Data Saturation vs Information Power: Sample Size

Choose a defensible qualitative research sample size, plan interview batches and decide when to stop using information power and data saturation.

The short answer: plan with information power, stop with evidence

There is no universally correct number of qualitative interviews. A defensible qualitative research sample size follows the decision at stake, the people whose experiences may differ, and the quality of evidence collected.

Use information power to set an initial recruitment range before fieldwork. Then analyse interviews as they happen and use a defined adequacy check to decide whether the evidence is sufficient. Information power holds that fewer participants may be needed when a sample provides more information relevant to the study aim source.

The practical sequence is:

Define the decision and population.
Assess the sample’s likely information power.
Recruit in reviewable batches.
Analyse alongside fieldwork.
Stop, continue targeted recruitment or pivot.

Published ranges are reference points, not guarantees. A systematic review found that empirical saturation studies commonly reported 9–17 interviews, chiefly with homogeneous groups and narrow objectives source. A UX study may need a different amount of evidence.

Choose methods that fit the decision before setting a sample plan. Our guide to UX research methods can help.

Data saturation vs information power: what changes?

Data saturation vs information power is not a choice between competing formulas. Information power shapes the initial plan; a defined adequacy check assesses the evidence collected.

	Information power	Data saturation or adequacy check
Purpose	Plan an initial sample	Assess collected evidence
Timing	Before recruitment	During concurrent analysis
Inputs	Aim, sample, theory, dialogue and analysis	Codes, meaning, variation, contradictions and coverage
Output	A justified recruitment range	A stop, continue or pivot decision
Limitation	Cannot guarantee rich interviews	Cannot repair an unfocused plan

Information power is a structured judgement about how much relevant information a planned sample can provide. Its considerations are the study aim, sample specificity, established theory, quality of dialogue and analysis strategy source. Recent methodological guidance also positions it as useful for estimating sample size before research begins source.

Define the adequacy target precisely. Code saturation means new topical codes rarely appear. Meaning saturation goes further: the team understands dimensions, explanations, exceptions and differences within those codes. In one 25-interview dataset, code saturation appeared at nine interviews, while meaning saturation required 16–24 interviews source.

Theoretical saturation is different again: it concerns whether a developing theory is sufficiently elaborated. Some qualitative traditions do not treat saturation as an appropriate goal. Your stopping logic must fit your analytical approach.

A stable codebook can arrive before you understand why behaviour differs by role or context. That is why when interviews beat bigger sample sizes depends on the evidence needed, not a target number alone.

Set an initial qualitative research sample size

Start with the decision, not the number. “What prevents people completing this workflow?” needs less coverage than “What should our product strategy be?”

Map characteristics that could plausibly change the answer: role, experience, organisation size, product maturity, accessibility needs, language, regulatory context or workflow frequency. Include segments because the decision could differ between them, not by habit.

Assess five information-power dimensions:

Aim breadth: Focused questions need less breadth than exploratory discovery.
Sample specificity: A tightly defined group can provide more relevant information than a broad mix.
Theory or prior evidence: Strong prior understanding focuses enquiry; uncertainty leaves more to learn.
Dialogue quality: Rich, well-probed accounts provide more information than short or uncertain conversations.
Analysis strategy: Cross-case and subgroup comparisons require broader coverage than describing one shared experience.

These are directional judgements, not a scoring system. Broad aims, uncertain dialogue and planned comparisons all point towards more evidence.

Treat relevant diversity as coverage to design for. If administrators and occasional users may face different barriers, each needs adequate representation for the comparison you intend to make. Where language affects participation or comprehension, plan appropriate support, including multilingual interviews through interpreters.

Choose a planning range rather than promising a final count. Reserve recruitment capacity for assumptions that may fail: a segment proving more varied than expected, weak interview depth or an early finding needing a disconfirming case.

Worked example: complete the sample-planning worksheet

Use this copyable worksheet to turn a UX research question into an operational plan.

Field	Planning prompt
Decision	What product or service decision will this evidence inform?
Research question	What focused question must the study answer?
Unit of analysis	Is the unit a person, account, team, workflow or organisation?
Required segments	Which characteristics could materially change the answer?
Aim breadth	Focused, moderate or broad? Why?
Sample specificity	How closely does recruitment match the decision population?
Prior evidence	What is known and what remains uncertain?
Dialogue quality	What might limit depth, recall or candour?
Analysis strategy	Are you describing one experience or comparing cases?
Initial recruitment range	What range follows from these judgements?
Review batches	After which interviews will you review evidence?
Contingency capacity	What extra recruitment is reserved, and for what trigger?
Decision owner	Who records and approves stop, continue or pivot decisions?

