Metric management
View the metric list
Go to Experiment → Metrics. The list shows each metric's name, display name, metric type, calculation method, group, metric direction, associated experiments, metric description, and creator.
Create a metric
Click Create Metric in the upper-right corner and fill in the metric information.
| Field | Description |
|---|---|
| Metric name | Starts with a lowercase letter and can contain lowercase letters, digits, and underscores, up to 80 characters |
| Display Name | Business name shown in experiment configurations and reports |
| Metric group | Used for categorized management; defaults to Default |
| Metric type | You can create Event metric on the current page |
| Calculation method | Currently Confidence Statistics is used. The system first aggregates data for each user, then runs statistical tests with users as samples |
| Calculation definition | Select a business event and an aggregation method. This example uses the event total of demo_business_conversion |
| Metric direction | A positive metric means higher values are better; a negative metric means lower values are better |
| Metric description | Enter the business meaning and statistical definition, up to 200 characters |
When you are done, click Create. After the metric is created, you return to the metric list, where you can search by name or click a display name to view details.
View metric details and associated experiments
Click a metric's display name to view Basic Information, Associated Experiments, and Finished Experiments.
Basic Information shows the metric ID, creation information, metric description, and the specific calculation definition. After a metric is referenced by an experiment, the Associated Experiments count in the list is updated accordingly. After an experiment ends, you can trace its name and status in Finished Experiments in the metric details.
After a metric is associated with experiments, its delete option may be grayed out. Before you adjust or delete a metric, check its associations to avoid affecting experiment reports.