The Offer That Everyone Got
A telecom configures Offer Decisioning with three offers: Gold (eligibility: tenure >= 24 months), Silver (tenure >= 12 months), Bronze (fallback, no constraint). A ranking formula weights 'priority' equally across all three. In production, 70% of customers receive Bronze even though 55% have tenure >= 24 months. The decision is requested via a single decision scope on a web property. A developer notices the fallback offer has priority 100 while Gold has priority 50, and the ranking uses 'priority' as the sole score. Collections are configured but the decision scope references the default collection only.
1. Explain why eligible Gold customers are receiving Bronze.
Read the worked answer
Two compounding causes. (1) The ranking formula scores solely on 'priority,' and Bronze (fallback) has priority 100 vs Gold 50 - so among all eligible offers, Bronze wins the ranking every time. Eligibility filters first, then ranking picks the winner; a fallback that outranks everything becomes the default winner for everyone. (2) The decision scope references only the default collection, so collection-based targeting isn't narrowing candidates. The net effect: Gold customers are eligible for Gold but Bronze outranks it. The fix is to make the ranking formula reflect business value (e.g., tenure descending, or Gold > Silver > Bronze), and ensure the fallback only surfaces when no constrained offer is eligible - typically by excluding fallback from ranking or scoring it lowest.
2. How would you redesign the ranking and fallback so the fallback only appears when appropriate?
Read the worked answer
Eligibility handles 'who can see what'; ranking handles 'which of the eligible wins.' The fallback should never compete on the same score axis as targeted offers. Design: rank by a business-value attribute (tenure or explicit tier weight) descending, and give the fallback the lowest possible score so it only wins when it is the sole eligible candidate. Alternatively, separate the fallback into its own decision path triggered only when the primary decision returns empty. Then verify: a 24-month customer should return Gold, a 12-month Silver, a new customer Bronze - test each cohort explicitly rather than trusting the aggregate. Also confirm the decision scope pulls the right collection.