Quick question about Predictive Routing Scorecard mismatch during Zendesk skill migration

Quick question about the discrepancy in routing scores when migrating from Zendesk’s round-robin skills to Genesys Cloud’s Predictive Routing. We are currently in the validation phase of our migration from Zendesk to Genesys Cloud, and the agent match percentages are wildly different. In Zendesk, we relied on simple skill tags and availability. Now, with Genesys Cloud’s predictive model, the score field in the routingScorecard endpoint returns values that don’t align with our historical Zendesk performance data.

We have mapped our Zendesk groups to Genesys Cloud teams and skills. However, when testing the predictive routing flow, agents with high Zendesk ticket resolution rates are getting low predictive scores (around 0.15) in Genesys. The skill match is correct, but the agentScore is unexpectedly low. We are using the GET /api/v2/routing/scorecards/{scorecardId} endpoint to debug this. The response shows agentScore is heavily weighted by utilization and handleTime, which we haven’t fully imported from Zendesk yet.

Here is the payload structure we are seeing in the scorecard response:

{
 "agentId": "12345678-1234-1234-1234-123456789012",
 "agentName": "Aki_Tester",
 "agentScore": 0.15,
 "agentScoreReason": "Low historical handle time data",
 "skillMatch": {
 "skillId": "87654321-4321-4321-4321-210987654321",
 "skillName": "Technical_Support",
 "matchScore": 1.0
 },
 "utilization": 0.85,
 "handleTime": 120.5
}

In Zendesk, this agent would have been top-priority due to high volume. How does Genesys Cloud handle the lack of historical data for predictive scoring? Do we need to backfill historical interactions via the Interaction API to boost these scores? Or is there a way to force a skillMatch-only routing mode during the migration period? Any advice on bridging this gap between Zendesk’s simple logic and Genesys’s predictive engine would be great.

I typically get around this by implementing a custom webhook payload transformation that normalizes Zendesk skill weights before ingestion into the Genesys Cloud routing engine. The discrepancy arises because Zendesk’s round-robin logic treats skills as binary flags, whereas Predictive Routing calculates a probability score based on historical interaction data, agent proficiency, and current queue depth.

Cause:
The routingScorecard endpoint returns a calculated match probability (0-100), not a static skill match. During migration, if historical interaction data is not imported or mapped correctly, the model defaults to a neutral baseline, resulting in lower scores compared to Zendesk’s deterministic availability checks.

Solution:
Configure a Data Action in Genesys Cloud to map Zendesk’s skill_group to specific Genesys Cloud routing_skills. Ensure the proficiency_level is explicitly set during the bulk import. For real-time validation, use the following cURL snippet to inspect the raw routing configuration:

curl -X GET "https://api.mypurecloud.com/api/v2/routing/users/{userId}/routingprofile" \
 -H "Authorization: Bearer {token}"

Verify that the skills array contains the correct proficiency values. Adjust the score threshold in your routing strategy to account for the initial cold-start period where historical data is sparse.