💡 Note: this feature is still in progress. See "Future work".
Rollouts are single-branch experiments that differ from a traditional experiment in a number of ways.
When I have a “winning branch” from an in-flight experiment, I want to easily set some percentage of the non-experiment population to that feature configuration. To create a rollout from an experiment, the experiment is cloned and a new rollout is created. Through this clone, the population percentage can be changed. Additionally, a rollout can be manually created (without cloning) through the "Create new" button on the Nimbus home screen. To mark the new item as a rollout, check the "This is a rollout" box on the "Branches" page:
A rollout is different from an experiment in the following ways:
- A rollout only has a single branch
- A client can be enrolled in both a single experiment AND rollout for a given feature
- The experiment feature value takes precedence over the rollout feature value
- Rollouts use a separate bucketing namespace from experiments so you don't need to worry about the populations colliding
- When configuring the audience for a rollout of a given feature, I should see information about all live experiments/rollouts of the same feature so I clearly understand how that will impact my population sizing. This can be done by filtering on the home screen by the given feature. All experiments/rollouts that have been created for that feature will appear in the filtered list.
- I should be able to view all live experiments and rollouts for a given feature. From the Nimbus home screen, all rollouts and experiments are viewable. There is no specific filtering to allow only rollouts at this time (see "Future work", separation of concerns).
Experimenter currently supports the following platforms:
- Fenix (Firefox for Android)
- Focus Android
- Firefox iOS
- Focus iOS
The minimum version supported for rollouts on all platforms (listed above) is currently 105. See Experimenter for more details.
- Live editability for rollouts
- Differentiate opt-out support for experiments vs rollouts
- Separation of concerns: separating experiment and rollout state
- I want to be able to create new experiments or update my rollout population without having to launch in a particular order to avoid unwanted interactions on the client.
⚠️ Unimplemented See "Future work", live editability.
When I need to increase or decrease the population percentage of a rollout configuration, I want to do this without creating a new rollout instance.
Currently, a rollout is created using the same workflow as an experiment (Nimbus home -> Create new -> ✅ This is a rollout). The population percentage is defined at create time, and is not editable while live (see "Future work").
When increasing/decreasing the population percentage, I should be able to request a review and approve/reject via the experimenter UI.
I should be able to differentiate between experiments and rollouts when looking at a directory listing.
Rollouts are not differentiated on the main home page. When creating a new rollout, you can see the "Rollout" tag added next to the experiment name. Behind the scenes, rollouts are differentiated from experiments by their bucket prefix.
The dashboards for rollouts that OpMon generates follow this URL pattern:
https://mozilla.cloud.looker.com/dashboards/operational_monitoring::<slug with underscores>