Learning about experimentation
The Experimentation section of a feature flag was covered in detail in Chapter 5, Experimentation, so it is best to read that chapter to understand how best to use this functionality, along with Metrics. However, there are a couple of additional pieces of functionality that were not covered in that chapter that we will cover next.
When working with the experimentation functionality, it is possible that testing can distort the results of the test. Usually, the only data that should be collected and analyzed is that of the percentage or ring rollout of a flag to customers. When testers or stakeholders are validating or signing off a feature, their usage of an experiment should not impact the data. Normally, the sign-off will be performed before an experiment is presented to any customers. However, testing might continue in production (even once live) to ensure that all the variations of an experiment work. In that case, specific users or targeting rules...
