Choose your research design by asking what evidence would answer your question, and what conclusion that evidence could justify. Then check whether you can obtain it ethically and within your resources. Choosing an impressive label first can leave you with a method that does not answer the question.
The USC Libraries research-design guide treats design as the strategy connecting a research problem with the evidence and analysis needed to investigate it. The decision worksheet below is an original planning aid, not a replacement for supervision or a discipline's methods training.
Once the design is agreed, your project document needs to explain how the question, method and evidence fit together. UniDraft helps organise those confirmed decisions into an editable Word report using your supervisor's instructions. Start your project document with UniDraft, and bring the methods and results you can verify rather than asking software to choose the design for you.
Separate the decisions students often mix up
| Decision | Example of what it describes |
|---|---|
| Research question | What you want to describe, explain, compare or evaluate |
| Study design | How evidence will be organised to address that question |
| Instrument | A questionnaire, interview guide, observation record or test tool |
| Selection | Which people, cases, records or materials enter the study |
| Analysis | How the collected evidence will be examined |
| Software | A tool used to manage or analyse information |
“I used Google Forms” identifies a delivery tool. It does not explain the design. “I used SPSS” identifies software. It does not justify a statistical procedure. Your methodology chapter needs the connected decisions.
Use the collection-method comparison to work out the records each question requires. If you plan to reuse a dataset, the primary-versus-secondary data worksheet checks provenance, definitions and suitability before analysis.
Match the question to what you can claim
These are fictional planning comparisons, not studies that have been carried out:
| Question | Possible direction to discuss with your supervisor | Main boundary to examine |
|---|---|---|
| What practices are reported at a defined time? | A descriptive survey with a clear selection and response plan | Responses may not represent the wider population |
| Are two measured variables related? | A correlational study with appropriate measurement and analysis | Association alone cannot establish causation |
| How do people describe an experience or process? | A qualitative approach with an explained analytic method | Findings need contextual interpretation, not automatic population percentages |
| What happens when a condition is deliberately changed? | An experimental or suitable quasi-experimental approach | Comparison, allocation, timing and alternative explanations need careful treatment |
| How does a process work within one bounded organisation? | A case study using appropriate evidence sources | A single context may not support broad generalisation |
| Does an artefact meet defined requirements? | A development and evaluation plan appropriate to the discipline | Passing selected tests does not establish every real-world benefit |
These rows are not an exhaustive taxonomy. A survey can investigate relationships as well as descriptions; interviews can be used within different approaches. Timing, selection and analysis determine what the work can support.
When both quantitative and qualitative parts are justified, the mixed-methods integration worksheet helps specify where they connect and what considering them together adds. Two separate methods without that explanation leave the design incomplete.
If a bounded case fits the question, use the case-study planning worksheet to define the case, justify its selection and connect evidence to interpretations. Studying one organisation does not by itself establish that design.
If your question concerns a defined set of texts, the content-analysis guide helps separate the selected item, coded extract and classification rule. Choosing that analysis does not by itself explain which documents belong in your study or what a finding can establish.
The open textbook Research Methods in Psychology: Correlational Research explains the problem of interpreting a measured association as cause and effect. Use that distinction when choosing verbs for your title and objectives. A question about “effect” needs more justification than a questionnaire measuring two things together.
A worked decision: relationship or cause?
Fictional example: a student wants to examine study time and assessment scores. Suppose the proposed plan measures both without manipulating study time or assigning conditions.
The first design question is whether the evidence supports an association. Calling the study an experiment would misdescribe the plan. A sentence claiming that more study time caused better scores would also need evidence addressing other explanations.
Before choosing an analysis, specify how study time and scores would be measured, whether access is permitted, how records would be linked safely and how missing or inaccurate information would be handled. Then discuss assumptions and analysis with your supervisor. The statistical-test guide is a starting checklist, not an automatic test selector.
If you are building a system
Keep the build process and evaluation evidence visible as separate parts of your plan:
- Requirement: what the system should do, with a measurable acceptance condition.
- Development: how the implementation will be produced and revised.
- Test: which permitted inputs and conditions will check the requirement.
- Record: expected result, actual result and evidence location.
- Boundary: what this test does not establish.
Fictional example: a test that prevents two allocations for the same room can address a defined allocation rule. It does not, by itself, prove that users find the system easy to use or that an institution saves money. Those are different questions requiring appropriate evidence and permissions.
Do not add a user survey merely to make a software report look like a social-science project. Follow your department's requirements and explain what the evaluation is intended to establish.
Fill in a design decision sheet
| Prompt | Your project-specific answer |
|---|---|
| Main question | One question with a defined scope |
| Required evidence | What you would need to observe or obtain |
| Planned design | The approach and why it fits |
| Selection and timing | Which units, how selected and when examined |
| Measurement or evaluation | Definitions, instruments or acceptance criteria |
| Analysis | How the evidence answers the question |
| Access and ethics | Permissions, consent and safe handling required |
| Main limitation | What the plan cannot establish |
| Feasible alternative | What changes if access or resources fail |
Use the sheet before collecting data. If your approved design changes, discuss and record the change; do not rename the study afterward to hide what actually happened. The sampling guide helps with selection decisions, and the proposal guide connects the plan to the wider document.
UniDraft helps organise your written project around confirmed methods, results and instructions. It does not conduct experiments, choose your design on your supervisor's behalf or run an evaluation for you.
