Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Contents

Academy · Research Data Analytics & AI · What is research design?

Research Design: Types, Methods and Best Practices

Research design is the plan that determines how data will be collected, analysed and interpreted, and therefore what the study can conclude. A poorly chosen design cannot be corrected by later statistical analysis.

Pubrica AcademyWritten by the medical writing and biostatistics teams Reviewed byPubrica biostatistics team Published2 October 2026 Last checked2 October 2026 Reading time8 minutes

Definition of Research Design

Research design is the structure of a study: the plan that links the research question to the data you will collect, the analysis you will run and the conclusions you can draw. It is not the same as research methodology, although the two terms are often used interchangeably 1. Methodology is the set of procedures, such as how you sample, how you measure and which test you run. Design is the logic within which those procedures operate, and it is decided first.

A simple test distinguishes the two. If what your study can prove depends on how the groups were formed and when the outcome was measured, the question concerns your design. If it depends on which assay or questionnaire you used, it concerns your methods.

How the Design Determines the Reporting Guideline

Many introductions to research design omit this point, and omitting it can cost authors a round of review. Your design determines which reporting standard governs the write-up: CONSORT 2025 for a randomised trial, STROBE for an observational study, PRISMA 2020 for a systematic review, STARD for a diagnostic accuracy study, TRIPOD+AI for a prediction model and ARRIVE 2.0 for animal research. Reading the relevant checklist when you choose the design is one of the most effective steps you can take for the eventual paper.

Why Research Design Matters

For researchers, the design is not a procedural step to complete before the main work begins. It determines whether the results of the study are meaningful.

  • It limits the claims you can make. A cross-sectional design cannot answer a question about change over time, and no statistical adjustment can change this. This is the most common reason a manuscript’s discussion section has to be rewritten.
  • It governs validity and reliability. Randomisation, blinding, control groups and pre-specification exist to minimise bias and error. They cannot be added once the data have been collected.
  • It determines what the study costs. A design matched to the question collects the variables the analysis needs and no more. A vague design collects excessive data and may still miss essential variables.
  • It defines generalisability. Who you recruited, and how, determines the population to which the findings apply. Reviewers will ask about this, so it should be decided in advance.

Four Types of Research Design

Research designs are usually grouped by what they aim to establish. Four families cover most health and life-science research, and each answers a different type of question 2.

Descriptive research design characterises a population or a phenomenon as it is. It answers what, not why, and it does not manipulate anything.

  • Use it to establish prevalence, document a pattern, or map a situation nobody has measured yet.
  • Methods — cross-sectional surveys, observational cohorts, case series, registry and archival analysis.
  • Example — a national survey measuring how many patients with a given condition are receiving guideline-concordant therapy 4.
  • Reported under STROBE.

Correlational research design looks for an association between two or more variables without manipulating any of them. It can establish that two variables change together, and the strength and direction of that relationship, but it cannot establish that one causes the other.

  • Use it to test whether a relationship exists before committing to an expensive experiment.
  • Methods — correlation and regression on survey, registry or observational data.
  • Example — examining the relationship between weekly physical activity and self-reported stress in adults 4.
  • Watch for inconsistent language: “associated with” in the results and “leads to” in the discussion is the most common overclaim in the literature.

Experimental research design manipulates one or more independent variables and measures the effect on a dependent variable. Random allocation and control of extraneous variables are what allow a causal claim.

  • Use it to establish cause and effect.
  • Methods — randomised controlled trials, laboratory experiments, field experiments; quasi-experimental designs where randomisation is not possible, with the weaker inference stated.
  • Example — a trial in which participants are randomly allocated to a new drug or a placebo and the primary outcome is measured at a pre-specified time point 5.
  • Reported under CONSORT 2025, with the registered protocol and the statistical analysis plan behind it.

Evidence synthesis design answers a question from existing studies rather than from new data. It is a design in its own right, with its own protocol, and it is distinct from a narrative literature review.

  • Use it to establish what the body of evidence shows, how certain it is, and where the gaps are.
  • Methods — systematic review, meta-analysis, scoping review, with the protocol registered before screening begins.
  • Example — pooling randomised trials of one intervention to estimate an overall effect and assess heterogeneity.
  • Reported under PRISMA 2020, with certainty of evidence assessed using GRADE.

