Context:
We operate 15 BYOC trunks across APAC regions, handling significant call volumes. Recently, we observed a discrepancy in our predictive routing analytics. Calls routed via trunks with average latency exceeding 200ms are showing a 15% drop in successful connection rates compared to low-latency trunks, despite identical IVR flow configurations in Architect. The SIP registration status remains stable (200 OK), and carrier failover logic is not triggering. However, the predictive routing engine appears to be misinterpreting the increased jitter as agent unavailability or poor quality, leading to premature call drops or misrouting to backup queues. We have verified that the SIP credentials and outbound routing rules are consistent across all trunks. The issue persists even after adjusting the jitter buffer settings in the trunk configuration. We suspect the predictive routing weight calculation might be overly sensitive to latency spikes specific to certain carriers in the APAC region.
Question:
What is the correct way to adjust the predictive routing quality thresholds to account for inherent carrier latency without degrading the overall routing efficiency? Are there specific API endpoints or Architect flow variables that can be tuned to decouple latency metrics from the predictive routing success score calculation?