Android 9 (API level 28) introduces a new battery management feature, App Standby Buckets. App Standby Buckets help the system prioritize apps' requests for resources based on how recently and how frequently the apps are used. Based on app usage patterns, each app is placed in one of five priority buckets. The system limits the device resources available to each app based on which bucket the app is in.
The system dynamically assigns each app to a priority bucket, reassigning the apps as needed. The system may rely on a preloaded app that uses machine learning to determine how likely each app is to be used, and assigns apps to the appropriate buckets. If the system app is not present on a device, the system defaults to sorting apps based on how recently they were used. More active apps are assigned to buckets that give the apps higher priority, making more system resources available to the app. In particular, the bucket determines how frequently the app's jobs run, how often the app can trigger alarms, and how often the app can receive high-priority Firebase Cloud Messaging messages. These restrictions apply only while the device is on battery power; the system does not impose these restrictions on apps while the device is charging.
Note: Every manufacturer can set their own criteria for how non-active apps are
assigned to buckets. You should not try to influence which bucket your app is
assigned to. Instead, focus on making sure your app behaves well in whatever
bucket it might be in. Your app can find out what bucket it's currently in by
The buckets are:
- Active: App is currently being used or was very recently used
- Working set: App is in regular use
- Frequent: App is often used, but not every day
- Rare: App is not frequently used
In addition, there's a special never bucket for apps that have been installed but have never been run. The system imposes severe restrictions on these apps.
Note: Apps that are on the Doze exemption list are exempted from the App Standby Bucket-based restrictions.
Note: The descriptions below are for the non-predictive case. By contrast, when the prediction uses machine learning to predict behavior, buckets are chosen in anticipation of the user's next actions rather than based on recency of usage. For example, a recently used app could end up in the rare bucket because machine learning predicts that the app won't be used for several hours.
An app is in the active bucket if the user is currently using the app or very recently used the app. For example:
- The app has launched an activity
- The app is running a foreground service
- The app has a sync adapter associated with a content provider used by a foreground app
- The user clicks on a notification from the app
If an app is in the active bucket, the system does not place any restrictions on the app's jobs, alarms, or FCM messages.
An app is in the working set bucket if it runs often but it is not currently active. For example, a social media app that the user launches most days is likely to be in the working set. Apps are also promoted to the working set bucket if they're used indirectly.
If an app is in the working set, the system imposes mild restrictions on its ability to run jobs and trigger alarms. For details, see Power management restrictions.
An app is in the frequent bucket if it is used regularly, but not necessarily every day. For example, a workout-tracking app that the user runs at the gym might be in the frequent bucket.
If an app is in the frequent bucket, the system imposes stronger restrictions on its ability to run jobs and trigger alarms, and also imposes a cap on high-priority FCM messages. For details, see Power management restrictions.
An app is in the rare bucket if it is not often used. For example, a hotel app that the user only runs while they're staying at that hotel might be in the rare bucket.
If an app is in the rare bucket, the system imposes strict restrictions on its ability to run jobs, trigger alarms, and receive high-priority FCM messages. The system also limits the app's ability to connect to the internet. For details, see Power management restrictions.
If your app is already following best practices for Doze and app standby, handling the new power management features should not be difficult. However, some app behaviors which previously worked well might now cause problems.
- Do not try to manipulate the system into putting your app into one bucket or another. The system's bucketing methods can change, and every device manufacturer could choose to write their own bucketing app with its own algorithm. Instead, make sure your app behaves appropriately no matter which bucket it's in.
- If an app does not have a launcher activity, it might never be promoted to the active bucket. You might want to redesign your app to have such an activity.
- If the app's notifications aren't actionable, users won't be able to trigger the app's promotion to the active bucket by interacting with the notifications. In this case, you may want to redesign some appropriate notifications so they allow a response from the user. For some guidelines, see the Material Design Notifications design patterns.
- Similarly, if the app doesn't show a notification upon receiving a
high-priority FCM message, it
won't give the user a chance to interact with the app and thus promote it to
the active bucket. In fact, the only intended use for high-priority FCM messages is to
push a notification to the user, so this situation should never occur. If you
inappropriately mark an FCM message as high-priority when it doesn't trigger
user interaction, it can cause other negative consequences; for example, it
can result in your app exhausting its quota, causing genuinely urgent
FCM messages to be treated as normal-priority.
Note: If the user repeatedly dismisses a notification, the system gives the user the option of blocking that notification in the future. Do not spam the user with notifications just to try to keep your app in the active bucket!
- If apps are split across multiple packages, those packages might be in different buckets and, thus, have different access levels. You should be sure to test such apps with the packages assigned to various buckets to make sure the app behaves properly.