Process Control-AB Test Components

AB testing works as follows: to achieve the same target event, N schemes are created. Users participating in the test are randomly divided into N groups, and each group uses a different scheme simultaneously. After a period of time, the target event achievement rate of each group is compared, and the scheme with the highest rate wins.

For example, you can randomly divide the same audience into two groups of 50% each, send different promotional SMS messages to each group, and then compare which message generates higher sales.

To configure a universal filter, perform the following steps:

1. Click the AB test component and configure it in the panel that appears on the right, as shown in the following figure.

2. Enter a step name.

3. Specify the name of the group and the percentage of the number of people in each group. You can click New Group to add a group.

   Note

   You can specify 2 to 30 groups. The total number of people in all groups must be equal to 100%.

4. Set end conditions:

   • Runs until all users participate in the AB test or until the automated campaign ends.

   • After the total number of users entering the AB test reaches your specified limit, the AB test ends. The remaining users are no longer randomly grouped but are all routed to a designated group.

     You need to choose which group the remaining users are routed to:

     • The group with the highest target achievement rate based on the AB test results.

     • A specified group.

5. Click OK. A blank branch is generated for each group on the canvas and is finally merged to the "Test Target" node, as shown in the following figure.

6. Set each branch based on your scheme, as shown in the following figure.

   You can perform operations on components in branches, such as adding, copying, pasting, and deleting components. However, you cannot add lines to branches.

   

7. Click the "Test Target" node and set the test target in the panel that appears on the right, as shown in the following figure.

   1. Specify the wait time before determining whether a user reaches the target event. You can set a relative time or a specified time.

      • Relative time: After the user finishes executing the previous component, wait for a specified duration before evaluating. The unit can be minutes, hours, or days.

      • Specified time: Wait until a specific point in time, and then evaluate at that point.

   Note

   If the specified time has already passed when the user enters the "Test Target" node, the system evaluates immediately without waiting.

   2. The following table describes how to configure the Behavior Event component:

      Note

      User behavior event data is collected by applications such as apps, mini programs, and web pages and then reported to Quick Audience. For more information, see Event Reporting (New Version) or Event Management (Old Version).

      Item	Example value
      The source of the event.	For Event Reporting (New Version), select an event source type. For Use Event Management (Old Vdition), select the name of the event source channel.
      Event Name	Select an event.
      Restriction Attributes	You can set event property requirements. Turn on the Limit Attributes toggle and set the attribute requirements for the event. The event counts toward the target only when the attribute requirements are met. You can click Add Attribute to add an attribute requirement. Multiple attribute requirements support the,, or relationship. You can click the corresponding text on the left to switch between them.
      Limit times	You can require the same user to meet the event and attribute requirements a specified number of times before it counts as a goal. Turn on the Limit Times toggle and enter the number of times.

   3. Set the planned target achievement rate to compare with the actual target achievement rate of each group after the test runs.

      Note

      Target achievement rate = number of users who achieve the target event in the group ÷ number of users in the group × 100%

   4. Choose whether to perform follow-up marketing based on goal achievement:

      • Yes: Two groups are generated, "achieved" and "not achieved", and different follow-up actions are performed based on whether the user has achieved the goal, as shown in the following figure.

      • No: All users share a single follow-up flow, as shown in the following figure.

   5. Click OK. The "Test Target" node is configured.

AB Test Execution Action Analysis

After the AB test component runs, the test target node records the target achievement status of each group, so you can select the best scheme based on the target achievement rate.

The execution data is displayed above the test target node, including the winning group, the total number of targets reached across all groups, and the total number of inflows, as shown in the following figure.

Click the Execution Data section. The Execution Action Analysis panel appears on the right side, showing the number of inbound users, the number of users who reached the target, and the target achievement rate for each group. Group 1 has the highest target achievement rate, so the scheme for Group 1 is recommended.

Note

• Because the number of users entering each group may differ, use the target achievement rate as the primary metric for evaluating scheme effectiveness.

• If the end condition of the AB test node is set to the maximum number of participants, some users may not be randomly grouped. In this case, you can select All Users or Test Users in the upper-right corner to display only the data of users who participated in the AB test.