Quick answer
Stratified sampling is a sampling method that divides a population into non-overlapping groups (strata) based on a shared trait, then draws a random sample from each group, so every group is represented in the sample.
Formula: Proportionate: nₕ = n × Nₕ / N; weight wₕ = Nₕ / nₕ
What is stratified sampling?
Stratified sampling divides a population into groups that do not overlap, called strata, based on a trait such as plan, country or device. You then draw a random sample from each group. Because every group is sampled on purpose, a small group cannot be left out by chance, and estimates are usually more precise than a simple random sample of the same size.
The four steps
- Pick the trait that splits the population into groups that differ on what you are measuring.
- Assign everyone to exactly one stratum. The strata must cover the whole population with no overlap.
- Decide how many to sample in each stratum (allocation, below).
- Draw a simple random sample within each stratum and combine the results, using weights if needed.
Proportionate vs disproportionate allocation
A team wants 400 customer interviews from 10,000 users (sample numbers):
| Plan | Users | Proportionate sample | Equal sample | Weight (equal) |
|---|---|---|---|---|
| Free | 7,000 | 280 | 133 | 52.6 |
| Pro | 2,500 | 100 | 133 | 18.8 |
| Business | 500 | 20 | 134 | 3.7 |
| Total | 10,000 | 400 | 400 |
- Proportionate: each stratum gets the same share it has in the population (Free is 70%, so 280 of 400). Results combine directly.
- Disproportionate (here, equal): small strata get more interviews than their share, so you can say something reliable about Business customers. To estimate an overall figure, weight each answer by stratum size ÷ stratum sample size, for example 7,000 ÷ 133 = 52.6 for Free.
A simple random sample of 400 would include about 20 Business users on average, and sometimes far fewer.
Stratified vs other methods
| Method | How it picks | Best when |
|---|---|---|
| Simple random | Anyone, with equal chance | The population is fairly uniform |
| Stratified | Some from every group | Groups differ and some are small |
| Cluster | A few whole groups, everyone in them | Groups are similar and reaching everyone in a group is cheap |
| Systematic | Every k-th item in a list | You have an ordered list and no repeating pattern in it |
Why it matters in analytics
Stratifying is not only for surveys. When you read a metric by stratum (conversion by device, retention by plan), you avoid an average that hides opposite trends in different groups and you can spot a confounding variable. It also protects against samples made of only active or loud users, a form of survivorship bias.
Drawing a stratified sample in TapCub
In Segments, create one rule segment per stratum with a property condition (for example plan = Business). Rule segments also support a random bucket condition, so you can combine "plan = Pro" with a random bucket to get a random subset of that plan. Open each segment's members with View users to get the list for outreach. Note that segments only contain identified people; anonymous visitors have no profile.
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.
Product Experience Specialist
Writes about product experience, practical workflows, and growth, turning complex processes into clear steps for everyday decisions.