What is the correct way to reconcile BYOC edge latency spikes in Performance dashboards?

What is the standard approach to isolate edge network latency from internal queue metrics when using a Bring Your Own Cloud (BYOC) configuration? Our organization operates in the EU1 region with a dedicated edge deployment. Recent audits of the Agent Performance view indicate anomalous increases in After Call Work and Talk Time for agents routed through the custom edge, yet the Conversation Detail view does not reflect corresponding network timeout errors or dropped packets.

The discrepancy suggests that the latency is being absorbed into the conversation metrics rather than flagged as a system failure. We are monitoring via the standard Performance dashboard and have verified that the Edge Status widget reports Healthy throughout the affected period. The issue persists across multiple queues utilizing the same Architect flow, which employs standard Add to Queue actions without complex logic or external API integrations that could introduce variable delays.

Is there a specific metric or log endpoint that correlates BYOC edge processing time with the agent-facing metrics? The current data makes it difficult to determine if the degradation is due to edge configuration, upstream carrier issues, or Genesys Cloud internal routing. We require a definitive method to attribute these time deltas to the edge infrastructure for our SLA reporting.

Make sure you isolate the edge latency from WFM metrics. The dashboard likely conflates network jitter with ACW. Check the telephony logs for timing discrepancies.

{
 "error": "Metric mismatch: Edge latency not excluded from Agent Performance calculation."
}

Adjust the reporting filters to exclude edge-specific timestamps.

The root of the issue is that the performance dashboard does not automatically strip out network jitter for byoc edges. when you see those acw spikes, it is often because the recording metadata includes the full duration from the sip dialog start, which captures the edge latency. to fix this for your legal hold exports, you need to query the recording api (v1) directly and filter by the media_type of screen or voice while excluding the edge_id from the timing calculation. use the bulk export job with a custom filter for conversation_id and check the created_date against the last_updated_date to find the delta. this delta is your actual network latency. do not rely on the standard agent view for chain of custody evidence. always verify the timestamps in the raw json response from the api to ensure the audit trail is accurate.