Purposive sampling in qualitative research: 6 types with examples

Purposive sampling (also called purposeful sampling) means deliberately selecting participants because they can best answer your research question, not because they were randomly drawn from a population. In qualitative research, you choose who can tell you the most, not who represents the most. There are six main types, each suited to different research questions and designs.

This guide covers all six types, when to use each, how to justify your choice to reviewers and supervisors, and how purposive sampling connects to data saturation and theoretical saturation. If you are a PhD student deciding how to frame your sampling strategy, start here.

What is purposive sampling?

Purposive sampling is a non-probability sampling approach in which the researcher uses deliberate judgement to select participants, cases, or documents that are most relevant to the research purpose. The term was popularised by Michael Quinn Patton, whose Qualitative Research & Evaluation Methods (2002) remains the most-cited reference on the topic. Patton argued that qualitative inquiry should focus on "information-rich cases": those from which you can learn the most about the phenomenon you are studying.

The underlying logic is different from quantitative sampling. In a survey study, random selection is crucial because you want your sample to represent the population statistically. In qualitative research, you are not aiming for statistical representation. You are aiming for conceptual depth. You want participants who have lived through the experience, made the decisions, or navigated the context you are investigating. Choosing them deliberately is not a weakness. It is the point.

"The logic and power of purposeful sampling lies in selecting information-rich cases for study in depth. Information-rich cases are those from which one can learn a great deal about issues of central importance to the purpose of the inquiry." Michael Quinn Patton (2002)

Purposive sampling is the dominant approach across most qualitative research methods: thematic analysis, interpretive phenomenological analysis, grounded theory, case study research, and ethnography all rely on it. When a methods section says "participants were selected purposively," it signals intentional, criteria-driven selection rather than chance.

What are the 6 main types of purposive sampling?

Patton's original typology identified 16 purposeful sampling strategies, but six of them account for the vast majority of qualitative studies. Each serves a distinct purpose and fits different kinds of research questions.

Type	Core logic	Best for	Typical sample size
Maximum variation	Capture the widest range of perspectives	Exploratory studies, policy research	20–40+
Homogeneous	Reduce variation to study a specific group in depth	IPA, phenomenology, focused studies	6–15
Typical case	Represent the most common or ordinary experience	Illustrative or descriptive studies	8–20
Extreme or deviant case	Learn from outliers	Explaining unusual outcomes, innovation research	3–10
Critical case	Find cases that can make or break a hypothesis	Theory testing, programme evaluation	1–5
Snowball	Reach hidden or hard-to-access populations	Marginalised groups, sensitive topics	Variable

1. Maximum variation sampling

Maximum variation sampling (also called heterogeneous sampling) deliberately seeks the widest possible range of participants across the dimensions most relevant to your research question. Rather than focusing on a uniform group, you map out key axes of variation and then select cases that span those axes.

For example, if you are studying how PhD students manage academic stress, your axes of variation might include discipline (STEM vs. humanities), year of study, funding status, and nationality. You would then recruit participants who differ substantially across these dimensions, aiming to represent the full spectrum of experience rather than a single segment.

The analytical payoff is distinctive. When you analyse your data and find that a theme or pattern holds across participants who differ in nearly every other respect, that convergence is analytically powerful. As Patton notes, the resulting common patterns across cases are especially useful because they cut across diversity.

Maximum variation sampling is a strong default choice for exploratory studies where you do not yet know which participant characteristics matter most. It is also common in policy and programme research, where the goal is to understand how a programme lands across a heterogeneous population of stakeholders.

Practical example: A study of how small businesses adapted to remote working during the pandemic might use maximum variation sampling to recruit owners across industry sectors, business sizes, urban vs. rural locations, and countries.

2. Homogeneous sampling

Homogeneous sampling is the opposite of maximum variation. You deliberately narrow your sample to a specific, well-defined group in order to study that group in depth. Reducing variation is not a flaw here. It is the strategy.

This approach works best when your research question is specifically about a particular type of person or experience. If you are asking "how do newly qualified secondary school teachers in Finland experience their first year in the classroom?", you want participants who are genuinely newly qualified, genuinely teachers, genuinely in Finland. Including experienced teachers, or teachers from other countries, would muddy the conceptual water.

