What is Non-Probability Sampling? Methods, Types, and Examples

Non-probability sampling selects participants through judgment, convenience, or referral rather than random processes.
It is best suited to qualitative, exploratory, and pilot research where generalizability is not the primary goal.
The six main types are: convenience, purposive, snowball, quota, theoretical, and self-selection sampling.
Purposive sampling is the most commonly searched term; it is identical to judgmental or selective sampling.
Sample size in non-probability studies is typically guided by data saturation rather than statistical formulas.
Bias can be mitigated through triangulation, transparent reporting, member checking, and purposeful diversification of participants.
To defend a non-probability sample to journal reviewers, researchers should acknowledge limitations openly, describe selection criteria precisely, and explain why random sampling was not feasible.
Non-probability sampling can be used in quantitative research, though it limits statistical inference and requires explicit acknowledgment in the methods section.

Table of Contents

What Is Non-Probability Sampling?

Non-probability sampling is a method of selecting research participants through subjective judgment, convenience, or referral rather than random processes. Unlike probability sampling, it does not guarantee that every member of the target population has an equal or known chance of inclusion. As a result, findings from non-probability samples cannot be statistically generalized to the broader population in the same way that findings from probability samples can.

This does not make the method inferior for all purposes. Non-probability sampling is particularly well suited to qualitative research, exploratory studies, and situations where the target population is hard to reach or a sampling frame does not exist. Its main advantages are speed, low cost, and practical accessibility.

When Should You Use Non-Probability Sampling?

Non-probability sampling is appropriate when random selection is not feasible, necessary, or aligned with the research design. Use it in the following situations:

Limited resources: When time, budget, or staffing prevents constructing or accessing a full sampling frame.
Exploratory or pilot research: When the goal is to generate hypotheses or test procedures rather than to make population-level inferences.
Qualitative research: When the aim is in-depth understanding of a specific group, context, or experience rather than statistical description.
Hard-to-reach populations: When the target group is hidden, stigmatized, or geographically dispersed, making random selection impractical.
Absence of a sampling frame: When no complete list of the population exists, making random selection impossible.
No generalizability requirement: When the research is a case study, an ethnography, or another design where results are explicitly bounded to the sample.
Urgent timelines: When findings are needed quickly and the depth of insight outweighs the need for statistical representativeness.

Types of Non-Probability Sampling

The six types below cover the full range used in contemporary research. Each is defined, characterized, and illustrated with discipline-specific examples.

Convenience Sampling

Convenience sampling, also called accidental or opportunity sampling, selects participants based on ease of access rather than any characteristic of the participant. It is the fastest and least expensive non-probability method, but it carries the highest risk of sampling bias because the sample is determined by what is available rather than what is representative.

Purposive Sampling

Purposive sampling, also called judgmental or selective sampling, involves the researcher deliberately choosing participants who possess characteristics relevant to the research question. It is the most widely used non-probability technique in qualitative research and is appropriate whenever the researcher needs specific expertise, experiences, or perspectives in the sample.

Purposive sampling has several recognized subtypes:

Snowball Sampling

Snowball sampling begins with a small number of participants who then refer the researcher to others in their networks, growing the sample incrementally. It is the method of choice when the population is hidden, stigmatized, or otherwise hard to reach through conventional recruitment.

Quota Sampling

Quota sampling sets predetermined targets for participant characteristics, such as equal numbers of men and women, or specific age brackets, and then fills those quotas through non-random selection. It is the non-probability method that most closely approximates stratified random sampling in structure, though it does not use randomization within strata.

Theoretical Sampling

Theoretical sampling is used specifically in grounded theory research. Data collection and analysis proceed simultaneously, with each round of sampling guided by emerging concepts from the previous round. The researcher continues sampling until theoretical saturation is reached: the point at which new data no longer modify the developing theory.

Self-Selection Sampling

Self-selection sampling occurs when individuals opt into a study of their own accord, typically in response to a public call for participants. It is distinct from volunteer sampling in that the researcher’s direct involvement in participant choice is minimal: participants choose to participate rather than being approached. This method is convenient for online surveys and open recruitment but introduces strong participation bias, because motivated or opinionated individuals are more likely to respond.

How Large Should a Non-Probability Sample Be?

There is no universal formula for non-probability sample size. The appropriate size depends on the method used, the research design, and the complexity of the phenomenon under investigation.

