Skip to main content

​
Implementing User Feedback

You can add feedback to traces using Maxim’s SDK across multiple programming languages:
  • JavaScript/TypeScript: trace.feedback({ score: 5, comment: "Great job!" })
  • Python: trace.feedback(score=5, comment="Great job!")
  • Go and Java: Similar methods available through respective SDKs
The feedback structure supports:
  • Numerical scores (typically 1-5 or custom rating scales)
  • Optional text comments for qualitative insights
  • Association with specific traces for granular analysis

​
Leveraging Feedback Data

Once feedback is collected, Maxim provides powerful analysis capabilities:
  • View feedback directly in trace details alongside technical metrics like latency, token usage, and cost
  • Filter and search traces based on user satisfaction scores to identify patterns
  • Track average user feedback over time through the Overview tab’s aggregated metrics
  • Correlate user satisfaction with specific prompt versions, model choices, or system configurations
  • Set up alerts when feedback scores drop below acceptable thresholds
This feedback loop enables data-driven improvements to your AI applications, helping you prioritize optimization efforts based on actual user experience rather than technical metrics alone.