Create ranking formulas create-ranking-formulas

Access the Strategy setup menu, then select Ranking formulas tab. The list of previously created formulas is displayed.

Click Create formula.

Specify the formula name, and add a description if desired.

{width="80%"}

Optionally, click Select AI model to set the model that will be used as a reference to build your ranking formula.

Every time you refer to a model score when defining your formula below, the AI model that you selected will be used.

Define the conditions that will determine the ranking score for the matching decision items. You can:

• Fill in the Criteria section using the formula builder, and/or
• Click Switch to code editor to define or refine ranking logic with PQL in the code editor.

If needed, click Switch to code editor to add an expression that uses PQL syntax alongside the formula builder. This option complements the user interface fields in the steps below, so you can combine both approaches in the same ranking formula. For more on how to use PQL syntax, refer to the dedicated documentation. Syntax for decision item attributes and copy-paste examples are provided in the Use the code editor section.

In the Criterion 1 section, specify the decision items that you want to apply a ranking score to by doing the following:

• select a decision item attribute
• select a logical operator
• add a matching condition - you can either type a value, or select a profile attribute or context data

{width="70%"}

Optionally, you can specify additional elements to refine the matching conditions for your criteria to be true.

{width="80%"}

For example, you defined Criterion 1 such as the Weather custom attribute Equals the warm condition. Additionally, you can add another condition such as if the first condition is met and if the temperature exceeds 75 degrees at the time of the request, then Criterion 1 is true.

Create an expression that will assign a ranking score to the decision items that meet the condition defined above. You can reference any of the following:

• the score that came out of the AI model that you optionally selected in the Details section above;
• the decision item’s priority, which is a value manually assigned when creating a decision item;
• any attribute that might live on the profile, such as any externally derived propensity score;
• a static value that you can assign in a free format;
• any combination of all the above.

{width="70%"}

Click Add criterion to add one or more criteria as many times as needed. The logic is as follows:

• If the first criterion is true for a given decision item, it takes precedence over the next ones.
• If not true, the decisioning engine moves on to the second criterion, and so on.

In the last field, you can build an expression that will be assigned to all decision items that do not meet the above criteria.

{width="70%"}

When your formula is ready, click Create.

In the ranking formulas list, click the red indicator icon next to the formula name.

The Optimize window opens, displaying the original PQL expression alongside the AI-suggested version.

To validate that both expressions produce identical ranking results, click Download Optimisation Analysis (TSV) to download a file showing how simulated profiles are evaluated against each version.

Once satisfied, click Apply to replace the original expression with the optimized one.