Sampling for M&E: a plain-English guide
Oct 12 2026 · Method
How many people to survey, and how to choose them — without a statistics degree.
Sampling in M&E is about choosing a subset of people or units to represent a larger group, so you can draw conclusions without surveying everyone. Two decisions matter: how many and who. For a rough estimate of a proportion in a large population, a sample of around 384 gives a margin of error of about ±5% at 95% confidence; smaller populations need fewer. For choosing who, random sampling is strongest, but cluster sampling (villages, schools) is often the only practical option in the field. Always report the design, sample size and limitations, because an unexplained sample invites doubt about every number built on it.
What is sampling in M&E?
Sampling is selecting a subset of a population to estimate characteristics of the whole. In M&E, it helps you get credible answers without interviewing everyone. The goal is a sample that is representative enough for your purpose, whether that is a baseline, midline or endline survey.
How many people do I need to survey?
Sample size depends on your population size, the confidence level you want, and the margin of error you can accept. For a large population, a common rule of thumb is about 384 respondents for a ±5% margin at 95% confidence. For smaller populations, you need proportionally fewer. Use a sample size calculator or table to find the right number.
| Population | Sample size for ±5% at 95% confidence |
|---|---|
| Large (10,000+) | ≈384 |
| 1,000 | ≈278 |
| 300 | ≈169 |
| 100 | ≈80 |
How do I choose who to survey?
The method you use to select respondents matters as much as the number. Random sampling gives every unit a known chance of selection, which reduces bias. But in practice, you often cannot list everyone. Cluster sampling — selecting villages, schools or health facilities first, then sampling within them — is common when a full list is unavailable. It is less precise but often the only feasible option.
What is the difference between random and cluster sampling?
| Method | How it works | Pros | Cons | When to use |
|---|---|---|---|---|
| Simple random | Every individual has an equal chance of selection | Unbiased, straightforward to explain | Requires a complete list; can be impractical | When you have a full sampling frame |
| Systematic | Pick every nth person from a list | Easy to implement if a list exists | Can be biased if the list has patterns | When a list is available and ordered randomly |
| Stratified | Divide into groups (strata) and sample within each | Ensures representation of key subgroups | Needs information on strata before sampling | When you want to compare subgroups |
| Cluster | Select groups (clusters) then sample within them | Practical when no full list exists | Less precise; needs a larger sample | When a full list is unavailable or travel is costly |
| Purposive | Deliberately select information-rich cases | Useful for qualitative or hard-to-reach groups | Not statistically representative | When you need specific perspectives, not numbers |
When is purposive sampling appropriate?
Purposive sampling is for when you need specific perspectives, not statistical representation. Use it for qualitative interviews, case studies, or to reach hidden populations. It is common in process evaluations and when you want to understand why something happened, not how many.
What should I report about my sample?
Always report the sampling design, the sample size, how you selected respondents, and any limitations. If you used cluster sampling, say so. If response rates were low, explain. An unexplained sample is the first thing an evaluator questions. Transparency builds credibility.
An unexplained sample invites doubt about every number built on it.
Checklist: designing a sample for M&E
- Define your population clearly.
- Decide on confidence level and margin of error.
- Calculate or look up the required sample size.
- Choose a sampling method that fits your context and resources.
- Prepare a sampling frame (list) if possible.
- Train enumerators on selection and replacement rules.
- Document any deviations from the plan.
- Report the design and limitations in your findings.
Common mistakes to avoid in M&E sampling
- Using convenience sampling and calling it random.
- Ignoring non-response bias.
- Sampling only easy-to-reach areas.
- Failing to account for cluster design in analysis.
- Not reporting the sample size or method.
Where can I get more help with M&E sampling?
If you need templates, tools or guidance, there are resources available. For a free handbook on M&E basics, see /book. For help choosing the right approach, try /which-ai. For logframe templates, see /blog/logframe-template-google-sheets. For practical tools, visit /tools.
Written for M&E professionals. Not a guarantee of accuracy — always verify every claim against its source, and let a named human sign off. Get the free handbook.