Homogeneous sampling is the standard approach in interpretive phenomenological analysis (IPA), where the goal is to understand a shared experience from the inside. IPA studies typically use small, homogeneous samples (often six to ten participants) because the method prizes depth of engagement over breadth of variation.

Practical example: A study of burnout among hospital nurses in intensive care units would recruit only ICU nurses, not nurses across all wards, to ensure the specific pressures of that environment are the focus.

3. Typical case sampling

Typical case sampling means selecting participants or cases that represent the most ordinary or average version of the phenomenon you are studying. You are not looking for extremes. You want to understand what the normal experience looks like.

This sounds straightforward, but it requires that you first have enough contextual knowledge to know what "typical" looks like in your study context. Researchers often use a preliminary survey, prior literature, or key informant interviews to establish what counts as typical before selecting their sample.

Typical case sampling is useful for descriptive or illustrative studies where the goal is to give readers a reliable portrait of the common experience. It is also a good choice when you are writing for audiences (practitioners, policymakers, clinicians) who need to understand what the ordinary situation looks like before you can introduce variation or exceptions.

Practical example: A study of how customers experience the onboarding process at a software company would recruit users who completed onboarding within the standard timeframe and were not flagged as unusually satisfied or unusually churned.

4. Extreme or deviant case sampling

Extreme case sampling (sometimes called deviant case sampling) deliberately selects participants at the ends of the distribution: cases that are unusually successful, unusually difficult, or otherwise exceptional. The idea is that extremes often illuminate mechanisms that are harder to see in the middle.

This approach has deep roots in organisational research and innovation studies. A company that has sustained quality under conditions where all others have failed is analytically interesting precisely because it has broken the pattern. By studying it closely, you may identify factors that are causally important but invisible in average cases.

Extreme case sampling is not about sensationalism. It is about using outliers as a lens. Because extreme cases often distil processes or mechanisms in concentrated form, they can generate conceptual insights that are later testable across a broader sample.

Practical example: A study of digital transformation in manufacturing might select one factory that completed a highly successful Industry 4.0 transformation and one that experienced a costly failure, in order to understand what differentiates the two.

5. Critical case sampling

Critical case sampling identifies the single case (or small number of cases) where, if the pattern holds, it is likely to hold elsewhere, or where, if it does not hold, it is unlikely to hold anywhere. The logic is economical: you use one strategically chosen case to make a strong inference.

Patton's canonical illustration is a community that opposes a proposed industrial development. If the most conservative, development-friendly community in a region is opposed, then it is very likely that all communities in the region are opposed. That one critical case does the analytical work of many.

Critical case sampling is particularly useful in evaluation research, where resources are limited and you need to establish proof of concept. It is also powerful for theory testing: a well-chosen critical case can disconfirm or strongly support a hypothesis more efficiently than a large sample of average cases.

Practical example: A researcher studying whether a new supervision model improves employee wellbeing might select the department with the most resistant management culture. If wellbeing improves even there, the model's effectiveness is credibly established.

6. Snowball sampling (and a note on theoretical sampling)

Snowball sampling (also known as chain-referral sampling) works by asking initial participants to refer the researcher to others who fit the study criteria. Those participants in turn refer further participants, creating a chain that "snowballs" outward through a social network.

This approach is most valuable when your target population is hard to identify through conventional means: because members are dispersed, because the topic is sensitive, or because no public register exists. Studies of undocumented migrants, people with stigmatised conditions, underground economic activity, or closed professional networks often rely on snowball sampling because there is no other practical route to participants.

The key limitation is that snowball samples tend to be homophilous: people refer others who are similar to themselves. This can introduce systematic bias if network position correlates with the variables you are studying. Good practice is to seed the snowball with several different starting points from different network positions, to maximise the range of perspectives you capture.

Practical example: A study of informal caregiving among elderly populations in rural communities might begin with participants identified through a local GP surgery, then use referrals to reach caregivers who are not registered with formal support services.

A note on theoretical sampling. Theoretical sampling is a related but distinct concept that comes specifically from grounded theory. The key difference is timing: in purposive sampling, you decide your sampling criteria before data collection begins. In theoretical sampling, sampling decisions emerge during analysis. As you code your initial data and begin to identify conceptual categories, you collect further data specifically to develop, refine, or saturate those categories. Theoretical sampling is inherently iterative: you sample in order to develop theory, not to answer a pre-specified question.

