Data collection methods

Data collection is where dissertations are won or lost, and the damage is usually invisible until analysis. A leading question, an unpiloted survey or a sample recruited from one seminar group cannot be fixed later. It is worth getting right at the point where it costs a day rather than a term.

Choosing a method

The gap between what people say and what they do is the single most useful thing to remember. Self-report measures beliefs about behaviour, not behaviour.

Surveys

Writing questions that work

Four more rules that catch most problems:

• Avoid absolutes such as always and never; almost nobody agrees with them
• Define vague terms: "regularly" means different things to different people
• Offer a neutral or "not applicable" option, or you force false answers
• Keep it short. Completion falls sharply past ten minutes

Sample size

Run a power calculation using G*Power, which is free. You need an expected effect size from similar studies, your alpha, usually 0.05, and your desired power, usually 0.80. The calculation tells you the minimum sample for a detectable effect.

"I recruited 43 because that is who responded" is not a justification. "A power calculation indicated 128 for a medium effect; 43 were recruited, so the study is underpowered and results are indicative" is honest and scores better.

Interviews

Building a topic guide

1. Opening. Easy, factual, to settle the participant
2. Core topics. Four to six, each with one main question and prompts
3. Sensitive material. Placed later, once rapport exists
4. Closing. "Is there anything I have not asked about that matters?" This question routinely produces the best data in the interview

Practical points that matter more than they sound

• Record with two devices. Recordings fail, and there is no second chance
• Allow an hour of transcription for every fifteen minutes of audio, or budget for a service
• Learn to leave silence. Participants fill it, and what they add is often the most useful material
• Write field notes immediately afterwards, while impressions are fresh

How many

Ten to fifteen is typical for a master's dissertation, six to ten for IPA. The principle is saturation: you stop when new interviews stop producing new themes. Say how you judged it rather than asserting the word.

Focus groups

Six to eight participants, sixty to ninety minutes. Useful when you want to see how views are negotiated between people rather than held individually.

• Avoid for sensitive topics, where people will not speak candidly in a group
• Avoid where participants have unequal power, such as staff and their managers together
• Plan for dominant voices: direct questions to quieter participants explicitly
• Transcription is harder because of overlapping speech; note who spoke as you go

Observation

Decide your position in advance and record it, because it shapes what you can see and what participants will do while you are there.

Pilot everything

This is the most skipped step and the highest return. Two pilot interviews or ten pilot survey responses will show you which questions are ambiguous, which produce one-word answers, and how long it really takes.

Report the pilot in your methodology and say what you changed. It demonstrates rigour and costs a paragraph.

Ethics, in the real timeline

1. Approval must be granted before any collection, with no exceptions
2. Allow four to eight weeks, longer for vulnerable groups
3. NHS research needs HRA approval on top of university approval, which can take months
4. Prepare a participant information sheet, a consent form, and a data management plan
5. Say how data will be stored, anonymised and eventually destroyed

Data collected before approval cannot be used, however good it is. Every year students discover this the expensive way.

Last updated: 30 July 2026 · Written by the AssignWise editorial team ·Spotted something wrong?