What Should You Decide First After Settling on a Research Question?

You have a research question. What should you decide next?

Settling on a research question can feel like a major turning point. The problem is what happens immediately afterward. Should you choose quantitative or qualitative research? Decide on a sample? Write a questionnaire? Pick a statistical test? Start preparing an ethics application?

Those decisions matter, but making them in the wrong order can create a study whose parts do not fit together. A questionnaire is useful only if questionnaire data can answer the question. A statistical technique is useful only if the study produces appropriate data for it. Even an elegant research design is of little value if you cannot obtain the participants, records, equipment, permissions, time, or expertise needed to carry it out.

The first decisions after a research question should therefore establish the logic of the study: what evidence would constitute an answer, where that evidence could come from, and what kind of study could generate or obtain it.

Start with the evidence you need, not the method you want to use

Move from question to evidence before moving from question to methods

A research question tells you what you want to know. It does not automatically tell you how to find out.

The bridge between those two points is a series of linked decisions. You need to determine what an acceptable answer would require you to observe, compare, measure, interpret, estimate, describe, or explain. From there, you can identify the appropriate sources of evidence and choose a design capable of connecting that evidence to the question.

This sequence matters because the research question should drive the design rather than being retrofitted to whatever method happens to be familiar. Methodological literature consistently treats the research question as foundational to study design, while feasibility includes practical considerations such as access to participants, expertise, resources, and time.

First ask: What would I need to know to answer this question?

Before naming a methodology, imagine that the study has been completed successfully. What evidence would allow you to answer the research question?

Suppose your question is:

How do first-year university students experience the transition from fully online learning to face-to-face classes?

An answer requires evidence about students' experiences, interpretations, challenges, and perhaps how those experiences vary. Simply obtaining their examination scores would not directly answer that question.

Now consider a different question:

Is participation in a supplemental mathematics tutorial associated with first-year students' final mathematics grades?

Here, detailed accounts of students' experiences might be interesting, but they would not by themselves provide the evidence needed to estimate the stated association. You would need information about tutorial participation and mathematics outcomes, together with whatever additional variables and design considerations are necessary for a defensible analysis.

The wording of the question therefore begins to constrain what counts as relevant evidence. WHO guidance similarly emphasizes that a clear research question helps determine study design, sampling, and analysis, while outcomes should be chosen so that the study can actually answer the question.

Identify the unit or source of evidence

Once you know what evidence is required, ask where it could come from.

Depending on the research question, the relevant source might be people, classrooms, organizations, documents, published studies, administrative records, social media posts, biological specimens, images, transactions, experiments, sensors, archival materials, or an existing dataset.

This decision sounds obvious, but it can substantially reshape a study. Consider the question:

How consistently do universities disclose their policies on generative AI?

If the object of interest is the content of official institutional policies, students may not be the primary source of evidence at all. The relevant units could instead be university policy documents or official webpages.

Clarifying the unit of analysis and evidence source early helps prevent a common methodological mismatch: collecting data from an accessible source simply because it is accessible, rather than because it can answer the question.

Determine the broad design that can produce the evidence

Only after clarifying the required evidence should you ask what broad research design could produce it credibly.

This does not mean you must immediately specify every procedure. At this stage, you may only need to determine whether the question points toward an experiment, observational study, survey, qualitative inquiry, case study, secondary-data analysis, systematic review, mixed-methods design, or another appropriate approach.

There is no universally superior design independent of the question. A design is appropriate insofar as it can generate evidence suited to the claim you intend to make. Research-methods literature consequently emphasizes that design selection should follow from the research question rather than from a generic hierarchy of supposedly superior methods.

These are orientations, not automatic conversions. Two similarly worded questions may require different designs because of the claims being made, the available evidence, disciplinary conventions, or practical constraints.

Clarify the population, setting, cases, or material of interest

The next decision is not necessarily the exact sample size. It is usually the broader question of what or whom the study is about.

If you are studying people, distinguish the population of substantive interest from the participants you happen to be able to recruit. If you are studying institutions, define what counts as an eligible institution. If you are analyzing documents, determine what kind of documents fall within the scope. If you are conducting secondary-data research, identify the records or dataset capable of representing the relevant population or phenomenon.

For some quantitative questions, structured frameworks make these elements explicit. In clinical and health research, for example, PICO and related formulations can help specify population, intervention or exposure, comparison, and outcome. Such frameworks are useful where they fit the question, but they should not be forced onto research traditions for which they were not designed.

Identify the central concepts, variables, outcomes, or phenomena

A question may sound clear in ordinary language while remaining ambiguous methodologically.

Consider:

Does social media use affect academic performance among university students?

What counts as social media use? Total screen time? Time on particular platforms? Academic versus recreational use? Self-reported use or device-recorded behavior?

