Skip to content
Free planNo credit card requiredStart writing

Types of research design: a guide to choosing the right one

Updated 13 min read

Short answer:

The main types of research design are experimental, quasi-experimental, correlational and descriptive (survey) designs on the quantitative side; case study, ethnography, grounded theory and phenomenology on the qualitative side; and action research and mixed methods, which combine the two. Each design answers a different kind of question, so you choose it from your research question, not from habit. Any design can collect data once (cross-sectional) or repeatedly (longitudinal).

Each design below has a definition, when to use it, an example and its main limit.

What is a research design?

A research design is the overall plan that links your research question to the evidence that will answer it: who or what you study, what you compare, when you collect data, and how you rule out other explanations. It sits between the question and the methods. If your question is "Does X cause Y?", the design decides whether you can actually claim that at the end.

Shadish, Cook and Campbell set out three conditions for a causal claim: the cause comes before the effect, the two vary together, and no other plausible explanation remains [1]. Most design choices are about how many of those conditions you can meet. An experiment can meet all three. A one-off survey usually meets only the second.

How are research designs grouped?

Designs are usually grouped by the kind of data they produce (numbers, words or both) and, for quantitative designs, by whether the researcher controls who gets the exposure.

FamilyDesignsResearcher controls the exposure?Typical question
Quantitative, experimentalTrue experiment (randomized)Yes, and assigns at randomDoes X cause Y?
Quantitative, quasi-experimentalNonequivalent groups, interrupted time series, regression discontinuityOften, but not at randomDid X probably cause Y in this real setting?
Quantitative, non-experimentalCorrelational, descriptive, survey, cohort, case-controlNoHow much? How are X and Y related?
QualitativeCase study, ethnography, grounded theory, phenomenology, narrativeNoHow and why? What is it like?
CombinedMixed methods, action researchVariesQuestions that need both numbers and meaning

For the difference between the three approaches in data, sampling and analysis, see quantitative, qualitative and mixed methods. For how study types rank as evidence in health research, see levels of evidence.

Experimental design

An experimental design is one in which the researcher changes the independent variable and assigns participants to conditions at random, so any difference in the outcome can be attributed to the change. Random assignment is what makes it a true experiment: it balances known and unknown differences between groups on average [1].

  • Use it when you want to show cause and you can ethically and practically assign the treatment.
  • Example (illustrative), education: 240 first-year students are randomly assigned to weekly low-stakes quizzes or to no quizzes, and final exam scores are compared.
  • Main limit: lab conditions may not match real life, and many exposures (smoking, poverty) cannot be assigned.

The building blocks are an independent and a dependent variable and a comparison group. Our guide to the controlled experiment covers control groups, randomization, blinding and placebos with worked examples.

Quasi-experimental design

A quasi-experiment is an experiment in which units are not assigned to conditions at random [1]. The researcher compares groups or time periods that already exist, and must argue away the differences that randomization would have removed.

Common versions:

  • Nonequivalent groups with pretest and posttest: two schools, one adopts a new reading program; both are tested before and after.
  • Interrupted time series: many measurements before and after a change, such as monthly hospital infection rates before and after a hand-hygiene policy.
  • Regression discontinuity: people just above and just below a cut-off (a test score that decides who gets a scholarship) are compared.

Example (illustrative), business: a retailer introduces a four-day week in 12 stores and compares staff turnover with 12 similar stores that kept a five-day week. Main limit: the stores that chose the change may differ in ways you did not measure.

Correlational design

A correlational design measures two or more variables as they occur, without changing anything, to see whether and how strongly they are related. It answers "Do X and Y move together?" and supports prediction, but on its own it does not show cause.

  • Example (illustrative), psychology: 400 employees report weekly hours of remote work and complete a job satisfaction scale; the study estimates the correlation and tests whether it holds after adjusting for age and role.
  • Main limit: a third variable may drive both, and the direction may run either way. Satisfied people may choose to work from home, not the reverse.

Descriptive and survey design

A descriptive design sets out what exists, how much and how often, without testing why; a survey is the most common way to do it. A survey asks a sample a fixed set of questions so the answers can be summarized and, with a probability sample, generalized to the population [2].

  • Example (illustrative), public health: a random sample of 1,200 households estimates the share with a working smoke alarm, with a 95% confidence interval.
  • Main limit: answers depend on who responds and on how questions are worded. Your sampling method decides how far you can generalize; see probability vs non-probability sampling. For closed yes/no items, see our guide to dichotomous questions.

Cross-sectional vs longitudinal designs

A cross-sectional design collects data at one point in time; a longitudinal design collects data from the same units, or the same population, at two or more points. This is a time dimension, not a separate family, so a survey, a correlational study or a qualitative study can be either.

