Feature Flag Rollout Planner
Plan staged feature flag rollouts with users per stage, guardrails and a deterministic bucketing simulator.
Plan updated: 5 stages, 6d 0h of observation.
Related Tools
What does the Feature Flag Rollout Planner do?
It turns a staged release into a concrete plan: users per stage, new and cumulative exposure, observation windows, the total timeline, and the conditions to proceed or roll back. It validates the plan (duplicate or descending percentages, missing observation time, values over 100%) and exports it as Markdown, JSON, YAML, a checklist or a compact AI context block.
How to use
- 1.Enter the feature name, flag key, owner and audience size, or load an example.
- 2.Choose a preset or edit the stages: rollout %, observation time and an optional note.
- 3.Add "Proceed if" guardrails and "Roll back if" conditions.
- 4.Review the stages and checks, then test user IDs in the Simulator.
- 5.Copy or download the plan from Export.
How the assignment simulator works
Each user gets a bucket from murmur3_x86_32 of "flagKey:userId" with seed 0, reduced to 0.00–99.99. A user is ON when the bucket is below the stage percentage, so the same user keeps the same bucket and stays ON as the percentage grows. Changing the flag key reshuffles everyone. This is a generic deterministic simulation; LaunchDarkly, Unleash, GrowthBook and other services use their own hashing, salts and bucket ranges, so their assignments differ.
Privacy and limits
Planning, simulation and exports run in your browser. The plan is not saved; only the selected export format is remembered. Plans hold up to 20 stages and 12 conditions per list, and the simulator accepts up to 1,000 user IDs (64 KB).
