You can track the effect of an experiment by defining experiment metrics, such as click-through rate (CTR) and conversion rate (CVR). The basic fields required to calculate these metrics have been registered in the source data table. You need to define only the computing logic of each metric. This topic describes how to define an experiment metric and its computing logic.
Prerequisites
A source data table is registered. For more information, see Register a table.
Create a metric
Go to the ABTest page.
Log on to the PAI console.
In the left-side navigation pane, click Workspaces. On the Workspaces page, click the name of the workspace that you want to manage.
In the left-side navigation pane, choose .
In the left-side navigation pane of the ABTest page, choose .
On the Experiment Metrics page, click Create Metric.
On the Create Metric page, configure the parameters. The following table describes the parameters.
Parameter
Description
Metric Group
Metric Group Name
The name of the metric group. A metric group is a collection of interrelated metrics.
Metric Group Description
The description of the metric group. The description helps identify the metric group.
Sub-metric
Sub-metric Name
The name of the sub-metric.
Sub-metric Description
The description of the sub-metric. The description helps identify the sub-metric.
Source Table
Select a registered source data table.
Metric Template
The system provides the following commonly used built-in metric templates:
Page Views
Unique Visitors
Cumulative Value
Total Number of Occurrences
Statistics by Condition
You need to configure the Condition Field and Computing Field parameters based on the metric template you select.
If you set the Metric Template parameter to Page Views, Cumulative Value, or Total Number of Occurrences, you must configure the Computing Field parameter.
If you set the Metric Template parameter to Statistics by Condition, you must configure the Condition Field and Computing Field parameters.
Metric Computing Logic
After you select a metric template, a condition field, and a computing field, the system automatically configures the metric computing logic. You can also modify the metric computing logic based on your business requirements.
After you configure the parameters, click OK.
Manage an experiment metric
You can perform the following operations on a created experiment metric:
View the details of the experiment metric.
After the experiment metric is created, you can click the
icon next to the experiment metric to view the metric details, including the source data table, number of associated experiments, and metric computing logic.
Add a sub-metric for the experiment metric.
If the current sub-metric does not meet your business requirements, you can click Add Sub-metric in the Actions column to add a sub-metric and its computing logic. For more information about the parameters of sub-metrics, see Create an experiment metric.
Delete the experiment metric.
You can click Delete in the Actions column of the desired sub-metric to delete the sub-metric. You can also click Delete in the Actions column of the experiment metric to delete the experiment metric.
Associate an experiment metric with an experiment
To associate an experiment metric with an experiment, perform the following steps:
In the left-side navigation pane of the ABTest page, choose .
In the experiment list, click the name of the desired experiment to go to the experiment details page.
On the Configure Basic Information tab of the experiment details page, click Edit in the Experiment Metrics section.
In the Experiment Metrics panel, associate the core metric or followed metrics with the experiment and click OK.
References
You can associate an experiment metric with an experiment when you create the experiment. For more information, see Create an experiment.