Set up a percentage rollout
Rolling a feature out to a slice of your fleet first is the single most useful targeting move. In a few minutes you’ll have a flag serving 10% of devices, with a clear path to 25%, 50%, and everyone.
You’ll need Developer or above for a non-production environment; production-class environments take Admin, or a proposal.
1. Add a rule with a percentage
Section titled “1. Add a rule with a percentage”Open your flag, pick the environment, and edit its targeting rules. Add a rule and set its rollout percentage to 10.
- Leave the rule’s conditions empty to roll out to 10% of all devices.
- Or add conditions first — say, cohort is
beta— and the percentage applies to the devices that pass them: 10% of your beta testers.
Set what the rule serves (for a boolean flag, true), save, and you’re live: one in ten
matching devices now gets the new value, everyone else falls through to the rules below or
the flag’s default.
2. Watch, then widen
Section titled “2. Watch, then widen”When you’re happy, edit the rule and raise the percentage. This is the part you can trust:
- Nobody flickers. A device’s place in the rollout is computed from its identity and the flag’s key — the same device gets the same answer every time.
- Raising the percentage keeps the original cohort in. Going 10 → 25 adds 15% of devices; the first 10% stay exactly where they were.
- Each flag rolls its own dice. A device in this flag’s 10% isn’t automatically in the next flag’s 10% — no permanently unlucky devices.
And because cohort membership is pure arithmetic, FortressFlag stores nothing about who’s in the rollout — there’s no cohort list anywhere, including for us.
3. Finish the rollout
Section titled “3. Finish the rollout”At 100%, every matching device gets the value — at that point the honest end state is to make the rolled-out value the flag’s default and delete the rule. A flag whose rollout finished months ago but still reads “100% rollout” is a small lie your future self has to re-check.
Rolling back is the same edit in reverse: lower the percentage, or delete the rule and let defaults serve. Either way the change reaches online devices within about a minute.
Where to next
Section titled “Where to next”- Targeting — the full rule machine this sits inside
- Require approval for production changes — rollouts in production, safely
- Schedule a change — set a flag to flip at a chosen time