Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.duckie.ai/llms.txt

Use this file to discover all available pages before exploring further.

Performance metrics help you measure how well your agents are handling customer conversations.

​
Key Metrics

​
Deflection Rate

Percentage of conversations handled without human intervention. 
Deflection Rate = (AI Resolved / Total Conversations) × 100
What it means:
  • Higher = more automation success
  • Target varies by use case (typically 40-80%)
How to improve:
  • Add more knowledge
  • Improve agent configuration
  • Fill knowledge gaps

​
Resolution Rate

Percentage of conversations successfully resolved (per your resolution rules). 
Resolution Rate = (Resolved / Total Conversations) × 100
What it means:
  • Higher = better customer outcomes
  • Depends on resolution rule definition
How to improve:
  • Review resolution rules for accuracy
  • Improve response quality
  • Address common failure patterns

​
Escalation Rate

Percentage of conversations escalated to humans. 
Escalation Rate = (Escalated / Total Conversations) × 100
What it means:
  • Lower is often better, but some escalation is healthy
  • Too low might mean missing issues that need humans
How to improve:
  • Tune escalation rules
  • Add knowledge for common escalation topics
  • Review if escalations are appropriate

​
Response Time

Time from customer message to agent response. Metrics:
  • P50 (Median) — Half of responses are faster
  • P95 — 95% of responses are faster
  • Average — Mean response time
Target: Most responses under a few seconds

​
Time to Resolution

Total time from first message to resolution. Metrics:
  • P50, P95, Average
  • By category
  • Trend over time

​
Viewing Metrics

1

Navigate to Performance

Go to Analyze → Performance.
2

Select Date Range

Choose the time period to analyze.
3

Filter by Deployment

Optionally filter by specific deployment.
4

Review Metrics

See metric cards with current values and trends.

​
Trend Charts

See how metrics change over time:
  • Daily, weekly, or monthly views
  • Compare to previous periods
  • Identify trends and anomalies

​
Filtering

​
By Date Range

  • Last 7 days
  • Last 30 days
  • Last 90 days
  • Custom range

​
By Deployment

View metrics for specific deployments:
  • All deployments
  • Individual deployment
  • Compare deployments

​
By Category

See performance per category:
  • Which categories resolve best?
  • Where does escalation happen most?

​
Benchmarks

​
Comparing Periods

See change from previous period:
  • ↑ 5% Deflection rate improved
  • ↓ 2% Escalation rate decreased

​
By Deployment

Compare deployments side-by-side:
MetricZendeskSlackIntercom
Deflection65%72%58%
Resolution82%85%78%
Response Time2.1s1.8s2.5s

​
Taking Action

​
When Deflection is Low

Causes:
  • Missing knowledge
  • Poor agent configuration
  • Complex issues
Actions:
  • Review knowledge gaps
  • Improve runbooks
  • Add more documentation

​
When Escalation is High

Causes:
  • Overly sensitive guardrails
  • Missing knowledge
  • Complex customer base
Actions:
  • Tune escalation rules
  • Add knowledge for escalation topics
  • Review escalation patterns

​
When Response Time is High

Causes:
  • Complex processing
  • Slow external APIs
  • Large knowledge searches
Actions:
  • Optimize runbooks/workflows
  • Check tool performance
  • Review execution logs

​
Exporting Data

Export metrics for external analysis:
  • CSV download

​
Next Steps

Breakdown

See distribution by category

Run History

Investigate individual runs