Cross-sectionalLongitudinal
Data collectedOnceRepeatedly over weeks, months or years
ShowsPrevalence; associations at one momentChange, sequence, who develops an outcome
Can it show which came first?RarelyYes, for measured variables
Cost and timeLowHigh; participants drop out
Example (illustrative)Survey of 800 nurses' burnout this monthThe same 800 nurses surveyed every year for five years

Longitudinal designs include panel studies (same people each wave), cohort studies (a group defined by a shared exposure or birth period, followed forward) and trend studies (new samples from the same population each time). Cohort studies are central to epidemiology; see what a cohort study is.

Case study

A case study is an in-depth study of one or a few bounded cases (a person, a school, a company, a policy) in their real-world context, using several sources of evidence. Yin describes it as the design to use for "how" and "why" questions about contemporary events the researcher cannot control [3].

  • Example (illustrative), engineering management: how one city water utility introduced sensor-based leak detection, using interviews with 15 staff, project documents and five years of leakage data.
  • Main limit: findings generalize to theory, not to a population. Say why this case was chosen (typical, extreme, critical, revelatory).

Ethnography

Ethnography studies a group's culture and everyday practices by spending extended time with them, through participant observation, informal interviews and documents [4]. The researcher's field notes are the main data.

  • Example (illustrative), sociology: nine months observing shift handovers on a hospital ward to understand how informal rules shape what information is passed on.
  • Main limit: it is slow, depends on access, and the researcher's presence can change what happens, so reflexivity is expected.

Grounded theory

Grounded theory builds a theory from the data itself, collecting and analyzing at the same time, comparing each new piece of data with what came before and choosing the next participants to fill gaps in the emerging theory [5][6]. These are constant comparison and theoretical sampling.

  • Example (illustrative), education: interviews with 25 mature students who left university in their first year, analyzed until a model of "gradual withdrawal" explains the cases.
  • Main limit: it is not a label for any interview study. If you do not sample theoretically or produce a theory, call your analysis thematic instead. Our guide to coding qualitative data explains the difference.

Phenomenology

Phenomenology describes the lived experience of a phenomenon from the point of view of the people who have it, usually through long, open interviews with a small, carefully chosen sample. Interpretative phenomenological analysis (IPA) is a widely used version in psychology and health [7].

  • Example (illustrative), nursing: how eight new mothers experience returning to work after a stillbirth.
  • Main limit: small samples and deep interpretation. It aims at understanding, not at counting how common an experience is.

Action research

Action research is a cycle of planning a change, acting, observing what happens and reflecting, carried out with the people affected, so that research and improvement happen together. The idea is usually traced to Kurt Lewin [8].

  • Example (illustrative), teaching: a team of four primary teachers runs three cycles of changes to feedback on writing, collecting pupil work samples and their own reflection logs each cycle.
  • Main limit: findings are tied to the setting, and the researcher is also a participant, so be open about that role.

Mixed methods designs

Mixed methods research collects both quantitative and qualitative data and integrates them deliberately to answer a question neither could answer alone. Creswell and Plano Clark describe three core designs [9]:

DesignOrderUse it to
ConvergentBoth strands at the same time, then mergedCompare and combine two views of the same question
Explanatory sequentialQuantitative first, then qualitativeExplain surprising or significant results
Exploratory sequentialQualitative first, then quantitativeBuild and test an instrument or intervention from what people say

See quantitative, qualitative and mixed methods for each design with an example and how to justify the choice.

How do you choose a research design?

Start from the verb in your research question: "how many" points to a descriptive design, "is related to" to a correlational one, "does it cause" to an experiment, and "how" or "what is it like" to a qualitative design. Then check what is ethical and feasible.

If your question asksChooseIf you cannot, fall back to
How many, how often, what proportionDescriptive survey (cross-sectional)Secondary analysis of existing records
Are X and Y related?CorrelationalSame, with more covariates
Does X cause Y, and can you assign X?Randomized experimentQuasi-experiment
Did a policy or program change an outcome?Quasi-experiment (time series, nonequivalent groups)Before-and-after with a comparison group
Who develops an outcome, and what predicts it?Longitudinal cohortCase-control (looks back from the outcome)
How does change happen over time?Longitudinal (panel)Retrospective interviews
How and why did this happen in this setting?Case studyDocument analysis
How does a group live and work?EthnographyObservation with interviews
What explains a process no theory yet covers?Grounded theoryThematic analysis
What is it like to experience X?PhenomenologyThematic analysis of interviews
How can we improve our own practice?Action researchEvaluation case study
Why did the numbers come out this way?Explanatory sequential mixed methodsSurvey with open questions

Four checks before you commit:

  1. Ethics. You cannot assign a harmful exposure. Your ethics application will ask why the design is needed.
  2. Time. A longitudinal study needs the time between waves. A 12-month master's rarely fits more than two waves.
  3. Sample. Experiments and surveys need numbers set by a power calculation; see statistical power and sample size.
  4. Measurement. Every variable needs a reliable, valid measure; see reliability vs validity.