How does purposive sampling relate to data saturation?

Purposive sampling answers the question of who you select. Data saturation addresses the question of when you stop. The two concepts are closely linked but serve different functions.

Saturation is reached when additional interviews or data sources stop producing new themes, categories, or conceptual insights. In practice, a systematic review of empirical studies found that thematic saturation typically requires between 9 and 17 interviews, while meaning saturation (understanding the full range of how people experience a phenomenon) may require up to 24 interviews. The exact number depends heavily on your sampling strategy.

A well-designed purposive sample can reach saturation more efficiently than a poorly designed one. If your sample lacks variation across relevant dimensions, you may reach apparent saturation quickly, but only because everyone is telling you the same thing, not because you have genuinely exhausted the conceptual space. Maximum variation sampling, by contrast, surfaces variation early and often, which means you are more likely to reach genuine saturation by the time participants start repeating patterns you have already identified.

For sample size decisions in qualitative research, your sampling strategy and your saturation criteria should be designed together, not independently. When you write your methods section, linking these two explicitly ("we used maximum variation sampling across three axes of diversity, and continued recruiting until no new themes emerged across three consecutive interviews") is far stronger than treating them as separate decisions. See our companion guide on how many interviews you need for a fuller treatment of saturation thresholds by method.

How do you justify purposive sampling to reviewers and supervisors?

The most common concern that reviewers raise about purposive sampling is generalisability: "How can we know these findings apply beyond your sample?" The short answer is that generalisability in the statistical sense is not the goal of qualitative research. The longer answer requires you to address three things explicitly in your methods section.

First, state your selection criteria clearly. Describe who qualified for inclusion and who did not, and why. Vague criteria ("experienced professionals") are harder to defend than specific ones ("professionals with at least five years of post-qualification practice in a UK NHS trust"). Specific criteria signal that selection was deliberate and systematic, not opportunistic.

Second, explain why your sampling strategy fits your research question. If you used homogeneous sampling, explain what you lose analytically by including variation and what you gain by controlling for it. If you used maximum variation sampling, explain what dimensions you varied across and why those dimensions are theoretically meaningful. The justification should connect back to your research question and epistemological position.

Third, acknowledge the limitations. Reviewers are not expecting you to claim your findings are universally applicable. They are expecting you to show you understand the scope conditions of your conclusions. A sentence like "findings are transferable to similar organisational contexts rather than statistically generalisable to a broader population" is stronger than saying nothing about generalisability, because it shows you understand the distinction.

One additional consideration: if you are working in a research design on a limited budget, document your access routes and any gatekeeping challenges that shaped your final sample. Reviewers understand that recruitment rarely goes exactly to plan, and transparency about how practical constraints shaped your sample strengthens rather than weakens your methods section.

How do you write purposive sampling in your methods section?

Your methods section needs to cover four things related to sampling:

The sampling approach and type. Name the strategy explicitly ("we used maximum variation purposive sampling") and cite Patton (2002) or another methodological authority.
The selection criteria. State your inclusion and exclusion criteria and the rationale for each.
The recruitment process. Describe how you identified and approached potential participants.
The final sample. Report the size, key characteristics, and (where relevant) a brief justification for why the sample was sufficient.

A common mistake is to say only "participants were selected purposively" without specifying which type of purposive sampling or why. This leaves reviewers unable to evaluate whether the sampling strategy was appropriate, which generates exactly the kind of vague methodological query you want to avoid.

For interview-based studies, pairing your sampling strategy description with a participant characteristics table (age, gender, years of experience, role, or whatever dimensions are relevant) makes it immediately clear whether your stated criteria were genuinely enacted. This is especially important if you are using maximum variation or homogeneous sampling, where reviewers will want to see that the variation (or lack of it) you claimed is visible in the sample.

What are the most common mistakes in purposive sampling?

Confusing purposive sampling with convenience sampling. Convenience sampling means selecting whoever is easiest to access. Purposive sampling means selecting whoever best answers your research question. The distinction matters both analytically and ethically. If your "purposive" sample consists entirely of colleagues you already knew, reviewers will question whether your criteria were genuinely applied or whether convenience drove selection.