Exploratory and Confirmatory Research

A second distinction applies across all four families and determines how results may be interpreted: whether the study was exploratory or confirmatory 3.

Exploratory research generates hypotheses. It addresses questions that have been little studied, is often qualitative or open-ended, and its findings indicate directions for further research rather than conclusions. Confirmatory research tests a hypothesis specified in advance. It is usually quantitative and structured, and its claims depend on how thoroughly it was pre-specified.

A Common Error and How to Avoid It

The common error is running an exploratory analysis and reporting it as confirmatory. Pre-registering the hypothesis, the primary outcome and the analysis plan before data collection keeps the two separate, which is why editors increasingly ask for the registration record alongside the manuscript. An exploratory finding reported as exploratory is publishable; the same finding presented as confirmatory is more likely to fail to replicate.

Six Steps to Developing a Robust Research Design

Developing a research design is a sequence, and each step constrains the next 2. Working out of order can lead to collecting data that cannot answer the question.

  1. State the research question precisely. A specific, answerable question that addresses a genuine gap. If it cannot be written in one sentence, the design cannot yet be chosen.
  2. Choose the design family. Descriptive, correlational, experimental or synthesis, chosen according to what the question asks rather than the data that happen to be available.
  3. Define the variables. Independent, dependent and control variables, each with its measurement and time point. Misclassifying a variable at this stage affects every subsequent step.
  4. Plan the data collection. Use methods that match the design and the variables, qualitative or quantitative, with the sample size justified rather than assumed.
  5. Write the analysis plan. Specify the tests, the handling of missing data and any adjustment for multiple comparisons, and record them before data collection. For trials, the ICH E9(R1) estimand framework is the clearest way to state the population, the treatment condition, the outcome and how intercurrent events are handled 6.
  6. Pilot it. A small pilot identifies instruments that do not work and recruitment assumptions that are too optimistic, while both are still inexpensive to correct.

Common Research Design Mistakes and How to Avoid Them

Nine problems account for most of the design issues identified at peer review. All of them are easier to correct before the study than after it.

Common Design Problems

  • An unclear purpose — a question too broad to answer. Write it as one measurable sentence before anything else.
  • A design mismatched to the question — most often an observational design asked to support a causal claim. Match the design to the inference you need.
  • Sampling problems — a sample too small, or drawn so that it cannot represent the population. Justify the size, state the frame, and report who was excluded and why.
  • Ethics and governance left late — informed consent, confidentiality and ethics committee approval are design decisions, not paperwork. Approval obtained after recruitment is a problem no journal will overlook.
  • Unreliable data collection — an instrument that has not been validated, or has been validated in a different population. Pre-test it.
  • Analysis chosen after the fact — running tests until one is significant. Fix the analysis plan in advance and report what it produced.
  • Bias and limitations unacknowledged — a limitations section that names no specific limitations suggests that limitations have not been considered.
  • No pilot — a design that has not been tested on a small scale usually encounters its first problem at full scale.
  • A thin literature review — this can lead to designing a study that has already been done, or to missing the measure that others in the field use.

The Most Important Check

Check the match between the design and the claim. Sampling and instruments can often be adjusted during a study, but a design that cannot support the conclusion cannot be corrected.

Take-home message for authors and researchers

  • Choose the design according to the claim you need to make, not the data you already hold. Analysis cannot compensate for a design that does not support the conclusion.
  • Read the reporting checklist for your design when you choose it, not when you prepare to submit.
  • Fix the hypothesis, the primary outcome and the analysis plan before data collection, and state clearly which analyses were exploratory.
  • Justify the sample size rather than assuming it; a power calculation performed after data collection does not justify the sample.
  • Pilot the instrument and the recruitment assumptions while both are still inexpensive to change.

The emphasis differs by field. Clinical trials are governed closely by CONSORT and the registered protocol; observational epidemiology depends on STROBE and on how the cohort was assembled; laboratory and life-science research is judged on replication and controls; engineering and materials studies usually require detailed documentation of the measurement protocol; and evidence synthesis in any discipline depends on a protocol registered before screening.

