Convenience, Purposive, Quota, and Snowball Sampling: When Is Non-Probability Sampling Defensible?

If You Cannot Randomly Sample the Population, What Should You Do Instead?

Researchers do not always have a complete population list or the resources and access needed for probability sampling. Sometimes the people most relevant to a question are rare, difficult to identify, or connected through networks. In qualitative research, you may deliberately want participants with particular experiences rather than a statistical cross-section of a population.

These situations often lead to non-probability sampling. But that label covers very different approaches.

Recruiting whoever is easiest to reach is not methodologically equivalent to deliberately selecting information-rich cases. Filling demographic quotas is not the same as asking participants to refer eligible peers. The relevant question is therefore not merely whether non-probability sampling is acceptable. It is which non-probability method fits your research purpose, why it is defensible, and what conclusions the resulting sample can support.

Non-Probability Sampling Is Defensible When Its Logic Fits the Research Purpose

Four Non-Probability Methods With Four Different Sampling Logics

What Makes a Sample Non-Probability?

In a non-probability sample, units are not selected through a probability mechanism that provides known selection probabilities for the relevant population under the design.

That distinction has inferential consequences. In probability sampling, known selection probabilities provide a formal design-based basis for estimating sampling variability. With non-probability sampling, that basis is absent, and population inference requires other assumptions or analytical approaches when it is attempted.

This does not mean non-probability sampling is one inferior method. It is a family of approaches used for different purposes. CDC guidance on sampling, for example, identifies convenience, purposive, quota, and chain-referral methods among common forms of non-probability sampling and notes that such approaches can be useful when probability sampling is impossible, impractical, or not central to the research purpose.

What Is Convenience Sampling?

Convenience sampling recruits participants primarily because they are readily available or easy to access.

Examples include surveying students in classes you already teach, recruiting patients who attend one accessible clinic during the data-collection period, distributing an open survey link through your own social network, or approaching people at an easily accessible location.

Convenience is attractive because it can reduce time, cost, and logistical complexity. CDC guidance notes that convenience sampling is fast and less resource intensive than more demanding approaches and can be useful particularly in exploratory work or when populations are very small.

The central limitation is selection. People who are easiest to reach may differ systematically from those who are difficult to reach. Their probability of entering the sample is unknown, and some eligible members of the target population may have little or no opportunity to participate.

This does not make every convenience sample useless. It does mean that the research claims should reflect how the sample was obtained. The more the research objective depends on estimating a broader population, the more consequential this limitation becomes.

What Is Purposive Sampling?

Purposive sampling, also called purposeful or judgmental sampling in some methodological traditions, deliberately selects people or cases because they possess characteristics, experiences, knowledge, or other qualities relevant to the research question.

Suppose you want to understand how university instructors redesign assessment after encountering student misuse of generative AI. Randomly selecting instructors might yield many people who have never encountered the phenomenon. A purposive strategy could instead identify instructors with direct relevant experience.

The defining logic is not ease of access but relevance to the inquiry.

Purposive sampling is especially common in qualitative research, where researchers may seek information-rich cases, variation across experiences, particular types of participants, or cases capable of illuminating a phenomenon in depth.

Purposive sampling still requires methodological discipline. Researchers should specify what characteristics guided selection, why those characteristics matter, how potential participants were identified, and how the evolving sample served the analytical purpose.

What Is Quota Sampling?

Quota sampling establishes target numbers for particular categories and recruits participants non-randomly until those quotas are reached.

Suppose a researcher wants 400 survey respondents and decides that 200 should be undergraduate students and 200 postgraduate students. Recruitment continues within each category until the required numbers are obtained.

More elaborate quotas might involve age, sex, geographic location, occupation, or combinations of characteristics.

Quota sampling can prevent an uncontrolled convenience sample from being dominated by one easily recruited subgroup. However, satisfying quotas does not automatically make the sample representative. Two samples can have identical demographic proportions while differing substantially in unmeasured characteristics because participants within each quota were still recruited non-randomly.

What Is Snowball Sampling?

Snowball sampling is a form of chain-referral recruitment in which initial participants or contacts help identify additional potentially eligible participants, who may then lead researchers to others.

This approach can be particularly useful when the population is difficult to enumerate or reach through conventional lists. People with uncommon experiences, members of socially connected populations, or groups without an adequate sampling frame may be more readily located through existing relationships.

The strength of the method is access. Its weakness is also the network structure through which that access occurs.

Participants tend to know people connected to their own social worlds. Individuals with large networks may be easier to reach, isolated individuals may never enter the referral chains, and recruitment may remain concentrated within a few interconnected groups.

