When to use qualitative coding

Qualitative coding is a way to work systematically with text, talk, images, or observation notes. Use it when the research question needs meaning, process, or context that a closed item cannot capture.

The question has to need it

It is the right tool when you need to know how people explain a decision, how a programme is experienced, or what language a community uses for a problem. It is the wrong tool when the question is a prevalence, a difference between groups that you already know how to measure, or a test of a numerical hypothesis written before the data were seen.

Do not code because interviews exist. Code because the question requires interpretation of meaning. If you already have a survey and you want a percentage, keep the survey. If you have transcripts and you want themes, code.

Sampling sets the claim

Start from the question and the sampling. A handful of information rich cases can support a deep account. They cannot support a claim about how common a view is in a population. Do not write "most respondents" unless the design can carry that sentence.

A codebook, even a short one, should exist before a full corpus is coded, or it should be built in a documented first pass and then applied. Two coders on a subset, with disagreement discussed, is a check on whether the codes are usable. Perfect agreement is not the goal. Shared meaning is.

Deductive codes come from the protocol or the theory. Inductive codes come from the material. Most serious studies use both. What matters is that you can say which is which.

Themes should be evidenced with excerpts. Exceptions should be shown. Claims should stay inside what the cases can support.

Analysis, not decoration

Qualitative coding is analysis, not decoration. A colourful word cloud is not a finding.

If the study mixes a survey and interviews, keep the two warrants separate. The interviews can explain a pattern. They cannot, on their own, measure it.