Selecting on the dependent variable. If you are studying why certain companies fail, and you only interview managers at failed companies, you cannot identify the factors that distinguish failure from success, because you have no comparison point. Depending on your research question, you may need variation on the outcome, not just on background characteristics.

Treating sampling strategy and sample size as independent decisions. As discussed above, your sampling type, your saturation criteria, and your eventual sample size are logically connected. Deciding on a sample size before you have committed to a sampling strategy (or before you have started analysis) is often premature. Build in flexibility, and justify your final size with reference to saturation.

Under-specifying selection criteria. "Experienced," "expert," and "senior" are not criteria. They are adjectives. Define them concretely: how many years of experience, what kind of expertise, what level within the organisation. Concrete criteria make your selection replicable in principle and defensible in review.

Neglecting reflexivity. Your position as a researcher shapes who you can access and who will talk to you. A brief reflexivity statement acknowledging how your background, networks, or disciplinary position may have influenced your sample is increasingly expected in academic qualitative work. See our guide on reflexive thematic analysis for more on researcher reflexivity in the write-up.

Frequently asked questions

What is the difference between purposive sampling and random sampling?

Random sampling gives every member of a population an equal chance of selection, aiming for statistical representativeness. Purposive sampling selects participants based on their relevance to the research question, aiming for conceptual depth. In qualitative research, the goal is to understand a phenomenon in depth, not to represent a population statistically, so purposive sampling is more appropriate than random sampling for most qualitative studies.

Can you combine different types of purposive sampling?

Yes, and in practice many studies do. A common pattern is to begin with maximum variation sampling to establish the breadth of the phenomenon, then follow up with extreme case or homogeneous sampling to investigate specific dimensions in more depth. Palinkas et al. (2015) found that combining sampling strategies was more appropriate for complex implementation research than relying on a single approach. When combining strategies, name each one and explain when and why you shifted between them.

How many participants do I need for a purposive sample?

There is no universal answer. Sample size depends on the type of purposive sampling, the homogeneity of your target population, the depth of your interviews, and the complexity of the phenomenon. Empirical research on saturation suggests thematic saturation is typically reached with 9–17 interviews in moderately focused studies, while studies with more heterogeneous samples or multiple sub-questions may require 25–40 interviews. For IPA studies using homogeneous sampling, 6–12 participants is standard. The most defensible approach is to recruit until saturation, with a minimum that you justify in advance.

Is purposive sampling the same as quota sampling?

Not exactly. Quota sampling is used in quantitative and survey research to ensure that specific demographic proportions are represented (e.g., 50% women, 30% under 35). Purposive sampling is a broader concept in qualitative research where the selection criteria are driven by conceptual relevance rather than demographic quotas. That said, maximum variation purposive sampling shares some structural similarity with quota sampling in that you actively seek variation across predefined dimensions.

How do I justify purposive sampling in a PhD thesis?

State your sampling type by name, cite a methodological authority (Patton 2002 is the most widely accepted reference), explain why your specific type fits your research question and design, list your inclusion and exclusion criteria with rationale, describe your recruitment process, and link your sampling strategy to your saturation criteria. A short paragraph in your methods chapter covering these five elements will address most examiner queries. If you used IPA or grounded theory, cite the method-specific sampling guidance in addition to Patton.

Want to analyse your qualitative data with the same rigour you brought to your sampling? Try Skimle for free and move from transcripts to transparent, auditable themes in a fraction of the time. Every insight traces back to the source quote, so your analysis is as defensible as your sample design.

If you are an academic researcher, Skimle is designed to meet the transparency and traceability standards that journals and PhD examiners expect.

Related reading:

About the authors

Henri Schildt is a Professor of Strategy at Aalto University School of Business and co-founder of Skimle. He has published over a dozen peer-reviewed articles using qualitative methods, including work in Academy of Management Journal, Organisation Science, and Strategic Management Journal. His research focuses on organisational strategy, innovation, and qualitative methodology. Google Scholar profile

Olli Salo is a former Partner at McKinsey & Company where he spent 18 years helping clients understand the markets and themselves, develop winning strategies and improve their operating models. He has done over 1000 client interviews and published over 10 articles on McKinsey.com and beyond. LinkedIn profile