Imagine you want to study freelance graphic designers who work entirely online, or people who have left a particular profession, or members of a small community that has no public list anywhere. You cannot draw names from a list, because no list exists. So how do you find your participants?
One answer is snowball sampling: you start with a few people you can find, and each one leads you to the next. Like a snowball rolling downhill, your sample grows as it goes. In this guide, you will learn exactly what snowball sampling is, how it works step by step, when it is the right choice, and where it can go wrong.
What Is Snowball Sampling?
Snowball sampling is a non-probability sampling method in which existing participants recruit future participants from among their own contacts. The researcher begins with a small number of initial participants, called seeds, and each seed refers other people who meet the study criteria. Those new participants then refer others, and the sample grows in waves until the researcher has enough participants or reaches data saturation.
It is called “snowball” sampling because the sample gets bigger as it rolls along, just like a snowball gathering more snow. It is also sometimes called chain-referral sampling or respondent-driven sampling, although respondent-driven sampling is a more formal variant with its own statistical adjustments, and the two terms should not be treated as identical.
The key idea is simple: when you cannot reach a population directly, you reach it through its own social networks.
Why Researchers Use Snowball Sampling
The main reason is access. Some populations are hidden, hard to reach, or simply have no sampling frame — no complete list from which to draw a random sample. Examples include:
- People with a rare condition or experience
- Members of stigmatized or underground communities
- Migrants or refugees who are not on official registers
- Professionals in a tiny, specialized niche
- People who have left a job, a group, or a religion
In these cases, random sampling is not just difficult — it is impossible, because there is no list to sample from. Snowball sampling solves the access problem by using trust: a referral from someone the participant already knows carries far more weight than a cold message from a stranger.
Snowball sampling is also practical. It is usually cheaper and faster than trying to build a sampling frame from scratch, and it works well for exploratory and qualitative research, where the goal is depth of understanding rather than statistical generalization to a whole population.
How Snowball Sampling Works: Step by Step
Step 1: Define your inclusion criteria clearly. Before you recruit anyone, write down exactly who counts as a participant. For example: “adults who have worked as freelance translators for at least two years.” Vague criteria produce a messy sample, so be precise.
Step 2: Find your seeds. Identify a small number of initial participants — often between three and ten — who meet your criteria and are willing to take part. You might find them through a community organization, a social media group, a professional contact, or a gatekeeper who is trusted by the community.
Step 3: Interview or survey the seeds. Collect your data from the initial participants just as you would with any other sample.
Step 4: Ask each participant for referrals. At the end of each interview or survey, ask the participant to recommend other people who meet your criteria. Explain what participation involves so they can pass on accurate information. Some researchers offer a small incentive for each successful referral, but this must be handled carefully and ethically — more on that below.
Step 5: Contact the referrals and repeat. Reach out to the referred people, screen them against your criteria, collect data, and ask them for further referrals. Each round is sometimes called a “wave.”
Step 6: Stop when you have enough. In qualitative research, you typically stop at data saturation — the point where new interviews stop revealing new themes. In a small survey, you stop when you reach your target sample size. Either way, decide your stopping rule in advance and record it.
A Worked Example
Suppose a student wants to study the work-life balance of home-based women entrepreneurs in a small city. There is no register of such entrepreneurs, and many do not advertise their businesses publicly.
She starts with two seeds: a cousin who sells homemade food online, and a neighbour who runs a tailoring business from home. She interviews both. Each one recommends two or three other home-based entrepreneurs she knows. The student contacts them, interviews those who agree and meet her criteria, and asks each for further names. After four waves, she has 24 participants, including women from different neighbourhoods and business types that her first two contacts alone would never have reached.
Notice what made this work: the referrals carried trust. A recommendation from a fellow entrepreneur opened doors that a flyer or a random message would not have.
Snowball Sampling vs Other Non-Probability Methods
Snowball sampling is often confused with convenience sampling and purposive sampling. Here is how they differ:
| Method | How participants are found | Best when |
|---|---|---|
| Snowball sampling | Participants refer other participants through their networks | The population is hidden or has no list |
| Convenience sampling | Whoever is easiest to reach | You need quick, exploratory data and generalization is not the goal |
| Purposive sampling | The researcher hand-picks people with specific traits | You need information-rich cases for a specific purpose |
| Quota sampling | Participants are recruited to fill set quotas (for example, equal numbers of men and women) | You want a sample with a certain composition but cannot sample randomly |
The difference matters for your methodology chapter: if participants found each other, say snowball sampling. If you chose them yourself for their expertise, that is purposive sampling, even if you found them through contacts.