Common mistakes when choosing a design

  • Claiming cause from a correlational or cross-sectional design. Write "associated with", not "leads to", unless the design supports it.
  • Calling any interview study "phenomenology" or "grounded theory". Each has its own sampling and analysis. If you did not follow them, call it a qualitative descriptive study with thematic analysis.
  • Calling a convenience sample "random". Name the sampling method you actually used.
  • Mixing methods without integrating them. Two separate chapters are not a mixed methods design. Say where and how the strands meet.
  • Choosing the design before the research question. The design follows from the question.

Once you have chosen, the methodology chapter guide shows how to write the design and its justification.

How to choose a research design with AI for Research

  1. See which designs others used (0 credits). In Find Papers, search your topic and filter by study type: RCT, cohort, case-control, cross-sectional, qualitative and more. Searching and filtering are free.
  2. Frame the question. The PICO question builder turns PICO, PICOT, PCC or SPIDER fields into a question and a search concept table. It runs in your browser, free and without an account.
  3. Draft the design (3 credits). The Methodology builder in the Thesis Suite takes your question, lets you choose quantitative, qualitative or mixed methods or suggest one, and drafts the design, sampling, analysis and limits with citations to studies from its search.
  4. Write the proposal. The research proposal generator drafts a proposal with a methods section and runs 3 times a day without an account.

The Free plan has 45 credits a month; see pricing. Limits: study-type labels can be wrong, and no tool knows your access, ethics rules or deadline. Your supervisor decides whether the design fits.

FAQ

What are the main types of research design?

Quantitative designs are experimental, quasi-experimental, correlational and descriptive (including surveys), each run cross-sectionally or longitudinally. Qualitative designs include case study, ethnography, grounded theory, phenomenology and narrative research. Action research and mixed methods designs combine elements of both. The right one depends on whether your question asks what exists, what is related, what causes what, or how people experience something.

What is the difference between research design and research method?

A research design is the overall plan for answering your question: who you study, when, with what comparison, and how you will rule out other explanations. A method is a specific technique inside that plan, such as a questionnaire, an interview or a regression model. A survey design can use several methods, and an interview can appear in a case study, an ethnography or a mixed methods study.

Which research design is best?

No design is best in general. A randomized experiment gives the strongest evidence that one thing causes another, but it cannot tell you how patients experience a treatment or how common a problem is. Match the design to the question: experiments for cause, surveys for how much, cohort studies for change over time, and qualitative designs for meaning and process.

Is a case study qualitative or quantitative?

A case study is usually qualitative, but it does not have to be. Yin defines a case study by its focus, an in-depth look at a bounded case in its real-world context, not by its data. Many case studies combine interviews and documents with numbers such as sales figures, test scores or admission records. Say which kinds of evidence you will use and how you will combine them.

How do I justify my research design?

Tie the design to the question in one sentence, then explain what it lets you claim and what it does not. Name the alternatives you rejected and why: an experiment was unethical, a longitudinal study did not fit a 12-month timeline. Cite a methods text for the design and one or two published studies in your field that used it. Examiners look for reasons, not labels.

Next steps

Work through the rest of the research methods series:

Sources

  1. Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and Quasi-Experimental Designs for Generalized Causal Inference. Houghton Mifflin.
  2. Dillman, D. A., Smyth, J. D., & Christian, L. M. (2014). Internet, Phone, Mail, and Mixed-Mode Surveys: The Tailored Design Method (4th ed.). Wiley.
  3. Yin, R. K. (2018). Case Study Research and Applications: Design and Methods (6th ed.). SAGE.
  4. Hammersley, M., & Atkinson, P. (2019). Ethnography: Principles in Practice (4th ed.). Routledge.
  5. Glaser, B. G., & Strauss, A. L. (1967). The Discovery of Grounded Theory: Strategies for Qualitative Research. Aldine.
  6. Charmaz, K. (2014). Constructing Grounded Theory (2nd ed.). SAGE.
  7. Smith, J. A., Flowers, P., & Larkin, M. (2009). Interpretative Phenomenological Analysis: Theory, Method and Research. SAGE.
  8. Lewin, K. (1946). Action research and minority problems. Journal of Social Issues, 2(4), 34–46. https://doi.org/10.1111/j.1540-4560.1946.tb02295.x
  9. Creswell, J. W., & Plano Clark, V. L. (2018). Designing and Conducting Mixed Methods Research (3rd ed.). SAGE.