Metric alerts

Last updated 10/02/2026

Metric alerts monitor the performance of key metrics and proactively send notifications when anomalies occur, helping you spot problems promptly and reduce unnecessary losses. With metric alerts, you can keep track of:

• Whether core product features are working properly
• Whether business performance meets expectations
• Whether the product ecosystem is healthy

Create an alert

To create an alert, click Tool → Metric Alerts at the top of the navigation bar, or add analysis metrics directly in Events Analysis.

Basic information

• Alert name: Up to 24 characters
• Time zone: Select the time zone used to calculate the alert metric. Available when project time zone is turned on
• Alert frequency: Select Hour or Daily to set how often the alert task runs
• Alert period: Monitored metric values usually fluctuate over time. For example, most products have fewer active users in the early morning, so alert rules designed for normal hours may not suit that period. In this case, you can configure an alert period so that no alert notifications are sent in the early morning.

Alert rules

Alert metrics

You can add simple metrics or custom formula metrics.

Rule settings

You can use the following monitoring rules to determine whether a metric has changed abnormally:

• Compare with a fixed value: The metric is compared with a fixed value. A value below or above the fixed value is considered abnormal.

  • For example, send an alert when active users fall below a fixed value
  • Suitable for periods when the business is relatively stable

• Compare with a relative value: The metric is compared with the previous period or the same period in a prior cycle.

  • For example, send an alert when active users are 30% lower than the same period yesterday
  • Suitable for steadily growing businesses

• Compare with a statistical distribution: The metric is compared with its average over a past period.

  • For example, send an alert when active users fall below the average over the past 24 hours
  • Compared with "Compare with a relative value", this rule eliminates the impact of single-point anomalies on alert decisions

• Based on predicted values: An alert is sent when the metric is above the predicted upper bound or below the predicted lower bound.

  • If you don't have a clear monitoring rule, choose predicted value alerts
  • The prediction model uses the Hannan-Rissanen algorithm. For the paper on the algorithm, see this document
  • The algorithm needs at least 10 data points for prediction and uses at most the 100 most recent data points

Group-by alerts

On top of the alert metric, you can also turn on Group by to monitor dimensions such as channel, platform, and level with more granular alert rules.

First, select an event property or user property. Then you can select specific group-by values in the alert rules.

For example, if the alert metric is the number of users who triggered the active event and you select Apple App Store in a rule, the system checks for abnormal changes based on the alert metric data for that group-by value, that is, whether DAU in the Apple channel is abnormal. This keeps overall data from masking what is actually happening.

Note that if a rule involves multiple group-by values, they are counted as a whole. For example, if you select Apple App Store and Google Play Store, the DAU of both channels is combined for evaluation. You can also leave the group-by values empty, in which case the alert metric data is counted in total (that is, without grouping).

Delay settings

Data takes time to travel from production to storage. To make sure alert tasks calculate on complete data, you can delay the calculation time of alert tasks based on your project's data latency.

• If you monitor hourly and set a 15-minute delay, the alert task for 6:00 runs at 6:15
• If you monitor daily and set a 30-minute delay, the alert task for the previous day runs at 0:30 every day

Notification methods

Multiple notification channels are supported, including email, SMS, DingTalk Robot, WeCom Robot, and Feishu Robot.

Notifications show all rules that triggered an abnormal change. If you use group-by alerts, notifications also show the specific group-by values.

Manage alerts

Number of alert tasks

A project can run up to 50 alert tasks at the same time. When this limit is exceeded, new tasks can't be saved.

Usage permissions

Permission details	Root	Project Owner	Analytics Admin	Analyst	Member
View the Custom Alert list	●	●	●	△	○
Create, edit, and delete own alert tasks	●	●	●	△	○
Edit and delete others' alert tasks	●	●	●	△	○

Permission notes:

● The role always has this permission

▲ The role has this permission by default, but it can be removed

△ The role doesn't have this permission by default, but it can be granted

○ The role never has this permission