Yes, you can use a decision policy across multiple API experiences. However, you must ensure that the eligibility rules and selection logic are appropriate for the content in those locations.

Each audience within a Personalize API experience can have a maximum of five decision policies distributed across all its variations. To keep API response times low, it is recommended to limit the number of decision policies lower than 5 in any single API experience.

If no offerings are eligible for a user, the decision policy will revert to the fallback offering. Fallback offering ensures that the surface is not left empty and can display general or default promotions (for example, brand awareness campaigns or app download banners).

When multiple offerings are eligible to be displayed, a precise ranking system determines which one is selected. The evaluation follows a strict order of rules:

• First, the offering with the highest priority value is selected.
• If two or more offerings share the same priority value, the system evaluates their end dates and chooses the one with the earliest end date.
• If the end dates are also identical, the offering with the earliest start date takes precedence.
• In the rare case that the priority value, end date, and start date are all identical, one offering will be selected randomly from the eligible group.

Yes, decision policies allow you to set the number of output offerings to restrict the number of offerings shown. For example, you can configure a policy to display exactly one promotion in a banner carousel or up to three promotions in a list.

A fallback offering is the default content shown to users who do not qualify for any primary offerings. When an offering serves as a fallback, it bypasses all eligibility rules, such as audience segmentation and frequency capping.

Yes, a decision policy can be configured to be dynamic. For example, you can dynamically pool offerings tagged with Holiday Sale in a single decision policy.

Yes, you can edit a live decision policy at any time, even if it is part of an active campaign. Changes take up to 30 seconds before they start reflecting in the campaign. Make sure to review your changes carefully before saving to prevent unintended consequences. Be watchful when switching between manual and dynamic offering pools, as this will alter the available offerings.

A decision policy accepts a maximum of 50 primary and 10 fallback offerings as input, and returns a maximum of 10 offerings in its output.

No. Each Decision Policy uses one ranking strategy.

No, but any Offering that does not have a value for a given attribute will receive a score of 0 for that parameter. If the attribute carries significant weight, this will consistently push that Offering lower in the ranking. It is good practice to ensure all Offerings in a policy are tagged with values for the attributes included in the formula.

Yes. You can edit the weights in a live Decision Policy. The updated formula takes effect immediately — all Offerings are re-ranked using the new weights from the next decisioning request onwards. No re-publishing of individual Offerings is required.

Deleting an attribute that is in active use within a Decision Policy is blocked. You must remove the attribute from the formula before deleting it. 

Yes, and this is the recommended approach for most use cases. Fixed Value attributes encode the strategic importance of the Offering (consistent across all customers), while Dynamic Value attributes encode individual relevance (varying per customer). Combining both gives you a formula that balances business goals with personalisation. 

The engine takes the raw CTR percentage of the Offering. This score is then multiplied by the weight assigned to Offering Performance in the formula. 

The manual priority rank assigned to the Offering (a value from 1 to 10) is used directly as the base score. A priority of 1 becomes a score of 1; a priority of 10 becomes a score of 10. Its influence is primarily determined by the weight you assign to it. 

Simulation of Custom Formula ranking is currently not available. To validate your formula, we recommend publishing to a small early-access audience first and reviewing Offering delivery data before rolling out broadly.