Research Design: Definition, Guide and Methods

Ask ten graduate students to describe their research design and eight will name a tool. “An online survey,” “a set of interviews,” and the conversation stops there. The gap shows up later, usually during analysis, when somebody on the committee asks what the study was built to demonstrate.

A research design is the overall strategy behind an empirical study. It settles the purpose of the research, the source of your data, whether that data is qualitative or quantitative, the setting, the number of times you measure and the number of cases you examine. Methods follow from the strategy. Work through this guide and you can lay out your own design in six decisions, each with a reason attached.

📌 Key points at a glance

• The research design is the strategy, methods sit inside it.
• USC Libraries lists 18 types of research design.
• Six decisions cover what a thesis actually has to settle.
• Cross-sectional data shows association, not causation.
• ICPSR charges non-member institutions about $825 per dataset.

What is a research design? Definition and purpose

A research design is the overall strategy that ties the components of an empirical study together. It describes how the investigation is set up, and it is settled before the first item is written and before the first participant is recruited.

The USC Libraries research guide “Organizing Your Social Sciences Research Paper”, last updated on 8 September 2026, states the priority plainly: the research problem determines the type of design you choose, not the other way around. The same page then lists 18 separate design types in alphabetical order, from action research through to systematic review.

Eighteen types is a lot to hold in your head, and the number itself is the useful part. No tidy three-way split is waiting to be memorized. What you get is a menu, and your job is to say which items you picked and why you picked them.

What your research design settles

• The kind of claim you intend to make.
• Where the data you analyze comes from.
• Whether that data is qualitative, quantitative or both.
• The setting, the timing and the number of cases.

What it deliberately leaves open

• The exact wording of your items.
• The software you collect and analyze with.
• The specific statistical test.
• Formatting and citation style.

Separating the two lists pays off at the defense, because a student who can name the design answers questions about the reach of the findings in one sentence.

Your design decides what the study can show. Your method only decides how cleanly it shows it.

Our reading tip: the USC Libraries guide “Types of Research Designs” by Robert V. Labaree is free to read and gives each of the 18 designs the same three headings: what it is, what these studies tell you, and what they do not tell you.

Research design vs. research method: what is the difference?

A research design describes the strategy of the whole study, and a research method describes the procedure you use to collect data. The two are not parallel choices. Design sits above method, because the strategy decides which procedures remain open to you.

Collapsing the two levels is the most common reason a methods chapter comes back marked up. “I am running an online survey” names a procedure and leaves five of the six design decisions blank. The next questions from a committee are about purpose, timing and case selection, and a procedure by itself answers none of them.

Each row in the table constrains the row below it. A cross-sectional design rules out several analyses, a qualitative design rules out the standardized instrument. Working from the top down is faster than starting with the instrument and reverse engineering a rationale. For the level below the design, see our overview of empirical research methods.

How to create a research design: the six decisions

A research design comes out of six decisions taken in this order: purpose, data source, data type, setting, number of measurement points and number of cases. Each one narrows the next, so skipping a decision costs time later rather than saving it now.

The table below is our own selection, cut down to what a thesis or dissertation has to settle rather than what a funded program can afford. Treat it as a questionnaire about your own project. Anyone who can answer all six rows has a finished research design, even with no prose written yet. For the wider process around it, see our guide to empirical research.

Decision 1: what is the purpose of your study?

The purpose sets how far your conclusions are allowed to reach. Exploratory studies open up a field that has barely been researched and generate hypotheses, descriptive studies record frequencies and distributions, explanatory studies test a suspected relationship. Picking one is a commitment rather than a label: exploratory work proves nothing at the end, and explanatory work needs testable hypotheses before collection begins. Our guide to exploratory, descriptive and explanatory research separates the terms.

Decision 2: collect new data or reuse existing data?

The data source saves the most time and gets chosen deliberately least often. Secondary analysis means working with data somebody else collected, and for US social science the main repository is ICPSR at the University of Michigan. Some collections are free to download, while members-only collections are restricted to member institutions: ICPSR charges individuals at non-member institutions an administration fee of approximately $825 per dataset (September 2026). Check your library's membership before you plan a study around a dataset you cannot reach.

Decision 3: qualitative, quantitative or mixed methods?

The data type governs everything downstream, from sample size to analysis. Quantitative work runs on numbers, standardized instruments and larger samples, qualitative work on text, open questions and few cases studied closely. Inside quantitative research the USC guide draws the sharper line, between descriptive designs, which answer who, what, when, where and how, and experimental designs, which require control, randomization and manipulation. Only the second kind supports a causal claim. Our comparison of qualitative and quantitative research covers the rest.

Decision 4: laboratory or field?

The setting governs how much control you have over interference. A laboratory gives you a controlled situation and lets you vary conditions deliberately, at the cost of everyday realism. Field research reaches people where they already are, with every influence that comes along. One point is specific to the United States: human subjects research goes to an Institutional Review Board rather than to a national ethics committee, and the IRB application asks you to describe the design, not just the consent process. Build the review time into your schedule.

