A new NBA Designer action that never appears on the web channel at Fabrikam Insurance
Fabrikam Insurance has 3 issues (Service, Retention, Growth), 8 groups, and 40 actions in Pega NBA Designer. Engagement policies are configured: eligibility when the customer is active; applicability by channel; suitability by risk score; contact policy of at most two impressions per week. A new switch-and-save action in the Growth issue is active, its group is enabled, and its start date is in the past. Its audience is a segment of 50,000 customers. The business complains that the action is never selected on the web channel. The action's only treatment is configured for email. The NBA Designer top-level strategy has an arbitration component using propensity and value. A separate suppression rule excludes customers with an open complaint. The action has no adaptive model yet. Web traffic is 300,000 interactions per week.
1. Walk through the order in which NBA Designer decides whether the switch-and-save action can appear on the web channel. Where is the most likely failure point, and what evidence would confirm it?
Read the worked answer
The order is: issue, group, and action must be enabled and in scope; the customer must be in the audience; the action must have a treatment for the web channel; then engagement policies apply - eligibility, applicability, suitability, contact policy; suppression applies; only survivors reach arbitration. The most likely failure is the treatment: the action only has an email treatment, so it is not applicable to web and never reaches arbitration. Confirm by previewing a web customer in NBA Designer and checking the action's treatment list and the applicability result. A second likely failure is the missing adaptive model: if arbitration uses propensity and the action has no model, its propensity may be null or zero and it loses or is excluded. Confirm by inspecting the arbitration input for that action. Suppression for an open complaint and the two-impression contact limit are also candidate exclusions, but they would affect only a subset, not all web customers.
2. The business wants to fix the problem by raising the action's arbitration weight. Why is that unlikely to work, and what would you change instead?
Read the worked answer
Raising the weight is unlikely to work because arbitration only ranks actions that have already passed scope, audience, treatment, engagement policy, and suppression. If the action has no web treatment, it is filtered before arbitration and the weight is never applied. If it has no propensity model, multiplying by a weight does not create a valid score; zero times any weight is still zero. The correct fix is to add a web treatment, provide a propensity (adaptive model or a defensible default), verify audience membership, and then test. Only after the action reaches arbitration should weight be used to influence rank. Weight inflation also risks crowding out other Growth and Service actions.
3. How would you distinguish an engagement-policy exclusion from an arbitration loss using NBA Designer diagnostics?
Read the worked answer
Use a controlled preview: pick a web customer who is active, in the audience, has no open complaint, and is under the contact limit; run NBA Designer preview or explain and record at which stage the action drops. Then repeat with a customer who has an open complaint and one who is over the contact limit to see the exclusion reasons. Compare a customer with the action to one without to isolate audience effects. If the action appears in preview but not in production, check issue/group channel enablement and whether production uses a different NBA Designer version or data flow snapshot. This proves whether the failure is engagement policy, treatment, model, or arbitration rather than guessing from final output.