Writing a Qualitative Sampling Plan
Justify whom a qualitative study needs to hear from and record achieved coverage, rather than treating an interview count as sufficient evidence.
A qualitative sampling plan is a short, defensible argument that links the decision a study must inform to the perspectives it selects, the reason each was selected, the route by which each can be reached, and what the coverage actually achieved permits a reader to conclude. It is written before fieldwork and revised after it, because the record of who could not be reached carries as much interpretive weight as the list of who was interviewed. The plan is not a recruitment schedule with a target number attached; it is the reasoning that makes the eventual findings assessable.
Start from the inference the study must support
Sampling decisions become arbitrary when they are detached from purpose. Where a study is asked to describe how a process works in practice, it needs coverage of the roles that carry out and receive that process. Explaining why outcomes diverge requires coverage of the circumstances associated with divergence, not simply a longer list of participants. To establish whether a documented procedure is followed, the study needs sites or cases where the procedure is under pressure as well as where it is routine.
State the intended inference in one sentence, then name the variation that could change the answer. Where the purpose is contested among sponsors, record the version the plan is built on; a later change of purpose is a reason to revise the sample, not to reinterpret the original one.
Deciding which variation matters
Variation is rarely a single axis, and a plan that treats it as one tends to over-claim. Four dimensions recur in policy and social research:
- Structural position: where a person sits in a process (commissioner, provider, front-line staff, recipient, oversight body).
- Circumstance: recency of contact, outcome experienced, duration of involvement, locality, household or organisational setting.
- Institutional interest: whether a source has a stake in the finding being favourable, unfavourable or unnoticed.
- Access route: whether a person can be identified and approached directly or only through an intermediary.
Name the dimension the study treats as primary, which are secondary, and why. A sample stratified by role but not by outcome may describe a service accurately while saying nothing about why experiences differ. A sample stratified by outcome but not by role may capture dissatisfaction without the operational context that produced it.
A perspective map, set out before recruitment begins, makes these choices visible and gives reviewers something specific to challenge. The format below is a suggested working structure rather than a validated instrument.
| Perspective and role | Knowledge sought | Why this variation is included | Access route | Potential interest or bias | Consequence if absent |
|---|---|---|---|---|---|
| e.g. front-line adviser | how rules are applied in practice | holds discretion not visible in guidance | team manager referral | may understate shortcuts taken | account of process would rest on documents alone |
Filling the final column honestly at the planning stage is the most useful discipline the table imposes. It converts a list of intended participants into a statement of what the study will be unable to say.
Identifying who is missing
Deliberate inclusion and foreseeable absence should be treated as separate judgements. Some groups are excluded by design because the question does not concern them; others are excluded by circumstance, and that exclusion changes what the findings mean.
Groups that frequently fall outside a sampling frame include former users and people who disengaged, people who declined at first contact, staff who have left since the period under study, organisations outside the commissioning relationship, and anyone whose contact depends on a caseworker, interpreter or advocate. Each absence has a specific consequence. If former users cannot be reached, the study can describe how current users experience a service; it cannot compare continuing with leaving, and statements about unmet demand become inferences from the accounts of those who stayed.
It also helps to distinguish four different facts that are often collapsed into the phrase "not sampled": not selected, selected but not identifiable in available records, identified but not reached, and reached but declined. These carry different weight when findings are reported, and a reader cannot tell them apart unless the plan records them separately.
How recruitment constraints shape interpretation
Access is an analytic constraint, not only an administrative one. Where recruitment runs through a provider, commissioner or caseworker, the intermediary's view of who is relevant and who can be approached becomes part of the selection mechanism. If staff are asked to refer participants, the sample may over-represent people whom staff consider articulate, cooperative or representative of the service at its best. If a public body will not release contact details, the sample may systematically exclude the people most likely to have a complaint.
