Trying to understand the discrepancy between real-time Bot Performance metrics displayed in the Performance Dashboard and the aggregated data found in the Conversation Detail views within our EU-West BYOC environment. The organization relies heavily on these dashboards for SLA monitoring, and a significant variance is causing reporting inconsistencies for stakeholders.
The specific issue involves the “Deflection Rate” metric. In the real-time dashboard, the deflection rate for our primary customer service bot is reported at 42%. However, when reviewing the historical data for the same 24-hour period via the Reporting module or by manually sampling Conversation Detail records, the calculated deflection rate drops to approximately 38%. This 4% variance is material for our quarterly performance reviews.
The environment is Genesys Cloud EU-West BYOC, running the latest stable release. The Architect flow in question uses standard NLP intents with no custom webhook logic that would alter conversation status. The bot is configured to hand off to an agent if confidence scores fall below 0.7. We have verified that the “Bot Deflection” tag is correctly applied in the flow logic upon successful intent resolution without handoff.
- Reviewed the Architect flow logs for the last 48 hours to confirm that the “Deflected” status is being set correctly at the decision node prior to the end of the conversation. No anomalies were found in the flow execution paths.
- Cross-referenced the Conversation Detail export with the Performance Dashboard data for a specific high-volume queue over a 1-hour window. The manual calculation based on conversation outcomes (Bot Only vs. Bot + Agent) consistently shows a lower deflection rate than the dashboard widget.
Is there a known latency or aggregation logic difference between the real-time performance widgets and the historical reporting engine? The documentation suggests these should align, but the persistent gap indicates a potential data synchronization issue or a difference in how “deflection” is defined across these two views. Any insights into the underlying calculation methodology would be appreciated to reconcile these figures for the upcoming audit.