Research Design FAQs

Design is the logic of the study — how groups are formed, what is manipulated, when outcomes are measured, and therefore what the study can conclude. Methodology is the set of procedures inside that logic: the sampling technique, the instrument, the statistical test. Design is decided first, and it constrains the methods rather than the other way round.

Start from the claim you need to make. To establish cause, you need an experimental design with random allocation. To show that a relationship exists, a correlational design is sufficient. To describe a population, use a descriptive design. If the evidence already exists, a systematic review can answer the question without new data, usually more quickly.

Not on its own. Methods such as instrumental variables, difference-in-differences and target trial emulation can strengthen causal inference from observational data, but each relies on assumptions that must be stated and justified. The appropriate default is to report the association and state clearly what the design cannot establish.

It depends on the design, the effect you consider meaningful, the variability of the outcome and the power you require. It is a calculation rather than an estimate, and it belongs in the protocol. A power calculation performed after data collection does not justify the sample.

Yes. The design determines which reporting guideline governs the manuscript, which items the journal will check and what the discussion can claim. Reading the relevant checklist from the EQUATOR Network while the design is still being planned is one of the most effective quality steps available to an author.

Sometimes, provided the change is declared. Record what changed, when, why, and whether the decision was made before or after anyone saw outcome data. Reviewers treat an undeclared change very differently from one explained in the methods.

References and Sources

  1. Abutabenjeh S, Jaradat R. Clarification of research design, research methods, and research methodology. Teaching Public Administration 2018;36(3):237–58. read 1 October 2026. doi.org/10.1177/0144739418775787
  2. Hanson BP. Designing, conducting and reporting clinical research: a step by step approach. Injury 2006;37(7):583–94. read 1 October 2026. pubmed.ncbi.nlm.nih.gov/?term=Hanson+designing+conducting+reporting+clinical+research+Injury+2006
  3. Nilsen EB, Bowler DE, Linnell JDC. Exploratory and confirmatory research in the open science era. Journal of Applied Ecology 2020;57(4):842–7. read 1 October 2026. doi.org/10.1111/1365-2664.13571
  4. Beauvais AM, Stewart JG, DeNisco S, Beauvais JE. Factors related to academic success among nursing students: a descriptive correlational research study. Nurse Education Today 2014;34(6):918–23. read 1 October 2026. doi.org/10.1016/j.nedt.2013.12.005
  5. EQUATOR Network reporting guideline index — CONSORT 2025, STROBE, PRISMA 2020, STARD, TRIPOD+AI and ARRIVE 2.0. read 1 October 2026. www.equator-network.org
  6. ICH E9(R1) addendum on estimands and sensitivity analysis in clinical trials. read 1 October 2026. www.ich.org/page/efficacy-guidelines

Each source shows the date it was last read. When a standard changes, this article is revised.

How to Cite This Article

Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.

Pubrica Academy. (2026, October 2). What is research design? Types, methods and best practices. Pubrica. https://pubrica.com/services/physician-writing-services/research-proposal/research-design-types-methods-best-practices/

This article is intended as a learning resource. Cite it when referring to its content; when using the evidence behind it, cite the sources listed above.

Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts who carry out this work. Articles are revised when the underlying standards change, and each shows the date it was last checked.

How Pubrica Can Help

Support is most valuable at the design stage, because the design cannot be changed once the study is complete. Research proposal writing covers the research question, the design and the protocol together; statistical analysis plan fixes the analysis before data collection; and our biostatistics team calculates the sample size the study requires. Where the question has already been addressed in the literature, a systematic review or meta-analysis answers it without new data. Before submission, reporting checklist review checks the manuscript against the guideline your design requires.

Search the Academy

Search Pubrica Academy

Fifty-eight free articles, templates, guidelines, checklists and worked examples, available without an account. Search for your topic, or open the full library to filter by category, sub-topic or resource type.

Request a Free Research Design Scoping Assessment

The scoping assessment is free and takes two business days. You receive a written assessment of what the work requires and which specialists would handle it.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.