For that reason, an ordinary snowball sample should not be treated as though each member of the broader population had an equal or known probability of selection.

Is Snowball Sampling the Same as Respondent-Driven Sampling?

No. The terms are sometimes used loosely, but they should not be treated as interchangeable.

Both use peer recruitment, but respondent-driven sampling was developed as a more structured network-based method involving controlled recruitment procedures and statistical estimation based on additional information and assumptions. Whether those assumptions are adequately satisfied is itself an important methodological question.

Calling an ordinary referral chain “respondent-driven sampling” merely because participants recruited acquaintances would therefore misdescribe the method.

How Do the Four Methods Compare?

Purposive Sampling Is Not Just a More Respectable Name for Convenience Sampling

This distinction is worth making explicitly because the terms are sometimes blurred in research reports.

If you recruit faculty members because they happen to work in your department and are easy to contact, the sample is primarily convenient. Describing them afterward as “purposively selected because they were faculty members” does not change the original selection mechanism if virtually every accessible faculty member qualified.

In purposive sampling, the characteristics guiding selection should follow from the analytical purpose. Researchers deliberately seek particular cases because those cases can contribute evidence needed for the inquiry.

When Is Non-Probability Sampling Defensible?

A defensible method is one whose selection logic makes sense for the research objective and whose limitations are acknowledged rather than concealed.

Non-probability sampling may be appropriate when the study is qualitative and seeks information-rich cases, when no suitable probability frame exists, when the population is difficult to locate, when the work is exploratory or feasibility-oriented, or when genuine resource and access constraints make probability sampling impractical.

CDC guidance similarly identifies circumstances such as unreachable target populations, limited resources, time pressure, and qualitative research as situations in which non-probability methods may be considered, while emphasizing their vulnerability to sampling bias.

The justification should still be specific. “Non-probability sampling was used because it was easier” is methodologically weaker than explaining why the particular recruitment strategy was appropriate to the study's purpose and what limitations followed from it.

When Does Non-Probability Sampling Become Harder to Defend?

The method becomes more problematic when the research claims require stronger population inference than the sampling and analysis can support.

For example, a voluntary online survey distributed through one researcher's professional network may be useful for exploratory evidence. Presenting the resulting percentages as precise estimates for all researchers in a country would require a much stronger inferential justification.

This distinction is central to deciding whether a convenience sample can still be useful. Practical sampling is not automatically fatal to a study, but practical access should not silently become evidence of population representativeness.

A Large Non-Probability Sample Can Still Be Selective

It is tempting to assume that selection problems disappear once enough people respond.

They do not necessarily disappear. If the process that brings people into the sample systematically favors particular kinds of participants, collecting more people through the same process may reproduce that selectivity on a larger scale.

This is why sample size and sampling bias should be treated separately. The question of whether a large sample can still be seriously biased has an unequivocal answer: yes.

Can Statistical Adjustment Make a Non-Probability Sample Representative?

Sometimes researchers use weighting, calibration, propensity modeling, matching, data integration, or related methods to improve inference from non-probability samples. Such methods can be valuable, particularly when strong auxiliary population information is available.

But adjustment is not alchemy. Its success depends on what variables are measured, how selection relates to those variables and the outcomes, model assumptions, the quality of external benchmarks, and the adequacy of the analytical method.

A sample that matches population age and sex distributions after weighting, for example, may still differ on unmeasured characteristics associated with the outcome.

Statistical adjustment should therefore be reported as an analytical strategy with assumptions and limitations, not as proof that the original sample was probability-selected.

Transparent Reporting Matters More, Not Less, With Non-Probability Samples

Readers should be able to understand how participants actually entered the study.

AAPOR's transparency standards call for researchers using non-probability sources to describe the lists, panels, websites, social media, respondent networks, or other recruitment sources used, together with participant contact and selection procedures, eligibility requirements, quotas where applicable, and related methodological information.

That level of transparency is far more informative than a single sentence stating that “non-probability sampling was employed.”

Four Different Samples for Four Different Research Needs

The label “non-probability sample” therefore provides only the first layer of information. The actual method and its relationship to the research question are what allow readers to evaluate the design.

Common Mistakes About Non-Probability Sampling

Name the Sampling Problem Before Choosing the Method

The strongest justification is usually specific: why this method, for this population, for this question, under these conditions? Once you can answer that, the sampling method becomes part of the research logic rather than a label added to the methodology after recruitment.

Before You Use a Non-Probability Sample

Questions About Non-Probability Sampling Methods

A Non-Probability Sample Needs a Reason, Not an Apology

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.