Stratified Sampling

What is stratified sampling

Stratified sampling involves dividing the entire population into homogeneous groups called strata (plural for stratum). Random samples are then selected from each stratum. For example, if you had XS and XL customers and randomized them into two groups (Control and Test), you would want both Control and Test to be balanced across XS and XL customers. You can also stratify based on a metric such as Revenue/User.

With large numbers, randomization typically solves this balance problem. However, in B2B scenarios and other relatively low-volume or high-variance scenarios, stratified sampling ensures this balance. Statsig supports both automated and manual stratified sampling. In tests where a tail-end of power users drive a large portion of an overall metric value, stratified sampling meaningfully reduces false positive rates and makes results more consistent. In Statsig's simulations, this approach produced around a 50% decrease in the variance of reported results.

Automated stratified sampling

How it works

The selection space for the salts is large enough that stratifying multiple experiments on the same metric will not result in overlap. In Statsig's simulations, the groups were as independent as expected, which matched the literature on this topic.

Enabling stratified sampling

You can enable this on an experiment under Advanced Settings on the experiment setup page. There are two ways to stratify on Statsig. If you choose a metric to stratify on, Statsig uses that metric to balance the groups.

If you instead choose an attribute or a classification (for example, S, M, L, XL), Statsig uses that to balance the groups.

• On Statsig Cloud, you'll upload a CSV (in Beta)
• On Statsig Warehouse Native, you'll use Entity Properties

  

After you select the Stratify button, Statsig analyzes a set of salts and picks the best one.

FAQ and best practices

• What population is used when balancing?

  • When evaluating salts, Statsig computes balance using pre-experiment data for the entire targeted population of the experiment’s unit type (for example, all userIDs or all customerIDs) over the selected lookback window. There is no filtering on exposure because the experiment has not started yet.

• How are new units handled after stratification?

  • Units that were not present in the pre-experiment data are still assigned deterministically by the chosen salt, i.e., effectively at random with respect to the balancing metric. They do not influence the salt selection and may introduce some drift from the initial balance.

• Should I use stratified sampling for every experiment?

  • Not necessarily. Stratified sampling is most useful when you expect imbalance due to heterogeneous units (for example, “whales”) or skewed metrics. The tradeoff is time and compute cost that scales with the number of units and adds steps before starting an experiment. If you do not expect meaningful imbalance, a standard random split is generally recommended.

• Does salt evaluation assume 100% allocation? What about running at less than 100%?

  • Yes. All candidate salts are evaluated assuming 100% of the targeted population is allocated. If you run the experiment at an allocation below 100%, random sampling of that subset can reintroduce imbalance (for example, by chance, some high-impact units may fall disproportionately into one arm). For the period you care most about inference, prefer 100% allocation to preserve the intended balance. Lower allocations are best used briefly for safe rollouts rather than for the full experiment duration.

• Across candidate salts, is it the same set of users being evaluated?

  • Yes. Candidate salts are assessed over the same targeted population; only the randomization induced by the salt changes.

• How long does stratification take?

  • Duration depends on the number of units and the metric/source being queried. There is no fixed SLA; larger populations take longer.

Manual assignment for stratified sampling

When setting up an experiment, you can configure overrides (for example, force user X or Segment A into Control, force user Y or Segment B into Test). Overrides are intended for testing; overridden users are excluded from experimental analysis in Pulse results. To use manual assignment for stratified sampling, select the Include Overrides in Pulse checkbox. This includes manually overridden users in each variant in all metric lift analyses. You can configure 100% of experiment participants into your test variants manually, or configure some subset of participants into variants manually and randomly assign the remainder.

When you use the Statsig SDK for assignment, it handles randomization. When you control assignment of users, you're responsible for making sure users are balanced across experiment groups.

Related resources