Lesson 3: Percentage Rollouts and Experimentation

So far, we've learned how to give specific feature flag variations to individuals and segments of users. But what if we want our users to be randomly bucketed into one of the available variations? There's a variety of reasons we might want to do this, for instance:

• Gradually rolling out a new feature
• Progressive cutover to a new service
• Sunsetting an old feature
• Experimentation

In these scenarios, you can use a percentage rollout rule type to serve a random variation, rather than serving a specific variation. In this section, we'll be using a percentage rollout rule to randomly bucket ourselves into one of the available variations of a multivariate flag.