Method	Size Guidance	Rationale
Purposive sampling	6 to 50 participants, depending on design	Qualitative interviews: 6 to 15 for homogeneous samples; up to 50 for maximum variation designs. Stop when saturation is reached.
Snowball sampling	Until saturation	No fixed target. Sample grows until new referrals no longer produce new themes or perspectives.
Convenience sampling (qualitative)	Until saturation	Same principle as snowball: continue until data are repetitive.
Convenience sampling (quantitative)	Minimum 30 for descriptive analysis; 100+ for regression	Follows general statistical guidance for exploratory quantitative work, with the caveat that findings are not generalizable.
Quota sampling	Defined by quota targets	Set by the number and size of categories. Each cell typically needs at least 30 respondents for descriptive analysis.
Theoretical sampling	Until theoretical saturation	The researcher stops when new data no longer modify the developing theory. This can require as few as 20 or as many as 60+ interviews.
Self-selection	As many as possible	Because of participation bias, larger samples allow for sub-group analysis and better demographic profiling of respondents.

The concept of data saturation is the most widely accepted stopping criterion for qualitative non-probability samples. Saturation is typically reached between 12 and 20 interviews in homogeneous purposive samples, though more complex studies may require more. Researchers should document their saturation judgment explicitly in the methods section.

Probability vs. Non-Probability Sampling: What Is the Difference?

Characteristic	Probability Sampling	Non-Probability Sampling
Selection process	Random: every population member has a known, non-zero chance of selection	Non-random: participants chosen by judgment, convenience, or referral
Types	Simple random, stratified, cluster, systematic	Convenience, purposive, snowball, quota, theoretical, self-selection
Generalizability	Supports statistical generalization to the target population	Does not support formal generalization; findings are bounded to the sample
Sampling error	Can be calculated and reported as a margin of error	Cannot be formally calculated; bias is acknowledged qualitatively
Sampling frame required	Yes: a complete list of the population is needed	No: a sampling frame is not required
Cost and time	Higher: random selection requires infrastructure and often larger samples	Lower: faster and less resource-intensive
Best suited to	Quantitative, confirmatory research; surveys intended to represent a defined population	Qualitative, exploratory, or pilot research; studies of hard-to-reach populations
Best used when	Results must be generalizable; adequate resources and a sampling frame exist	Generalizability is not required; resources are limited; the population is inaccessible

Advantages and Disadvantages of Non-Probability Sampling

Advantages

Advantage	Explanation
Cost-effectiveness	Requires fewer resources than probability sampling because no sampling frame or randomization infrastructure is needed.
Speed	Samples can be assembled quickly, making this approach ideal for pilot studies, urgent research questions, or iterative qualitative work.
Access to hidden populations	Snowball and purposive methods can reach groups that are inaccessible through any form of random selection, such as undocumented migrants, illicit drug users, or people with stigmatized conditions.
Flexibility	The researcher can adapt selection criteria as the study progresses, which is especially valuable in theoretical sampling and grounded theory research.
Targeted expertise	Purposive methods ensure that participants have direct, relevant experience of the phenomenon under study, increasing the depth and relevance of data.
Practical feasibility	In many real-world research contexts, probability sampling is simply not possible. Non-probability methods make research feasible where it would otherwise be impossible.

Disadvantages

Disadvantage	Explanation
No formal generalizability	Because participants are not randomly selected, it is not statistically legitimate to generalize findings to the broader population.
Sampling bias	The selection process introduces systematic distortions: convenience samples overrepresent accessible individuals; snowball samples overrepresent densely connected network members.
Inability to calculate sampling error	Unlike probability samples, there is no formula for estimating the margin of error or confidence intervals.
Reduced perceived credibility	Peer reviewers and grant committees may view non-probability sampling as less rigorous, requiring the researcher to provide a stronger methodological justification.
Researcher subjectivity	Purposive and judgmental methods rely heavily on the researcher’s judgment, which can introduce unconscious bias in participant selection.
Limited inferential statistics	Many parametric statistical tests assume random sampling. Applying them to non-probability samples without explicit caveats is methodologically problematic.

How Can Researchers Reduce Bias in Non-Probability Samples?

