The number of unanswered questions → Generated answer rate and quality; The number of abandoned sessions → Conversation outcomes
Why “Generated answer rate and quality” is correct
The requirement is to identify the number of questions unanswered by the agent . In Copilot Studio, unanswered-question behavior is tied to how often the agent successfully generates answers and the quality of those answers.
The Generated answer rate and quality insight is the right place to evaluate whether the agent is:
answering user questions
failing to generate answers
producing low-quality responses
missing knowledge coverage
From an AI business solutions standpoint, unanswered questions are a direct signal of knowledge gaps, grounding gaps, or prompt-response weaknesses. This is exactly what generated answer analytics are meant to surface.
Why “Conversation outcomes” is correct
The requirement also asks for the number of abandoned sessions between users and the agent.
Abandonment is a conversation-level outcome , not a reaction or survey result. The Conversation outcomes insight tracks what happened to the interaction, such as whether the conversation was:
That makes it the correct metric for identifying abandoned sessions.
Why the other insights are not correct
Reactions
Reactions capture signals like positive or negative user feedback, but they do not directly measure unanswered questions or abandoned sessions.
Survey results
Survey results reflect user satisfaction feedback after interactions, but they do not directly quantify unanswered questions or abandonment counts.
Submit