Strengths of Snowball Sampling
First, it reaches the unreachable. Populations with no sampling frame become researchable. Second, it builds on trust, which often improves response rates and the honesty of answers, especially on sensitive topics. Third, it is cost-effective: referrals do most of the recruitment work for you. Fourth, it can reveal the social structure of a community — who knows whom — which is itself useful data in network and ethnographic studies.
Limitations and How to Handle Them
Snowball sampling has real weaknesses, and examiners will expect you to name them.
- Sampling bias. People tend to refer people like themselves, so your sample can become too similar — same age group, same neighbourhood, same views. To reduce this, start with several diverse seeds from different networks rather than one tight group, and keep recruiting until new waves add genuinely different voices.
- No statistical generalization. Because participants are not chosen randomly, you cannot calculate a margin of error or claim your findings represent the whole population. Report your findings as insights about your participants, not as population estimates.
- Lack of control over who joins. Referrals can bring in people who do not quite meet your criteria. Screen every referral carefully, and be ready to exclude people politely.
- Privacy and ethical concerns. A participant who refers a friend has shared that friend’s contact details with you. Get permission before passing on details, protect identities in your reporting, and never pressure anyone to name names. If you use referral incentives, keep them modest so nobody feels pushed to recruit.
- Over-reliance on a few seeds. If one seed’s network dominates, the whole sample tilts in that direction. Track which seed each participant came from, and if one chain is dominating, start new seeds elsewhere.
When Should You Use Snowball Sampling?
Use snowball sampling when your population is hidden, hard to reach, or has no sampling frame, and when your study is exploratory or qualitative. It fits interview studies, case studies, and ethnographic work particularly well.
Do not use it when you need a representative sample, when you must generalize percentages to a larger population, or when a complete list of the population is available — in that case, a probability method such as simple random or stratified sampling is the stronger, more defensible choice.
How to Report Snowball Sampling in Your Thesis
Examiners look for transparency. In your methodology chapter, state: your inclusion criteria, how many seeds you started with and how you found them, how many waves of referral took place, your final sample size, your stopping rule (for example, data saturation), and the limitations you recognized. A short diagram showing the waves of recruitment makes the process easy to follow and shows you understood the method rather than just naming it.
Common Mistakes to Avoid
- Starting with a single seed and letting one network dominate the sample.
- Calling it snowball sampling when you actually hand-picked participants yourself (that is purposive sampling).
- Forgetting to screen referrals against your criteria.
- Claiming your results generalize to the whole population.
- Failing to protect the privacy of people who were referred by others.
- Writing no stopping rule, so the sample size looks arbitrary.
Frequently Asked Questions
Is snowball sampling a probability or non-probability method? It is a non-probability method, because participants are not selected randomly and not every member of the population has a known chance of being included.
How many seeds should I start with? There is no fixed rule, but starting with several diverse seeds — often three to ten — is better than relying on one. Diverse seeds reduce the risk of a sample that all comes from one social circle.
How do I know when to stop snowball sampling? In qualitative research, stop at data saturation, when new participants stop adding new themes. In small surveys, stop at a pre-decided target sample size. State your rule in your methodology chapter either way.
Can snowball sampling be used for quantitative research? Yes, but with caution. It is sometimes used for small surveys of hidden populations. The results describe your sample only and cannot be generalized statistically, so be honest about that limit in your write-up.
Is snowball sampling ethical? It can be, if handled properly: get informed consent from every participant, protect the privacy of referred people, do not pressure anyone to give names, and keep any referral incentives modest. Check your institution’s ethics requirements before you begin.
Key Takeaways
- Snowball sampling grows a sample through participant referrals, starting from a few seeds.
- It is the go-to method when a population is hidden or has no sampling frame.
- Start with several diverse seeds, screen every referral, and decide your stopping rule in advance.
- Its biggest weakness is bias toward similar participants — and its results cannot be generalized statistically.
- Report the process transparently: seeds, waves, sample size, stopping rule, and limitations.
If your participants are hard to find, let the people who know them help you find them — that, in one sentence, is snowball sampling.