How to Set Up Call Evaluations
Call Evaluations allow you to automatically measure the success of your calls based on specific criteria you define. Instead of manually listening to every recording, our AI analyzes the conversation transcript against your requirements and provides an objective success rating immediately after the call ends.
Why use Call Evaluations?
Track conversion rates objectively: Get clear data on how many calls are actually meeting your goals.
Identify what works and what doesn't: Spot patterns in successful vs. failed conversations.
No manual review needed: Save hours by letting AI handle the initial screening of your calls.
Data-driven improvement: Use the results to refine your agent instructions and sales scripts.
How to Access Evaluations
To set up evaluations for your agent:
Go to the Agents section in the sidebar.
Select the Agent you want to configure.
Click on the Evaluations tab.
Adding an Evaluation Criterion
Click the "Add Criteria" button.
Enter Name: Give your criterion a clear name (e.g., "Client Ready to Buy").
Write Evaluation Prompt: Define exactly what the AI should look for to determine Success, Failure, or Unknown.
Click "Add Criteria" to save.
Writing Good Evaluation Prompts
A good prompt provides clear boundaries for the AI. Always define:
SUCCESS conditions: What specific words or outcomes mean the goal was met?
FAILURE conditions: What indicates the goal was definitely not met?
UNKNOWN conditions: When should the AI stay neutral (e.g., the topic wasn't mentioned)?
Built-in Examples
You can use these patterns for your own criteria:
Criterion Name | Evaluation Prompt Logic |
Client Ready to Buy | Success: Explicit agreement, scheduling next steps, or price acceptance. |
Budget Qualified | Success: Budget range confirmed (e.g., $5k-$15k), no major price objections. |
Appointment Set | Success: Specific date and time confirmed, address or contact info verified. |
Understanding Results
After each call, each criterion will be marked with one of three statuses:
✅ Success — The criterion was clearly met during the conversation.
❌ Failure — The criterion was clearly not met or was rejected.
❓ Unknown — The topic wasn't discussed or the outcome was ambiguous.
Where to View Results
Evaluation results are available in two places:
Call Details Page: See the status and the AI's rationale for each criterion.
Webhook Payload: Results are sent to your integrated CRM or analytics tools automatically.
Tips for Success
Start small: Begin with the 2-3 most important metrics for your business.
Be specific: Instead of "was the client nice," use "did the client confirm their email address."
Use examples: Include specific phrases in your prompts that indicate success or failure.
Quick Reference
Evaluations turn subjective conversations into objective data. Set them up once, and let the AI do the heavy lifting of measuring your agent's performance.