Start with your question, not your method

The most common mistake in undergraduate and master's dissertations is choosing a method before knowing what you actually want to find out. It usually happens innocently enough: you attend a workshop on questionnaire design, get excited about a dataset, or hear that interviews are "more interesting", and suddenly the method comes first. Your research question then gets bent to fit it.

Try the reverse. Write your aim as a single sentence, then ask what kind of answer would genuinely satisfy it. If the answer is a number, a rate or a comparison, you are heading towards quantitative work. If the answer is a description of meaning, experience or process, you are heading towards qualitative work. If you need both — a sense of how widespread something is and an explanation of why it happens — mixed methods may be the honest answer.

A useful test: ask yourself what a reader would need in order to believe your claim. That tells you what kind of evidence you need, and therefore what kind of method.

Quantitative methods: measuring patterns at scale

Quantitative approaches suit questions about prevalence, frequency, difference, association and prediction. They work best when your key concepts can be turned into variables — age, income, hours studied, score on a validated scale — and when you can reach enough participants to detect a pattern rather than a coincidence.

• Typical designs: closed-question surveys, secondary analysis of large national datasets, experiments, quasi-experiments, structured observation, content analysis with numerical coding.
• Sampling: probability sampling (random, stratified, cluster) lets you generalise to a population; convenience sampling does not, so describe it honestly.
• Analysis: descriptive statistics first, then inferential tests — t-tests, chi-square, correlation, regression — chosen to match your variables and design.

Two practical warnings. First, statistical significance is not the same as importance: with a large sample, a trivial effect can be significant, so report effect sizes and confidence intervals alongside p-values. Second, the quality of your questionnaire determines everything downstream. Pilot it on a handful of people before you circulate it properly, and use established scales where they exist rather than inventing your own.

Qualitative methods: understanding meaning and process

Qualitative approaches suit questions beginning with how, why and what is it like. They are the right choice when the point of the study is to understand how people interpret their situation, or to trace how something unfolds over time in a particular setting.

• Typical designs: semi-structured interviews, focus groups, ethnography and observation, documentary and policy analysis, narrative inquiry, visual methods.
• Sampling: purposive and often small — you select participants because they can illuminate the question, not because they represent a population.
• Analysis: thematic analysis, interpretative phenomenological analysis, discourse analysis, grounded theory, framework analysis.

Roughly eight to fifteen interviews is often enough for a focused undergraduate project, but the real test is informational: stop when new interviews stop producing new insights. Depth beats breadth here, and your own position as a researcher is part of the picture. A short reflexivity note explaining how your background or assumptions may have shaped the data is expected, not a confession of bias.

Mixed methods: when one approach isn't enough

Mixed methods combine both, and they are genuinely powerful when your question has two halves. A common pattern is the explanatory sequential design: a survey establishes how common something is, then interviews explore why. The exploratory sequential design runs the other way — interviews first to generate themes, then a survey to test how widely they hold. A convergent design collects both at once and compares the findings.

Be careful here. Bolting two extra interviews onto a survey is not mixed methods; it is a survey with decoration. You need a clear rationale for why each strand is necessary and how the two speak to each other. Mixed methods also roughly double your workload, so discuss it with your supervisor before committing.

Fit, feasibility and ethics

Before you finalise anything, run through the practical checks. A method that is intellectually perfect but undeliverable will cost you marks.

• Access: can you actually recruit these participants, or get permission to use this dataset or site?
• Time: ethnography, longitudinal designs and large surveys rarely fit a ten-week project.
• Skills: are you confident with the statistical test or coding approach you have chosen? If not, build in time to learn it.
• Ethics: approval takes longer than you think, especially with vulnerable participants or sensitive topics.
• Support: does your supervisor have expertise in this method? Their guidance is worth more than a trendier design.

Justifying your choice in writing

Your methods chapter should read as a series of reasoned decisions, not a list of what you did. For each choice — design, sampling, instruments, analysis — state what you did, then why it suited your aim, then what the trade-off was. "A cross-sectional survey was chosen because the aim was to estimate prevalence across a large student population; the limitation is that it cannot establish causal direction."

Cite methodological literature to show your choices are grounded rather than arbitrary, and frame limitations as trade-offs rather than failures. A reader who finishes your chapter should be able to say: the question demanded this evidence, and this is the evidence they gathered. Get that alignment right, and the rest of the dissertation becomes far easier to write.