Random cohorts

How it works

branching_random-cohort-branch

If you use custom objects in your workspace, pay attention to the Cohort by setting. By default, we add people to random cohorts. But if you use custom objects in your workspace, you can group people into cohorts by a related object so people in the same company, account, class, or other object “group” travel down the same path. See Random cohorts with custom objects for more details.

In this video, we show an automation to convert people from free to paid plans, and we use a random cohort branch to randomly distribute people down different paths to test the best time to send an offer.

Set up a random cohort branch

When you drag a Random Cohort Branch into your workflow:

  1. Pick the Cohort by setting:
    • Profile: each person is added to random cohorts as an individual. If people can re-enter this automation, decide whether you want them to go down the same path each journey.
    • Objects, like Companies, group people into cohorts by a related object. This ensures that people related to the same object—in the same company, account, class, or other object—all travel down the same path. If someone belongs to multiple objects, they’ll exit the automation immediately. See Random cohorts with custom objects for more details.
  2. Add up to 20 paths and set the percentage of cohorts you want to send to each path. The total distribution must add up to 100% before you can save your changes.
A random cohort branch with two paths

Small sample sizes may not reflect your exact distribution settings

Because random cohort distribution is truly random, you may not see people distributed exactly as you expect until your automation experiences a statistically significant sample size. The larger the sample size, the easier it is to reach an exact distribution.

Send each person down the same path every time

This setting is available when you cohort by people. If you cohort by objects, you won’t see this setting; people related to the same object already go down the same path every time.

When you cohort by people, Customer.io picks a random path each time a person reaches the branch. If a person enters your automation more than once, they can go down a different path each time.

To keep people on the path for each journey, select Send people to the same path when they re-enter. Turn this setting on when a person’s path should last across journeys. For example, use it for a long-running holdout, or when you send each path different messages in an automation that people re-enter.

The setting applies to people who reach the branch after you save it. If a person re-enters the automation after you enable the setting, they go down a random branch, and then with subsequent journeys, they would follow the same path.

People may not repeat a path if you change the path percentages

If you change the percent of people that go down each path, then in subsequent journeys, some people will go down a different path in order to respect the new distribution. For example, if you change a 50/50 branch to 70/30, 20% of people enter a different path than before. Adding, removing, or reordering paths impacts people the same way.

Cohort by: random cohorts with custom objects

Random cohort blocks typically send people down different paths in your workflow. But imagine a scenario where you want to use a random cohort but you want to make sure that people related to the same account all travel down the same path—you don’t want people in the same account to have different experiences!

That’s what the Cohort by setting in a random cohort branch does: it treats custom objects as cohorts rather than people, so people related to the same object all travel down the same path.

A random cohort branch with the Cohort by setting open to show options

When someone reaches a random cohort block, Customer.io picks their path based on their related object, so everyone related to the same object goes down the same path. If you change the percentages or add or remove paths, some objects move to a different path, along with everyone related to them.

If someone belongs to multiple objects of the same type, we won’t know which cohort to assign them to, and they’ll exit the automation immediately! If people could have relationships to multiple objects of the same type, you should use the Cohort by: Profile option instead.

flowchart LR a(Person reaches a<br>random cohort block) a-->b(Customer.io picks a path<br>based on the person's<br>related object) b-->c(Person goes down<br>the same path as<br>other people related<br>to the same object)

Random cohorts in active automations

When you drag a Random Cohort into an active automation, one path is set to 100% and the other to 0%. While you can add more paths, you should leave the distribution at 100% for the first path and 0% for all other paths.

A random cohort branch with two paths, one set to 100% and the other set to 0%

Then you can edit your new paths without people flowing through them right away. This helps you prevent unintended consequences as you add paths to your automation.

When you’re done editing your automation to account for the new paths, return to your random cohort branch and update the distribution to reflect your new paths.

If your branch sends each person down the same path every time, or cohorts by objects, updating the distribution sends some people down different paths in subsequent journeys.

Updated November 10, 2026