Discrepancy in Conversation Entity Counts Between Conversations API and Analytics API

What’s the best way to reconcile the entity count differences when querying active sessions via the Genesys Cloud Platform APIs?

I am maintaining a comprehensive Postman collection for validating our contact center’s real-time state against historical reports. I have a Newman CLI script that runs hourly to audit data integrity. The script performs two parallel requests:

  1. GET /api/v2/conversations?active=true
  2. GET /api/v2/analytics/conversations/summary?interval=real-time&metricTypes=interactionCount

The issue is that the response from the Conversations API returns a list of conversation objects with a total count of 142. However, the Analytics API response shows an interactionCount of 138 for the same exact second.

Here is the relevant JSON snippet from the Analytics response:

{
 "data": [
 {
 "metricTypes": ["interactionCount"],
 "interval": {
 "start": "2023-10-27T14:00:00.000Z",
 "end": "2023-10-27T14:00:01.000Z"
 },
 "values": [
 {
 "metricType": "interactionCount",
 "value": 138
 }
 ]
 }
 ]
}

My pre-request script logs the timestamp to millisecond precision. Both requests are fired within a 50ms window. I suspect the Analytics API is aggregating data from a different source or has a slight latency in the real-time buffer, but the documentation implies both should reflect the current state of the platform.

Is there a specific filter or query parameter I am missing on the /api/v2/conversations endpoint to exclude certain types of interactions (like internal-only or test conversations) that might be inflating the count? Or is the Analytics API excluding completed-but-not-yet-archived conversations?

I need a deterministic way to validate that my monitoring dashboard (which uses the Conversations API) matches the reporting layer (which uses the Analytics API). The 4% variance is triggering false positives in our alerting pipeline.

Thanks for the help.