What counts as academic performance? Grade point average? Course grades? A standardized assessment? Student perceptions of their performance?

You do not necessarily need to finalize every operational definition immediately, but you should identify the concepts that will eventually require definition or measurement. Otherwise, you may discover later that different parts of the project have been using the same words to mean different things.

This is especially consequential when a study has a primary outcome. WHO guidance on research questions recommends careful advance consideration of outcomes so that the resulting evidence can answer the intended question.

Check what is already known before locking in the design

A settled research question should still be tested against the literature.

The literature may reveal that the question has already been answered convincingly, that an important variable has been overlooked, that a proposed measure performs poorly, or that previous researchers repeatedly encountered the same recruitment problem. It can also reveal established designs, instruments, datasets, theoretical perspectives, and methodological disagreements relevant to your choices.

This does not mean copying the most common design. Prior studies are evidence about how the problem has been approached, not instructions that must be followed.

The review may even require you to revise the question. Research planning is iterative. A question can be sufficiently settled to guide planning without being protected from evidence that exposes a serious problem.

Test the emerging design against feasibility

An appropriate design is not automatically a feasible design.

Feasibility includes whether you can obtain enough relevant participants or cases, access the required setting or dataset, use the necessary equipment or software, complete the work within the available time, meet the financial costs, and obtain the methodological or technical expertise the study requires. These considerations are widely recognized as part of assessing whether a research question can become a viable study.

This creates an important feedback loop:

Feasibility work can itself be empirical. In some settings, pilot or feasibility studies are used to test recruitment, retention, intervention delivery, measurement, or other procedures before a larger study is undertaken.

Consider ethics while the design is taking shape

Ethics should not be treated as a final checkpoint after all methodological choices have been made.

The population, recruitment method, data collected, intervention, consent process, privacy protections, incentives, risks, and data-sharing arrangements may all have ethical implications. A design that cannot be conducted ethically is not an acceptable design simply because it answers the question efficiently.

At this stage, you may not yet have all the documentation required for formal review. You should, however, be able to recognize whether your emerging choices raise ethical or regulatory issues that could materially affect the study.

Do not confuse the broad design with the complete protocol

Knowing that you will conduct a qualitative interview study does not yet tell you the sampling strategy, recruitment procedure, interview guide, recording arrangements, analytic approach, data-security procedures, or stopping logic.

Likewise, deciding on an observational quantitative study does not settle the exact variables, measurement procedures, sample-size considerations, missing-data strategy, statistical model, or quality-control procedures.

Those decisions matter, but they are downstream of the basic question-to-evidence logic. A research protocol eventually makes many of these choices explicit. WHO's recommended protocol structure, for example, asks researchers to specify objectives, study design, population, sampling, data collection, analysis, ethical considerations, and other operational details.

The useful distinction is between knowing what kind of study you are building and having every construction detail finalized. The latter develops progressively as you determine how detailed the research plan needs to be before you begin.

The decisions are sequential, but not perfectly linear

It would be convenient if research planning worked as a one-way sequence in which the question permanently determines the design, the design determines the sample, and everything thereafter simply follows. In practice, researchers move back and forth.

You may discover that the desired population cannot be accessed. A validated instrument may not exist in the necessary language. A dataset may lack a crucial variable. A preliminary sample-size assessment may show that the intended comparison is unrealistic. An ethics issue may make a proposed procedure inappropriate.

When that happens, revisiting an earlier decision is not necessarily a planning failure. It is often precisely what planning is for.

See how the first decisions emerge from the question

Notice what has not yet been decided. The researcher has not necessarily selected the final AI system, written every prompt, determined the exact sample size, completed the scoring rubric, chosen the statistical model, or finalized the participant schedule.

Those decisions will follow. What now exists is more fundamental: a defensible connection between the question, the evidence required to answer it, and a broad design capable of producing that evidence.

Avoid letting convenient methods make the important decisions for you

Make the next decision by asking what the question logically requires

If your research question is now reasonably stable, resist the urge to fill in an entire methodology section immediately. Work outward from the question.

This is the point at which the broader set of decisions that should be made before starting the research becomes easier to organize. Instead of making dozens of unrelated choices, you can evaluate each one according to the study you are actually trying to conduct.

You also do not need to freeze everything at once. Some choices can legitimately remain flexible during the early stages of a research project, while others become increasingly consequential as you approach recruitment and data collection.

Eventually, the emerging design needs to become operational. That means specifying the procedures, instruments, sampling decisions, data management, analysis plans, approvals, and other elements required by the study. A research protocol can provide a structured place to document those decisions when the project reaches the appropriate stage.

After settling on the question, work through these decisions

Common questions about what comes after the research question

Let the question determine what evidence you need next

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.