Research Design: Choosing the Right Method for Your Research Question

Start with the decision the study must inform

A research design is the argument connecting a question, the evidence needed to answer it and the conclusions that evidence can reasonably support. Begin by writing the decision or knowledge problem in plain language. An institution asking why students discontinue a course needs a different design from one estimating how many students leave or testing whether a particular support programme changes completion.

Write one primary question, then identify the population, setting, phenomenon or intervention and time period. A broad interest such as digital learning becomes researchable when narrowed to a defined experience or outcome. The purpose is not to make the project small for its own sake; it is to establish what evidence would count as a useful answer.

Separate the main types of question

Descriptive questions examine what occurs, how often or under what conditions. Exploratory questions investigate meanings and experiences that are not yet well understood. Comparative questions examine differences between groups or settings. Evaluative questions assess implementation, outcomes or value. Causal questions ask whether a change would occur because of an intervention rather than other influences.

These purposes can coexist, but each needs its own evidence. Interviews may explain how learners experience feedback without estimating the frequency of those experiences across a university. Administrative records may describe completion patterns without explaining students' reasons for leaving. Treating either source as an answer to every question creates avoidable overstatement.

Build a question-to-evidence map

For each question, record the unit of analysis, relevant evidence, collection method, analysis approach and major limitation. Units might be students, classes, institutions, documents or partnerships. Clarify whether repeated observations belong to the same person or whether participants share a classroom, because the structure affects analysis.

Next, decide whether evidence should be collected once or over time. Existing records can be valuable if their definitions, completeness and permissions fit the question. A new survey should have a specific purpose rather than being the default choice. NIH's research methods resources highlight that design decisions need to reflect the structure of the study and its methodological challenges. ^[1]

Assess feasibility without weakening the question silently

List access, staffing, language, time, ethical review, data security and analytic expertise requirements. If the preferred design is infeasible, revise the question explicitly. A modest study with a defensible claim is more useful than an ambitious question answered with unsuitable evidence.

Consider how recruitment will work and who may be excluded. Identify what happens if participation is lower than expected, a partner withdraws or key records are incomplete. Agree which changes require supervisor, partner or ethics approval. Keep a decision log so later readers can understand why the final design differs from the original proposal.

Illustrative example: evaluating a mentoring initiative

Suppose a college introduces voluntary mentoring for first-year students. Comparing participating students with non-participants at the end of the year would be easy, but motivation, prior achievement and available time may differ between groups. A positive association would not, by itself, establish that mentoring caused improvement.

The team could first examine implementation: who joined, how often meetings occurred and what support was delivered. Interviews could explore barriers to participation. If causal impact is the main question, the team should seek appropriate methodological support for a credible comparison strategy, potentially including random allocation where ethical and practical. If that is unavailable, the report should present the study as an observational evaluation and describe remaining uncertainties.

Review the design before collecting data

• Can each research question be answered by the proposed evidence?
• Are recruitment and inclusion criteria clear?
• Are measures suitable for the population and language?
• Are foreseeable alternative explanations addressed?
• Are consent, data access and storage arrangements workable?
• Can the team perform the proposed analysis with available expertise?
• Have changes, exclusions and reporting responsibilities been agreed?

Reporting guidance can help expose missing design information early. EQUATOR organises reporting guidance by study type, although many resources originate in health research and should be applied with attention to disciplinary context. ^[2]

State the limits of the conclusion

End the protocol with a short statement of what the study will and will not establish. Specify the settings to which findings may reasonably transfer, the sources of uncertainty and the questions left for later work. This statement should guide the final discussion as well as the initial planning. Design quality comes from alignment and transparent judgement, not from selecting the most complicated method.