Designing a study

“My supervisor says my sample is too small. Can I defend it?”

Saturation, information power, and a sample-size argument you can show.

You did not pick your number carelessly. But “it felt like enough” is not an argument, and a supervisor or reviewer who says too small is asking for one. This page builds the argument, in words you can paste into your methods.

By Heidi Andersén, MD, PhD · Updated 18 July 2026 · 7 min read

“I really have to justify my decision to include 20 participants for an interview… there are hardly any guidelines for establishing sample size.” — a doctoral student, on defending a qualitative N

The problemThe number is fine. The justification is missing.

The event is almost always the same: you name a sample size, and someone with power over your project says it is too small. Sometimes it is a supervisor in a meeting, sometimes Reviewer 2 in writing. What they are really asking is not “add more people” — it is “show me why this many is defensible.” Qualitative work rarely comes with a power calculation to hide behind, so the burden falls on your reasoning. The good news: qualitative methodology has two well-established ways to reason about “enough”, and a supervisor who hears either one, applied honestly, has very little left to push on.

Why it mattersAn undefended N is the easiest thing to attack.

A sample size with no stated logic is a gift to a reviewer: it is concrete, it looks like a weakness, and it costs nothing to flag. Left unaddressed it can hold up an examination or trigger a revision. Addressed well — in one clear paragraph — it moves from your weakest sentence to one of your most defensible, because you have shown you understood the question before it was asked.

Common mistakesFour ways the defence goes wrong

Citing a magic number. “12 to 15 interviews is standard” borrowed from a paper in another field, with another question, is not a justification — it is a coincidence.
Claiming saturation you did not track. Saying “saturation was reached” with no record of when new codes stopped appearing invites the obvious follow-up: how do you know?
Treating saturation as a formula. Saturation is a judgement about your data, not a threshold that fires at interview 17. Presenting it as automatic reads as a shortcut.
Confusing sample size with information power. A narrow question with rich, on-target participants needs fewer people — not more. Arguing headcount alone misses the stronger case.

The minimal explanationTwo lenses, one paragraph

You only need two concepts, and you should name the one that fits your study rather than reaching for both.

Saturation. The point at which further data collection stops changing your analysis — no new codes, then no new dimensions of existing codes. It is a claim about your coding, so it only means something if you kept a record of when new codes appeared and stopped. Distinguish code saturation (no new codes) from meaning saturation (you fully understand each code); the second takes longer and is the stronger claim.

Information power (Malterud et al., 2016). The idea that the more information your sample holds relevant to your question, the fewer participants you need. It turns five things into an argument: a narrow aim, a specific sample, established theory, strong dialogue in the interviews, and a case (rather than cross-case) analysis each lower the N you need. This is the lens that lets you defend 20, or even fewer, without appealing to saturation at all.

Pick the lens that matches how you actually decided, and write the argument around it. Assess your own study against the five information-power dimensions; do not claim a number the dimensions do not support.

Worked exampleDefending 20 interviews

A study you could write tomorrow

Study. 20 semi-structured interviews with newly qualified nurses about moral distress in their first year.

The argument, information-power style:

Aim — narrow: one profession, one career stage, one phenomenon. (Fewer needed.)
Sample — specific: participants selected precisely for first-year experience. (Fewer needed.)
Theory — an existing moral-distress framework guided the questions. (Fewer needed.)
Dialogue — 60–90 minute interviews with a skilled interviewer produced dense accounts. (Fewer needed.)
Analysis — cross-case thematic analysis, not a single deep case. (Slightly more needed — the one dimension pushing up.)

Written up: “A sample of 20 was judged adequate on the information-power model (Malterud et al., 2016): a narrow aim, a highly specific sample, theory-informed questioning, and strong interview dialogue each support a smaller sample, while the cross-case analytic strategy was accommodated by extending recruitment to 20. Coding was monitored for new codes; the final three interviews produced none, consistent with code saturation.”

That is a paragraph a supervisor can read once and accept. It names a model, applies it to this study, and reports what actually happened in the coding — not a number pulled from another field.

Checklist & templateBuild your own defence

Downloadable · sample-size defence

The four-part defence one-pager

State your aim and how narrow it is.
State how specific your sample is to that aim.
Name the theory or framework guiding your questions.
Describe the quality of the interview dialogue (length, depth, rapport).
Name your analytic strategy (case vs cross-case) and its effect on N.
Report what your coding showed: when new codes stopped, if they did.
Write the single paragraph, then read it aloud to your supervisor’s likely objection.

Sample-size defence — one-pager Study: ____________________________________________ Aim (how narrow?): ________________________________ Sample specificity: _______________________________ Guiding theory/framework: _________________________ Interview dialogue (length, depth): _______________ Analytic strategy (case / cross-case): ____________ Coding record (when new codes stopped): ___________ Draft paragraph: "A sample of ___ was judged adequate on the information-power model (Malterud et al., 2016): [narrow aim], [specific sample], [theory-informed questions] and [strong dialogue] support a smaller sample; [analytic strategy] was accommodated by ___. Coding was monitored for new codes; ______________." Records your reasoning — not a verdict on adequacy. You make the judgement; this keeps it explainable.

Download the one-pager (Markdown)

The honest limit. No model makes a sample size correct — saturation and information power make your reasoning visible and checkable, which is what a defence actually needs. The benefit is exactly that: a reviewer can follow your logic instead of guessing at it.

Related guidesRead next

Recommended tool · the next step

MethodVahti — Sample Defence Pack

Want the written defence to paste into your methods? MethodVahti sets the stopping logic, keeps an adequacy log as you code, and drafts the sampling paragraph from your own numbers — without treating saturation as a formula. You check it, you own it.

See the Sample Defence Pack · €49 →