PM Feature Flags
(2026 Edition)
How do PMs actually use feature flags? Mainly for five things — gradual rollout with kill switches, fast reverts without code changes, segment targeting, in-production A/B variants, and trunk-based development. Discipline matters more after launch than before: sunset flags, document who owns each one, and audit the count quarterly, since hundreds to low thousands is normal at scale and the real risk is untracked, unowned flags lingering past their purpose.
By Naman Goyal · Product manager · Builder of PM Streak · Updated July 3, 2026
5 uses and 4 practices for feature flag PMs.
Build Feature Flag PM Skills — Free →5 Uses
Gradual rollout — 1%, 10%, 50%, 100% with kill switches
Kill switches for fast revert without code change
Segment targeting — beta users, geo, plan tier
A/B testing variants in production
Trunk-based development — merge fast, ship slow
4 Practices
Sunset flags after rollout — tech debt accumulates fast
Document who owns each active flag
Use config-only changes for ops controls; code changes for product
Audit flag count quarterly
FAQ
How many feature flags is too many?
Hundreds to low thousands is normal at scale. The problem isn't count — it's untracked, unowned flags that linger past their purpose. A flag is a fork in code that increases test surface. Sunset discipline matters more than initial restraint on creating them.
Keep learning