Snowball Sampling: When Your Users Find Your Users
Snowball sampling has your first participants recruit the next ones from their network. It's vital for reaching hidden groups, like specific professional communities. The footgun is selection bias: you're sampling social networks, not the whole population.
WHY IT EXISTS How do you study a group you can't find? Some populations aren't on public lists or easily identifiable, making standard random sampling impossible. Snowball sampling was created to solve this access problem by leveraging trust and existing social connections to build a sample from the inside out.
THE MENTAL MODEL Think of it like finding the best local coffee shop in a new city. You don't have a master list. Instead, you ask one person you trust for a recommendation. Then you ask the barista they recommended who else makes great coffee. Your search "snowballs" through a network of connections, rather than starting from a complete, top-down directory.
HOW IT WORKS You start by identifying and recruiting one or a few initial participants who fit your criteria (the "seeds"). After interviewing or studying them, you ask them to refer you to other people they know who also meet the criteria. You then contact these referrals, and the process repeats. The sample size grows with each "wave" of recruitment until you have enough data. When this is done via online social networks, it's called virtual snowball sampling.
WHEN TO USE IT Use this technique when your target population is hard to reach, hidden, or lacks a clear sampling frame. This is common for studying specific subcultures, users of illicit services, members of exclusive communities (like C-level executives in a specific industry), or people with rare conditions. It's a qualitative, exploratory tool to understand a group's dynamics.
WHEN NOT TO USE IT Do not use snowball sampling if you need to make statistically valid generalizations about an entire population. The method is inherently biased. Because subjects are recruited through social networks, you're more likely to sample people who are well-connected and share characteristics with the initial seeds. This is not a representative sample and will skew quantitative findings.
ONE CANONICAL EXAMPLE A researcher wants to study the work habits of software developers who contribute to a niche, invite-only open-source project. Unable to get a list of contributors, they find and interview two known developers. They ask each for referrals to other contributors, who in turn provide more referrals. This allows the researcher to build a rich, qualitative picture of this hidden community, even though the findings can't be generalized to all open-source developers.
Read the original → en.wikipedia.org
Get five bites like this every day.
Tezvyn delivers a daily feed of 60-second tech bites with quizzes to lock in what you learn.