Why does this setting trigger compliance gaps in Architect flow data retention?

Context:
Our organization operates within the EU-West BYOC environment, adhering to strict GDPR and local data sovereignty requirements. We have recently updated our Architect flows to include specific data masking actions for sensitive customer information (PII) before logging conversation details. The intention was to ensure that the Performance Dashboard and associated conversation detail views reflect only sanitized data, thereby aligning with our internal security compliance standards.

However, since implementing these changes, we have observed a discrepancy in the data retention metrics displayed in the Performance Dashboard. Specifically, conversations routed through the updated Architect flows are showing incomplete metadata in the ‘Conversation Detail’ view, while the queue activity metrics remain accurate. This suggests that the data masking action within the flow is inadvertently stripping necessary operational metadata required for the dashboard to render full performance analytics, rather than just masking PII.

We are not seeing any explicit error codes in the standard system logs, but the analytics query results for these specific flow paths return null values for fields such as ‘handle_time’ and ‘wrap_up_code’ in certain edge cases. This is causing significant confusion for our team leads, who rely on these metrics for agent performance reviews. The issue appears to be isolated to flows that utilize the ‘Set Variable’ action with complex JSON payloads for masking, rather than simple string replacements.

Question:
Why does the implementation of data masking actions within Architect flows cause the Performance Dashboard to omit critical operational metadata in the EU-West BYOC environment? Is there a specific configuration or limitation in how Genesys Cloud handles data transformation actions that impacts downstream analytics reporting? We need to understand if this is a known behavior or if our flow design is inadvertently conflicting with the platform’s data indexing mechanisms for compliance views.