New features can go out switched off. Then the team flips them on, for a few people first, whenever they're ready.
Imagine a new lamp installed in every room of a hotel, but all of them switched off. The manager can turn them on one room at a time, see how guests like them and switch any off in an instant.
A feature flag is that switch. It’s a setting inside an app that turns a feature on or off, for everyone or just for some people, without releasing a new version.
What teams use them for
- trying a feature on a small group before everyone
- switching something off instantly if it causes problems
- giving paying customers features others don't get
- testing two versions to see which works better
Why it’s safer
Without flags, every change goes live for everyone the moment it’s deployed. With flags, going live and switching on are separate steps.
What it isn’t
It isn’t a rollback. A rollback puts back an old version. A flag just switches one feature off while everything else stays.
Where the comparison breaks down
Light switches stay forever. Old feature flags should be removed once a feature is settled, or the app fills up with forgotten switches.
How the words fit together
You deploy with the feature flag off. A beta group gets it first. If it goes wrong, flip it off instead of a rollback.
If a friend has a new feature in an app and you don't, it's probably behind a feature flag that's been switched on for some people first.

When you'll hear it
“It's behind a flag.” The feature is built but switched off until we decide to turn it on.
“Roll it out to 10 per cent.” Switch the feature on for one in ten users first, to see how it goes.
“Kill the flag.” Switch the feature off straight away, because it's causing trouble.
“Can we switch this on for a small group first and turn it off quickly if something goes wrong?”


