researchers should therefore justify trade-offs rather than claim that a design is flawless.

Common Research-Design Mistakes

Choosing the method before the question

Beginning with “I want to conduct interviews” or “I want to use a questionnaire” reverses the correct logic.

Begin with the problem and question.

Giving an incomplete design label

“Quantitative design” or “survey design” is rarely sufficient.

A stronger description might be:

Quantitative descriptive cross-sectional survey design.

Treating classification dimensions as competing options

A study does not have to choose between being longitudinal, correlational, and quantitative. It can be all three.

Confusing correlation with causation

A statistically significant association does not by itself show that one variable caused another.

Claiming mixed methods without integration

Two disconnected datasets do not automatically form a mixed-methods design.

Selecting analysis after seeing the results

Unplanned analytical flexibility can increase the risk of selective reporting and misleading findings. Confirmatory analyses should be distinguished from exploratory analyses.

Ignoring the unit of analysis

Analysing individual responses while making unsupported claims about organisations, schools, or countries creates an ecological or aggregation problem.

Using an invalid or inappropriate instrument

A popular scale may not be valid for the target language, culture, population, administration mode, or purpose.

Failing to plan for missing data and attrition

Missingness should be anticipated in recruitment, measurement, follow-up, and analysis plans.

Treating sample size as the only quality criterion

A large biased sample does not automatically produce credible conclusions. Sampling method, measurement quality, comparison structure, and implementation remain important.

Overstating generalisability

Researchers should identify the population and conditions to which the study reasonably applies.

Ignoring feasibility

A design requiring inaccessible participants, unavailable records, unaffordable software, or several years of follow-up may need to be redesigned.

Research Design in Modern Research

Modern research design increasingly includes decisions about transparency, digital environments, data governance, and reproducibility.

Preregistration

Preregistration creates a time-stamped record of research questions, hypotheses, design decisions, data-collection procedures, and planned analyses before outcomes are known.

It can help distinguish confirmatory decisions from later exploratory work. It does not prevent legitimate changes, but changes should be documented and explained.

Registered reports

In a registered report, the question and proposed methods may be peer reviewed before data collection or before outcomes are known. Publication decisions are therefore less dependent on whether the results are statistically significant or surprising.

Protocols

Complex studies benefit from a detailed protocol covering:

• Objectives.
• Eligibility.
• Recruitment.
• Interventions.
• Outcomes.
• Analysis.
• Data management.
• Ethics.
• Deviations.
• Dissemination.

Reporting guidelines

Reporting guidelines help researchers disclose essential design and implementation information.

Examples include:

• CONSORT for randomised trials.
• STROBE for observational studies.
• PRISMA for systematic reviews.
• COREQ and SRQR for qualitative research.
• SPIRIT for trial protocols.
• APA Journal Article Reporting Standards for quantitative, qualitative, and mixed-methods studies.

A reporting checklist does not create a good design. However, consulting the relevant guideline during planning can reveal decisions that need to be specified.

Reproducible workflows

Depending on the field and ethical constraints, researchers may preserve:

• Analysis code.
• Data dictionaries.
• Version histories.
• Instrument versions.
• Decision logs.
• Search strategies.
• Anonymised or synthetic datasets.
• Computational environments.
• Qualitative coding frameworks.

Transparency must be balanced with participant privacy, legal restrictions, intellectual property, and community agreements.

Online and digitally mediated research

Digital research may involve:

• Online surveys.
• Remote experiments.
• Video interviews.
• Digital ethnography.
• Platform data.
• Mobile sensors.
• Learning analytics.
• Electronic health records.
• Social-media archives.

The convenience of digital collection does not remove sampling, consent, identity-verification, platform-bias, accessibility, security, and context concerns.

Artificial Intelligence and Research Design

Artificial intelligence can assist research planning, but it should not make unreviewed methodological decisions.

Potential uses include:

• Generating alternative question formulations.
• Identifying possible variables and confounders.
• Drafting interview prompts for expert review.
• Testing survey wording.
• Supporting literature screening.
• Checking code.
• Producing simulation scenarios.
• Assisting transcription or translation.
• Organising documented research decisions.

