Analytics
Beyond delivery metrics — see how your AI bot actually performs. Conversation success rate, drop-off points, tool call accuracy, customer satisfaction, handover reasons. Optimize from data, not guesses.
Bot metrics
Every angle covered
Drop-off analysis
Find weak nodes
Success rate
Optimize APIs
Insights
Pattern detection
What % of conversations end in the desired outcome (booking, lead captured, sale closed)? Trends over time.
At which message does the customer disengage? Identify weak steps; iterate on copy or logic.
Per-tool success rate, latency, error patterns. Identify slow APIs or broken integrations.
% of customer questions where bot found relevant context vs hallucinated. Indicator of KB quality.
Why does the bot escalate? Reasons logged: explicit request, can't answer, tool failure, sentiment. Optimize the most common reasons.
Post-conversation CSAT survey (optional). Average score per bot, per agent, per use case. Tie satisfaction to conversion.
Drop-off report shows 40% of users abandon at step 3 (asking for SSN). Test: replace 'SSN' with 'tax ID' — drop-off down to 12%.
Conversation completion up 28%
Tool call analytics show order_lookup taking avg 12 seconds. Investigate, optimize the API, latency down to 800ms.
Customer perceived speed dramatically improved
KB hit rate at 65% — for 35% of questions, bot couldn't find relevant context and either hallucinated or escalated. Add missing docs; hit rate up to 92%.
Escalation rate down from 28% to 9%
Handover analysis shows 'customer requested human' as #1 reason — but happens after 8+ messages. Investigation: bot wasn't acknowledging frustration. Improve sentiment handling → handover reasons shift.
Better customer experience
Free workspace, bot analytics included from day one.