Arbitration
This domain tests how Pega picks the winner among eligible actions, and how business levers let a business tune that choice without rewriting strategies.
The core mechanic
Arbitration runs after applicability, suitability and contact policy have removed actions. It ranks survivors using business levers:
- Propensity – the model score for the action.
- Value – the business worth defined on the action.
- Context weighting – how much the current context (channel, customer segment, direction) matters for this group.
Raising a lever's weight increases its influence on the ranking. Lowering it lets other levers dominate. Weights are configured per issue/group, so tuning one group leaves others untouched. For inbound channels arbitration produces the single best action; outbound flows can deliver several, still bounded by contact policy and volume constraints.
Worked example
Two eligible actions compete. Action A has propensity 0.30 and value 120; Action B has propensity 0.55 and value 15. With a heavy value weight, A wins because its business worth dominates. The business then decides short-term revenue matters less than engagement, so the propensity weight is raised and the value weight lowered; B now outranks A. Nothing in the strategies changed - only the levers. Reasoning: arbitration is a ranking, so the answer depends entirely on the configured weights and the context, not on which action scored higher.
Traps that cost marks
- Placing arbitration before contact policy; contact policy removes candidates before ranking.
- Assuming the highest propensity always wins.
- Expecting one group's lever change to apply globally.
- Forgetting that arbitration cannot rescue an action with no treatment or no value.