Bias cannot be eliminated from non-probability sampling, but it can be meaningfully reduced and transparently managed. The following strategies are widely recommended:

Triangulation: Use multiple data sources, methods, or analysts to cross-check findings. If interviews, observations, and documents all point to the same conclusion, confidence in the finding increases even without a representative sample.
Member checking: Share preliminary findings with participants to verify that their perspectives have been accurately captured.
Maximum variation selection: Deliberately seek participants who differ on key variables to ensure the sample does not reflect only one segment of the target population.
Reflexive journaling: Document selection decisions and the researcher’s own assumptions throughout the study so that readers can assess how subjectivity may have shaped the sample.
Transparent reporting: Describe the selection process in detail in the methods section, including who was excluded and why.
Combining methods: Use non-probability sampling for qualitative phases and probability sampling for confirmatory quantitative phases in a mixed-methods design.
Demographic benchmarking: For quantitative non-probability samples, compare respondent demographics to known population benchmarks and discuss any discrepancies.

Frequently Asked Questions

Does non-probability sampling affect research outcomes?

Yes, and the effect depends on the method and the research goal. Because participants are not selected randomly, non-probability samples may overrepresent certain groups, exclude others, and introduce systematic bias into the data. This does not make the findings invalid: it means they must be interpreted within the scope of the sample rather than generalized to the broader population. Transparent reporting of selection criteria and acknowledged limitations protects the credibility of the findings.

Why would a researcher choose non-probability over probability sampling?

The most common reasons are practical: probability sampling requires a complete sampling frame (a list of all population members), adequate time, and sufficient budget. Many real-world research contexts lack one or more of these. Beyond practicality, some research questions genuinely call for non-probability methods: studying hidden populations, conducting grounded theory research, or running exploratory pilot work are all contexts where random selection is either impossible or theoretically inappropriate.

Can non-probability sampling be used in quantitative research?

Yes, though with important caveats. Non-probability sampling is commonly used in quantitative studies when probability sampling is not feasible, such as online surveys distributed through social media or surveys of professional networks. The key requirement is transparency: the methods section must clearly state that the sample is not random, and the discussion must acknowledge that inferential statistics should be interpreted cautiously. Findings from such studies are best treated as indicative rather than definitive.

My thesis supervisor says my convenience sample is a limitation. How do I respond?

This is a very common concern among student researchers. The appropriate response is to acknowledge the limitation clearly rather than defend it away. In your methods section, explain why convenience sampling was necessary given your resources and timeline. In your limitations section, identify specifically which groups may be underrepresented and how this could affect the findings. Describe any steps you took to mitigate bias: for example, sampling across different times or locations, or using member checking. Then frame the contribution appropriately: if the study is exploratory or qualitative, the standard is depth of insight rather than statistical representativeness.

Can I use both purposive and snowball sampling in the same study?

Yes, and this combination is common in practice. A researcher might purposively select the first several participants based on explicit inclusion criteria and then use those participants as the starting point for snowball referrals to expand the sample. This approach is particularly effective when the target population is partially accessible: the researcher can identify and recruit initial participants directly, then use the snowball method to reach individuals who would not have been accessible through direct recruitment alone. Document the rationale for each phase of recruitment separately in the methods section.

Will journals reject my paper if I used non-probability sampling?

Not automatically. Non-probability sampling is standard practice in qualitative research and is widely accepted in exploratory and mixed-methods work across most disciplines. The risk of rejection arises when a paper makes strong statistical generalizations based on a non-probability sample without acknowledging the limitation, or when the sampling method is poorly described. Reviewers expect researchers to name the method, justify the choice, describe the selection criteria, state the sample size and rationale, and acknowledge the implications for generalizability. A clearly reasoned methods section with an honest limitations discussion will satisfy most peer reviewers.

References

Levy, P. S., and Lemeshow, S. (2013). Sampling of Populations: Methods and Applications. Wiley.
Pandey, P., and Pandey, M. M. (2021). Research Methodology Tools and Techniques. Bridge Center.
Patton, M. Q. (2002). Qualitative Research and Evaluation Methods (3rd ed.). Sage Publications.
Creswell, J. W., and Plano Clark, V. L. (2017). Designing and Conducting Mixed Methods Research (3rd ed.). Sage Publications.
Etikan, I., Musa, S. A., and Alkassim, R. S. (2016). Comparison of convenience sampling and purposive sampling. American Journal of Theoretical and Applied Statistics, 5(1), 1-4.
Baltes, S., and Ralph, P. (2022). Sampling in software engineering research: A critical review and guidelines. Empirical Software Engineering, 27(94).

This article was published on November 28, 2024, and updated on June 26, 2026.