What qualitative research is for, which design answers which kind of question, and how to defend a small sample properly.
Qualitative research answers questions about meaning, process and experience: how people understand something, how a situation unfolds, why a practice persists. It is not the option you take when numbers feel intimidating — it answers a different kind of question, and it answers it in depth rather than in breadth. Its findings are strong on understanding and are not designed to establish how common something is.
Qualitative work earns its place when the thing you want to know cannot be counted without being destroyed. If the answer you need is a number, this is the wrong family. If the answer you need is an account — of how something is understood, negotiated, resisted or lived with — nothing else will get you there.
Typical questions it handles well:
And the questions it cannot answer, however good the study: how widespread something is, whether one group differs from another in a measurable way, or whether an intervention caused an outcome. Asking a qualitative study to do any of those is the most common way a well-run project ends up in trouble at examination.
“Qualitative” is a family, not a method. Naming a design is not a formality: each one carries commitments about what you collect, how you analyse it and what you may conclude. Adopting the label without the commitments is something examiners spot immediately.
Studies the lived experience of a phenomenon — what it is actually like to go through something, described as closely as possible to how it is experienced. Commits you to participants who have genuinely lived it, to interviews that stay with experience rather than opinion, and to an analysis that resists explaining too early. Descriptive and interpretative traditions differ on how far the researcher’s own reading belongs in the account, and you are expected to know which you are following.
Builds an explanatory theory upwards from data rather than testing one downwards. Commits you to collecting and analysing at the same time, to sampling decided by what the developing theory needs next rather than fixed in advance, and to constant comparison. Doing thematic analysis and calling it grounded theory is one of the most frequent mislabels in doctoral work; if you did not sample theoretically, it is not grounded theory.
Studies something in depth within its real setting, where the boundary between the case and its context is genuinely unclear. Commits you to defining the case explicitly, to more than one source of evidence, and to a claim that generalises to theory rather than to a population. A single case is entirely defensible — provided you say why this case earns the attention.
Studies a culture or community from inside, over time. Commits you to sustained presence, to fieldnotes as data, and to reflexivity about how your being there changed what happened. It is not a short study, and calling three site visits an ethnography will not survive examination.
Treats the story itself as the unit of analysis — how people construct accounts of their lives and what those constructions do. Commits you to keeping accounts whole rather than fragmenting them into codes, which sits awkwardly with standard thematic coding and needs to be handled deliberately.
Studies how language itself constructs the objects it appears merely to describe. Commits you to treating text as action rather than as a window onto what people think, and to a theoretical position about language that has to be stated, because the different traditions here are genuinely incompatible.
Qualitative sampling does not aim at representativeness, and treating it as a failed attempt at representativeness is what produces indefensible chapters. You are selecting for information, not for coverage: the people, sites or documents most able to inform the question.
| Strategy | What it selects for | Best when |
|---|---|---|
| Purposive | People who meet criteria the question requires | The default for most doctoral qualitative work |
| Theoretical | Whatever the developing theory needs next | Grounded theory, and only there |
| Maximum variation | Deliberate spread across relevant differences | You want patterns that hold despite variety |
| Snowball | Referral through existing participants | Hard-to-reach or hidden populations |
| Convenience | Who was available | Rarely defensible alone; state it plainly if used |
Whichever you use, the chapter has to say what you selected for and why that supports the question. “Twenty participants were recruited” explains nothing; “participants were selected to span both hospitals and all three grades, because the question concerns how the policy is interpreted differently across them” explains everything.
Saturation is the most over-claimed word in qualitative methods. It is routinely used to mean “I stopped interviewing”, and examiners know it.
Used properly, it means that further data collection stopped changing the developing analysis — which presupposes that you were analysing as you collected. If all twenty interviews were done first and coded afterwards, you cannot claim saturation, because there was no developing analysis to stop changing. Claiming it anyway is a straightforward invitation to be questioned about it.
Two honest alternatives are available and are increasingly respected. You can state the sample was determined by what the design and the question needed, and defend that directly. Or you can report that you continued until new material was adding refinement rather than anything new, and show the evidence — a record of when each code first appeared, for instance. The second is more work and considerably stronger.
Either way, be wary of numerical rules quoted as though they were settled. Sample size in qualitative work depends on how narrow the question is, how homogeneous the participants are, and how rich each account is; no fixed number covers those.
Qualitative research is judged, but not by reliability coefficients. The usual criteria are credibility, transferability, dependability and confirmability, and each is demonstrated by something you actually did rather than by asserting it.
The strongest single move available here is the deliberate search for disconfirming cases. A chapter that reports what did not fit, and what was done about it, is markedly harder to challenge than one where everything agrees.
Research that seeks to understand meaning, process and experience through material such as interviews, observation, documents and open text, rather than through measurement. It asks how and why questions, works in depth with relatively few cases, and produces interpretive accounts rather than estimates of how common something is.
There is no fixed number, and any figure quoted as a rule is a rough convention rather than a requirement. What decides it is how narrow the question is, how similar participants are to each other, how rich each account is, and which design you are using — grounded theory samples until the theory stops developing, while a phenomenological study may work closely with very few. Justify the number by what the design needed, and be honest about practical constraints.
Qualitative research covers the whole design: the question, the approach, who you recruit, what you collect and how quality is judged. Qualitative data analysis is what you do with the material once you have it — coding it, developing themes, and interpreting them. The design decides what the analysis is allowed to conclude, which is why the two need to be argued separately in the thesis.
No, though it is judged by different criteria. Rigour in qualitative work means a defensible sampling logic, systematic analysis, an audit trail, active attention to cases that do not fit, and reflexivity about the researcher’s influence. A qualitative study that shows all of that is more rigorous than a survey with a large sample, a poorly validated instrument and unchecked assumptions.
Yes, and for some questions it is clearly the right choice — when you need both the size of an effect and an explanation of it. The requirement is genuine integration: saying in advance where the two strands will meet and what you will do if they disagree. Strands reported one after another without ever meeting produce two thin studies rather than one strong one.
Sampling logic and design commitments are cheap to fix beforehand and difficult afterwards. Describe your question and who you plan to talk to, and a PhD in your field will tell you whether it holds together.
Leave an email or a WhatsApp number — whichever you prefer — and tick how we should reach you. We reach out within 30 minutes.