The plan should therefore record the route for each perspective and the likely direction of the distortion it introduces, rather than noting only that a route exists. Conditions matter here: where ethical or legal constraints rule out direct approach (for example, contacting people in acute distress, or contacting service users without the operator's cooperation), the plan should state the constraint and the resulting limit on claims, not substitute an assurance of coverage that the method cannot deliver.
Justifying the extent of sampling
There is no universal sample size for qualitative work, and a plan that offers one without argument invites the reader to substitute intuition for reasoning. A defensible justification states the stopping rule in terms the study can actually apply: completion of a defined matrix of perspectives and circumstances, coverage of the specified variation, or the point at which additional cases cease to change the analysis, with the basis for that judgement described in advance. A numerical target may be useful for planning, but the justification should rest on the variation the number is intended to deliver, and the plan should say what will happen if the target is met before the variation is covered, or the reverse.
HM Treasury's Magenta Book guidance, in the chapter Quality in Qualitative Evaluation, sets out an appraisal framework for evaluation in UK central government that asks whether design and sampling are justified, whether coverage and missing perspectives are explained, and whether collection, analysis and reporting are transparent. The framework does not prescribe a sample size or favour a particular sampling method; its use here is to confirm that justification and the explanation of absence are part of the quality of the design, not a retrospective caveat. A practical plan should therefore be written so that these questions can be answered from the document itself.
A sampling justification record
The worksheet below is a suggested working format. It is not a validated instrument or an institutional requirement, and it should be adapted to the study's purpose; its value lies in keeping justification, access and absence in one place through revision.
| Field | What to record |
|---|---|
| Decision or question | the inference the study must support, in one sentence |
| Perspective and role | who is being sampled, described in terms of position rather than name |
| Selection rationale | the variation this perspective represents and why that variation matters to the question |
| Access route | how the perspective is identified and approached, and which intermediaries are involved |
| Anticipated interest | the stake a source may have in the finding, and the direction of possible distortion |
| Achieved coverage | approached, identified, reached, recruited and included, with dates |
| Absent or thin cases | who is missing, at which stage, and on what grounds |
| Consequence for interpretation | claims the evidence supports, claims withheld, and language that should not be used |
| Revisions | what changed after fieldwork began and why, dated |
Completing the record is not the same as resolving every gap. Its purpose is to make the gap legible to a reader who will not see the fieldwork.
A fictional worked example
The following example is invented to illustrate the worksheet; it is not a description of a real study or client.
A fictional district council commissions a study of a dial-a-ride transport service for people with mobility restrictions. The intended inference concerns why eligible residents stop using the service. The perspective map includes current users, carers who book on behalf of others, service administrators, and former users who left within the past two years. Access to current users runs through the operator's booking list; access to former users would require the operator to write to them, since records cannot be released.
The operator agrees to the first route and declines the second, citing data protection and concern about complaints. Fieldwork yields fourteen current users, six carers and three administrators, and no former users. The record notes that former users were identified as a perspective, that the route existed in principle, that the intermediary declined, and that no alternative route was available within the study period.
The consequence is stated plainly. The study can describe how the service is used and where booking breaks down for people who continue to use it. It cannot explain why eligible residents stop using the service, and any statement about unmet demand rests on current users' accounts of the service rather than on evidence from those who left. Reports and summaries should describe the sample as current users, carers and administrators of the service, not as users of the service.
Keeping the record current
Date each version of the plan and record decisions to broaden, narrow or substitute perspectives. When a gatekeeper changes the route, or a perspective proves thinner than expected, the revision is part of the evidence rather than an admission of failure. A plan that shows its own adjustments gives a reader the means to judge the findings; one that presents the final sample as if it had been the intended sample conceals exactly the information the reader needs.
For organisations commissioning qualitative work, the for-organisations page sets out how project-based research support is scoped; researchers considering such projects may find the guidance on what to establish before accepting an expert project a useful companion, particularly where sampling and access constraints need to be agreed before fieldwork is costed.