Agent Analytics metrics and optimization
What each Agent Analytics metric measures (deflection, escalation, abandonment, engagement, success, task resolution), how LLM-judged quality scores point to retrieval or response problems, how intents are extracted and clustered, and the loop that turns a weak subagent or intent cluster into a measured fix.
Explain agent analytics and agent optimization.
~6 min read · 16 practice questions ·
What the exam expects
- Define the effectiveness metrics (deflection, escalation, abandonment, engagement, success, task resolution) and read them together.
- Interpret the quality score and its dimensions (answer faithfulness, answer relevance, context relevance) and know they are judged by an LLM.
- Explain intents, the intent and clustering pipelines and their timing.
- Use Performance Insights and quality score reasoning to find the subagent or intent behind poor results.
- Choose the optimization action (knowledge gaps, overlapping articles, unclear instructions) and re-measure.
Key terms: deflection rate is the share of sessions that ended without escalation; task resolution rate is the share judged fully resolved by the Trust Layer Task Resolution detector; answer faithfulness measures how closely a response sticks to grounded source data; context relevance measures whether the right source was retrieved; intent is what the user was trying to accomplish within a set of turns.
Key facts
- Detail
- Default 2 hours
- Detail
- 1–5, from an LLM evaluator
- Detail
- Faithfulness, answer relevance and context relevance are 0–1
- Detail
- Audit and Feedback plus Knowledge/RAG Quality Data and Metrics
- Detail
- About 4–5 hours after a session closes
- Detail
- Weekly; tag needs 10+ similar intents
- Detail
- Performance Insights by subagent, intent or action
- Detail
- Built on the Session Tracing Data Model
- Why can an agent show high deflection and still leave customers unresolved?
- When will a new cluster tag appear for a brand-new kind of request?
- Which field explains why a session got a low quality rating?
- Which three dimensions make up the quality view besides the overall score?
The questions are free to read. Answers with their sources and self-grading are included with the pack.
Keep reading with the Agentforce Specialist pack
- Core concepts
- Exam traps
- Worked scenario
The pack includes:
- All 26 lessons of the Agentforce Specialist guide, with reading progress
- Interactive decision walks and step-by-step sequence steppers
- Inline practice questions and one-click topic practice (16 questions for this lesson)
- An official source cited for every claim
- An AI tutor that answers from the guide
- Unlimited timed practice exams