Decision 5: one measurement point or several?

The number of measurement points separates cross-sectional from longitudinal designs and decides whether you may talk about change. A cross-sectional study collects data at one point in time, a longitudinal study follows the same sample and repeats the observation. The National Longitudinal Survey of Youth 1997, run by the Bureau of Labor Statistics, shows the scale involved: 8,984 people born between 1980 and 1984 were first interviewed in 1997, then annually through 2011 and every two years since. For a thesis the cross-section is the default, because a second wave rarely fits the calendar.

Decision 6: how many cases do you study?

The number of cases runs from a single one to the whole population. A case study examines one case, perhaps one school district or one company, in real depth. A sample is a selection from the population and carries conclusions back to the whole. A census covers everyone, which outside a national count or a small, clearly bounded group is rarely realistic. How to choose procedure and size is set out in our guide to sampling.

Research design examples from two student projects

A research design becomes concrete as soon as the six decisions are filled in for a real question. The two examples below come from ordinary thesis projects and differ in every single row, although both sit under the same broad topic.

Put them side by side. On the left a student who wants to know how common something is, on the right a student who wants to know how people experience it. The same curiosity produces two designs with almost nothing in common.

Both designs are complete and neither is the better one. Twelve interviews are not worth less than 300 questionnaires, they answer a different question. Fill in the left-hand column for your own project and the outline of your methods chapter is already settled.

💡 Tip

Put your six rows on half a page and send them to your adviser before the proposal deadline. A design can be fixed in a ten minute conversation. Data collection that has already started cannot.

Writing up: what belongs in your methods chapter

The research design appears in your write-up as its own chapter between the literature review and the results, and it explains in a few pages what you did and why. Committees grade the justification far more often than the description.

American terminology runs the opposite way to British usage. In the United States a thesis is the write-up for a master's degree and a dissertation is the write-up for a doctorate, while UK universities use the two words the other way round. Check which convention your program follows before you title anything.

The order inside the chapter is much the same across disciplines because it follows the route the study took. Your program handbook overrules it, since some departments want the hypotheses in the literature review instead.

1. Research question and hypotheses. What the study follows from.
2. Purpose and design type. The six decisions in prose.
3. Choice of method. With a reason, not just a name.
4. Sample and recruitment. Selection procedure, size, response rate.
5. Instrument and pilot. Structure, scales, changes after piloting.
6. Analysis technique. Which technique, applied to which question.
7. Limitations of the design. What this setup cannot show.

The gap between a strong chapter and a flat one sits almost always in point 3 and point 7. Four questions turn a report into an argument: why this method and not another, why these cases, why this analysis, and what each decision costs you.

⚠️ Warning

Naming your own limitations earns credit rather than losing it. Writing that a cross-section cannot support a causal claim shows methodological understanding. Leaving it out means somebody on the committee raises it instead.

Where the chapter sits in the shape of the whole document is covered in our guide to the empirical paper. For the instrument itself, use our guide to questionnaire design.

Common mistakes in research design

Most trouble in a methods chapter comes not from missing knowledge but from an early decision nobody revisited. Four mistakes turn up again and again.

All four are cheap to fix in the first few weeks and impossible to fix after data collection. Timing is the reason a research design belongs before the proposal is approved rather than after.

The method is fixed before the question is

Starting with “I am running an online survey” means hunting for a question that fits it. Committees notice, because the methods chapter then contains no justification: no alternative was ever weighed. Turn the order round and narrow the research question until a purpose falls out of it. The method follows almost by itself after that.

The design does not fit the calendar

A longitudinal study across two semesters, a census of a whole county, thirty interviews with transcription: interesting on paper, undeliverable in one term. Budget the collection phase realistically, because responses do not arrive on day one and transcription takes roughly five times the recording length. A small design carried out properly beats a large one handed in unfinished.

A cross-section turns into a causal claim

The most common factual error in empirical student work shows up in the results chapter, as a sentence of the form “X leads to Y”. A cross-sectional study only shows that two characteristics occur together, not which one produces the other. Report such findings as association and say in the discussion which design a causal claim would have required.

The design is stated but never justified

Many papers name the design correctly and say nothing about why it was chosen. Readers then cannot tell whether the decision was deliberate or accidental. Write one sentence with because for each of the six decisions, even where it sounds obvious. Four sentences like that change how the whole chapter reads.

Conclusion

A research design is more predictable than the term suggests: six decisions, each with two or three usual answers, and the rest of the project follows almost by itself. Take them deliberately and in order before you build the first instrument. Write a because next to each one, since that is what your methods chapter is judged on.

Where to go next

• Still missing the overview of the whole research process? Fundamentals of empirical research
• Choosing between qualitative and quantitative? Qualitative vs. quantitative compared
• Hypotheses not written yet? How to write a hypothesis
• Planning your own data collection? Running a survey for your thesis

Design settled, data collection still ahead?

With empirio.ai, an online survey tool from Germany with hosting in the EU, you build the instrument that matches your design and export the responses straight into your analysis.

How to build the questionnaire for your research design