Important risks include:

• Fabricated references.
• Incorrect methodological advice.
• Hidden training-data biases.
• Disclosure of confidential information.
• Unreliable coding or classification.
• Loss of context.
• Automation bias.
• Unclear reproducibility when models change.
• Intellectual-property and data-protection problems.

Responsible practice requires human accountability, verification, transparency, appropriate disclosure, and compliance with institutional, funder, publisher, and legal requirements. AI-generated recommendations should be treated as unverified suggestions rather than methodological authority.

Researchers should document, where relevant:

• The tool and model used.
• The date or version.
• The task performed.
• The inputs provided.
• Human review procedures.
• Whether outputs affected design decisions.
• How privacy and confidentiality were protected.

Practical Research-Design Checklist

Before approving a design, confirm that:

• The problem and question are clear.
• The intended conclusion is explicitly stated.
• The design fits that conclusion.
• The unit of analysis is correct.
• The population and setting are defined.
• The sampling strategy is justified.
• Measures represent the intended constructs.
• Data collection and analysis align.
• The comparison strategy is credible.
• Time order is adequate for the claim.
• Bias and confounding have been considered.
• Qualitative reflexivity or quantitative assumptions are addressed.
• Mixed-methods integration is explicit where relevant.
• Ethics and data governance are covered.
• The study is feasible.
• Deviations can be documented.
• A relevant reporting guideline has been identified.
• Claims will remain within what the design can support.

Conclusion

Research design is the logical architecture of a study. It is not simply a label such as qualitative, quantitative, survey, or interview-based. A complete design connects the research question to the population or cases, sampling, measurement, procedures, time frame, comparison strategy, analysis, validity safeguards, ethics, and intended conclusion.

The most appropriate design is the one that answers the research question credibly within the study’s ethical and practical constraints. Researchers should therefore begin with the question, identify the type of conclusion required, compare realistic design options, and justify the final choice transparently.

References

1. Creswell, J. W., & Creswell, J. D. (2022). Research design: Qualitative, quantitative, and mixed methods approaches (6th ed.). SAGE Publications. SAGE describes this edition as covering research questions, proposals, and qualitative, quantitative, and mixed-methods approaches.
2. Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs for generalized causal inference. Houghton Mifflin. This remains an important foundational source for experimental and quasi-experimental causal designs.
3. Fetters, M. D., Curry, L. A., & Creswell, J. W. (2013). Achieving integration in mixed methods designs—Principles and practices. Health Services Research, 48(6, Pt. 2), 2134–2156. https://doi.org/10.1111/1475-6773.12117. The article explains integration through convergent, explanatory sequential, exploratory sequential, embedded, and other mixed-method structures.
4. EQUATOR Network. Use its searchable library to identify reporting guidance by study type. It maps randomised trials to CONSORT, observational studies to STROBE, systematic reviews to PRISMA, protocols to SPIRIT or PRISMA-P, and qualitative studies to SRQR or COREQ.
5. CONSORT 2025. The updated guideline for reporting randomised trials was published in April 2025.
6. von Elm, E., Altman, D. G., Egger, M., Pocock, S. J., Gøtzsche, P. C., & Vandenbroucke, J. P. (2007). The Strengthening the Reporting of Observational Studies in Epidemiology statement: Guidelines for reporting observational studies. PLoS Medicine, 4(10), e296. Use the official STROBE resources indexed by EQUATOR.
7. Page, M. J., McKenzie, J. E., Bossuyt, P. M., et al. (2021). The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ, 372, n71. https://doi.org/10.1136/bmj.n71.
8. Tong, A., Sainsbury, P., & Craig, J. (2007). Consolidated criteria for reporting qualitative research: A 32-item checklist for interviews and focus groups. International Journal for Quality in Health Care, 19(6), 349–357. https://doi.org/10.1093/intqhc/mzm042. COREQ remains listed by EQUATOR for qualitative reporting.