What You’ll Learn
- What Metrics Lab is and when to use it
- How to prototype and test evaluation prompts
- How to promote metrics from experimentation to production
How Metrics Lab Works
You use Metrics Lab to prototype measurement ideas, compare scoring approaches, and mature Custom Metrics before you depend on them across simulations, alerts, and reporting workflows. Metrics Lab provides an interactive environment where you write evaluation prompts, test them against sample transcripts, and iterate on scoring criteria. Once a metric performs reliably, you promote it to production where it runs automatically across simulations and observability evaluations.Key Capabilities
- Interactive prototyping — write and test evaluation prompts against sample transcripts in real time
- Side-by-side comparison — compare different scoring approaches to find the most reliable one
- Safe experimentation — iterate without affecting production data or live evaluations
- One-click promotion — move a validated metric directly into your simulation and observability workflows
Common Use Cases
- Draft a new “compliance check” metric and test it against 10 sample transcripts before deploying
- Compare two different prompts for scoring empathy to see which produces more consistent results
- Validate that a formula metric correctly computes a composite score before attaching it to alerts
Next Steps
Custom Metrics
Learn about the metrics that Metrics Lab helps you create.
Create Custom Metrics API
Create metrics programmatically via the API.