AI Campaign Decisioning helps you choose the best message copy for each end user in your Push campaigns. A Campaign Decisioning Policy defines the goal and guardrails. AI learns from campaign performance to understand which content works best for different end users and delivers the variation most likely to convert for each end user.

​

Policy Components

A Campaign Decisioning Policy includes the following:

• Reward event: Defines success for the policy, such as product purchase. Each policy has one reward event, and the selected event must be positive.
• Performance metrics: Additional metrics are tracked as positive or negative signals to measure impact beyond the reward event. Performance metrics are grouped into four event groups:
  • Reward Event
  • Brand Engagement
  • Message Interactions
  • Custom Events
• You can have up to 5 conversion metrics in a policy. Of these, 2 metrics are added by default and cannot be removed:
  • Reward event
  • App/site opened
• You can add up to 3 more conversion events in Brand Engagement or Custom Events, as either Positive or Negative metrics.
• In Message Interactions, you can add only 2 negative metrics. They are optional and are not part of the conversion metrics:
  • Email Spam Complained
  • Email Unsubscribed
• The Reward Event row is auto-populated and locked. It always matches the reward event selected in Basic Details. The following rules apply:
  • Events can be classified only as Positive or Negative.
  • The same event cannot be added to both the Positive and Negative columns.
• Guardrails: Define the eligible audience, daily send limits, optional time slots, and policy control group. AI operates only within these boundaries. Your workspace global frequency cap also applies.

​

Content Selection

MoEngage selects content in the following way:

1. You create multiple complete message variations for a campaign.
2. Each variation is labeled using your Content Labels taxonomy.
3. The policy learns which labels each end user responds to and delivers the variation most likely to convert for that end user.

​

Sending Logic

Once AI selects the best message variation for an end user, the message is sent only if it meets the policy limits and global frequency cap:

• Policy limits: The policy’s maximum sends per end user and any defined time-slot limits are checked first.
• Global frequency cap: The workspace-level limit on messages per end user is checked next.

If the global frequency cap is reached, the AI-selected message will not be sent. This is expected behavior.