Using Feature Flags for Controlled System Changes and Rapid Feedback Loops

Feature flags allow us to make controlled changes to a system, observe the impact of those changes, and adjust as needed.

If a new feature increases a business growth metric (e.g., conversion rate) by 20% with statistical significance, we keep the change and roll it out to all users. Conversely, if a new feature causes a technical metric (e.g., request latency) to spike by 200%, we aim to roll back the change quickly.

Thus, when using feature flags we operate within a rapid business feedback loop: we change, observe the effect, and use those observations to decide the next change, as illustrated in Figure 11‑1.

The greatest value of unlocking the "feature flag" practice lies in emphasizing the importance of a mature feedback mechanism. Making changes without seeing their impact is like driving a car with a fogged windshield.

Despite the huge value of this feedback loop, many modern product delivery organizations have not tightly integrated their feature flag platforms with their analytics foundations, preventing an effective mechanism for rich feedback and iteration.