Here is a fully worked hypothetical B2B SaaS example. Its numbers are planning choices, not universal thresholds.

Field	Hypothetical plan
Decision	Improve setup for single sign-on configuration
Question	Where and why do administrators struggle to configure single sign-on?
Unit of analysis	An administrator completing or overseeing configuration
Required segments	First-time configurators; experienced administrators; administrators using complex identity-management arrangements
Aim and specificity	Focused workflow investigation; all participants recently configured or attempted configuration
Prior evidence	Support themes exist, but causes and workarounds are unclear
Dialogue quality	Uncertain: technical recall may be incomplete and work may be delegated
Analysis	Compare barriers, workarounds and context across segments
Initial range	12–18 interviews
Allocation	Four planned interviews per segment, with six places reserved for targeted follow-up
Review batches	Review after four, eight, 12 and then every three interviews
Owner	Research lead, with the product decision-maker reviewing the log

After each batch, check:

Were participants relevant to the decision population?
Did interviews contain concrete, recent examples?
Was probing consistent enough to compare accounts?
Did participants explain causes, not only problems?
What evidence contradicts the emerging pattern?
Which required segments remain unresolved?

Weak dialogue does not automatically mean recruiting more people. First inspect the guide, moderation and participant fit. A clearer task prompt or better follow-up question may add more information than further shallow interviews. Use a workflow-specific guide, such as this approach to write a discussion guide for user interviews.

Decide when to stop, continue or pivot

Analyse while fieldwork is running. A retrospective saturation claim cannot guide recruitment if coding waits until every interview is complete.

Maintain a review log with one row per interview or batch. Record segment coverage, new decision-relevant codes, new meanings or explanations, contradictions, evidence-quality concerns and the next action. Track new codes separately from new dimensions, exceptions and segment-specific patterns.

Evidence state	Decision	Action
Required segments are represented; accounts are rich; evidence answers the stated decision; planned checks show stable meaning	Stop	Document the judgement, limitations and final sample
New explanations, exceptions or segment differences still affect the decision	Continue targeted recruitment	Recruit for the unresolved segment, context or disconfirming case
Interviews are shallow, poorly matched or not eliciting relevant evidence	Pivot	Improve recruitment, revise the guide or reconsider the question
Recruitment or budget ends before evidence is sufficient	Stop for practical reasons	Report the constraint separately from methodological adequacy

Repetition among similar participants is not proof of sufficiency. Test plausible disconfirming cases. Predefine how apparent stability will be checked, such as a further batch covering unresolved segments or interviews testing an emerging explanation.

Good interviewing improves each unit of evidence. See our guidance on how to conduct user interviews.

Report the sample-size decision without overclaiming

Report both the plan and what happened.

We set an initial recruitment range using information power: the study had a focused aim, a specific participant profile and planned cross-case analysis across defined segments. We reviewed evidence after each batch, tracking segment coverage, new codes, new meaning, contradictions and interview quality. We stopped after [final sample] because [defined adequacy judgement]. Evidence remained limited for [unresolved segment, context or question].

Name the form of saturation or adequacy assessed. Do not write only that “saturation was reached”. State whether you assessed code saturation, meaning saturation, theoretical saturation, segment coverage or another defined standard.

Avoid treating 5, 9, 12 or 17 interviews as universal rules. The influential Guest, Bunce and Johnson study used a particular dataset of 60 interviews; it did not establish a threshold for every study source.

Participant count is not analytical sufficiency. Justify the plan from the study design, then justify the stopping decision from the evidence collected.

How many qualitative interviews are enough?

There is no universal count. The often-cited 9–17 range comes from empirical saturation studies, chiefly involving homogeneous populations and narrow objectives source. Plan from the aim, relevant variation, dialogue quality and analytical goal.

Can information power and data saturation be used together?

Yes. Use information power to set an initial recruitment range, then use a defined saturation or evidence-adequacy check during concurrent analysis to decide whether to stop, continue or pivot.

What is the difference between code and meaning saturation?

Code saturation means few or no new topical codes are appearing. Meaning saturation asks whether the team understands the dimensions, explanations, exceptions and variation within those codes.

Does participant diversity always mean more interviews?

Relevant diversity usually increases required coverage, especially for subgroup comparisons. Do not add variation mechanically: segment choices should follow the research question and product decision.

About Glasgow Research — Glasgow Research helps B2B SaaS teams turn customer and market research into product decisions. Work with us.