Percentage-based feature flag targeting divides the contexts eligible for a rollout—such as users, accounts, or devices—among the flag’s possible variations. The system typically uses a stable identifier and a provider-specific bucketing method so the same context receives the same variation on later evaluations. The configured percentage controls the allocation rule; it does not guarantee an exact headcount, especially in a small group.
What percentage targeting does
A feature flag can return different values or experiences, often called variations. A percentage rollout assigns eligible evaluation contexts to those variations according to configured weights. For example, a two-variation rollout might assign 50% to a control experience and 50% to a new experience. LaunchDarkly documents manual percentage rollouts in which the variation weights total 100% (LaunchDarkly JSON targeting).
Two decisions are involved: first, whether a context qualifies for the rollout; second, which variation it receives. Keeping eligibility separate from allocation makes it easier to reason about why a person did—or did not—see a feature.
How the assignment is made
- The application supplies evaluation context. This contains a targeting key identifying the subject—such as a user or service—and may include attributes used by rules. OpenFeature explains that many implementations need a unique targeting key for deterministic fractional evaluation. It also cautions that providers may handle or persist context data, so avoid supplying unnecessary personal information (OpenFeature evaluation context).
- Rules determine eligibility. Individual targets and conditional rules can match before the percentage allocation is applied. In LaunchDarkly, the fallthrough rule applies when no higher-priority rule matches; the API documentation describes flag fallthrough behavior (JSON targeting; Feature Flags API).
- The provider calculates a bucket. It uses a stable key and provider-specific inputs to place the context in a rollout range. In Unleash, the documented method combines a context field with a strategy
groupIdand hashes it with MurmurHash to produce a value from 0 to 100. The defaultgroupIdis the flag name; sharing group IDs can correlate assignments across flags, while changing one can reshuffle them (Unleash stickiness). - The bucket maps to a variation. The provider compares the bucket with configured weights. LaunchDarkly’s API uses a 0-to-100,000 weight scale: 60,000 represents 60%, and weights across variations should total 100% (LaunchDarkly Feature Flags API).
- Later evaluations usually reproduce the assignment. With the same relevant inputs and configuration, deterministic bucketing can return the same result without storing a separate assignment record for every context. LaunchDarkly describes deterministic assignment in its experimentation traffic documentation; that description is specifically about experiments and should not be taken as proof that every flag product uses the same algorithm (LaunchDarkly experiment traffic assignment).
Which identity should be bucketed?
The rollout unit is the entity whose experience needs to stay consistent. A user key may give each person an independent assignment; an account key can keep everyone in a customer organization on the same variation; a device key can make exposure consistent on one device. Sessions are another possible unit, but a session-based assignment may change on a later visit. The right choice depends on the feature’s consistency and risk boundary, not on a universal default. LaunchDarkly documents user, device, and account context kinds, while Unleash exposes stickiness options (LaunchDarkly progressive rollouts; Unleash gradual rollout).
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstall#1 Best Overall
Use a stable identifier available across the journey in which consistency matters. Anonymous users who later log in can otherwise be assigned differently before and after authentication. LaunchDarkly documents device contexts and multi-contexts as one way to associate anonymous and logged-in identities (LaunchDarkly percentage rollouts by context attribute). If targeting rules use one context kind while allocation uses another, check the provider’s behavior: LaunchDarkly warns that contexts without the expected multi-context may receive the first variation with a positive weight.
Why the observed count may differ from the setting
A percentage is an allocation proportion, not a promise to select an exact number of named contexts. The provider maps identities into buckets, and a small eligible population can land unevenly. LaunchDarkly illustrates the difference with a 10% rollout: among 10,000 contexts, the result would be about 1,000; among 20, it could be zero, one, or two. These are vendor examples, not independent measurements (LaunchDarkly progressive rollouts).
Rank #2
- 4LessCo UNDER NEW MANAGEMENT Windless Swooper Flag Feather Banner Sign 2.5x11.5 ft Tall Large (Hardware NOT Included) yb
- 2.5 ft by 11.5 Ft Tall Flag.
- Printed on one side, backside same image but in reverse.
- This flag only works with windless swooper pole.
- Pole and spike are NOT included.
When a close aggregate share matters, assess the size of the eligible population and the rollout unit together. Bucketing accounts instead of users may mean fewer independent contexts, for example, even if the number of people affected is large. Preserve the unit needed for product consistency rather than changing it solely to make the percentage look closer.
What happens when a rollout changes?
Assignment behavior after edits depends on the provider and the specific rollout mechanism:
Rank #3
- UNDER NEW MANAGEMENT Windless Feather Swooper Flag Kit - No Wind Is Needed
- 2.5x11.5 Ft Tall Flag
- 15ft Tall Heavy Duty Deluxe Aluminum/Faberglass Pole
- Steel Ground Spike
- Unleash gradual rollout: its documentation says increasing the percentage retains contexts already inside the rollout and adds more; lowering it removes contexts above the new threshold. The stickiness key and group ID therefore matter to cohort stability (Unleash stickiness).
- LaunchDarkly percentage rollout: its documentation says the same contexts remain assigned when a percentage rollout is stopped and restarted if the configuration and context kind remain unchanged. A newly created progressive rollout may allocate a different set (LaunchDarkly progressive rollouts).
Do not assume those behaviors apply across products. A provider may use different hash inputs, group identifiers, or algorithms, and a configuration edit can alter the cohort even when the displayed percentage stays the same.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Why provider migration can reshuffle a cohort
The same stable identifier and percentage do not guarantee the same people remain in a rollout after a provider change. Unleash’s migration guidance says its hashing differs from LaunchDarkly’s, so a 50% allocation in each system need not select the same contexts (Unleash migration guidance). If continuity is important, plan and validate the migration’s assignment behavior instead of relying on matching percentage settings.
Quick Recap
Best Value
Implementation checklist
- Choose the rollout unit—user, account, device, or another supported context—that matches the feature’s consistency boundary.
- Use a stable targeting or stickiness key and decide how anonymous-to-known identity changes should be associated.
- Define eligibility rules separately from percentage weights, and confirm which rule applies when no higher-priority condition matches.
- Check how the provider represents weights, selects its bucket, and handles changes to percentage, context kind, group ID, or rollout lifecycle.
- For migrations, verify whether the new provider preserves the same cohort; a matching percentage alone is not enough.
- Include only context attributes needed for targeting, since providers may handle or persist evaluation context.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
