Custom Metrics & Post-Call Processing

What are Custom Metrics?

Custom metrics allow you to define AI-powered analysis rules that extract specific insights from your calls automatically. Combined with post-call webhooks, you can send this data directly to your systems for further processing.

Key Features

AI-Powered Extraction

AI analyzes call transcripts to extract custom metrics

Multiple Data Types

Text, Numbers, Yes/No, and List data types

Post-Call Webhooks

Send metrics to your external systems automatically

Default Metrics

Standard metrics included by default

How It Works

Configure Metrics

Define custom metrics with AI analysis prompts in agent settings

Call Completes

Dialgen records and transcribes the call

AI Analysis

AI extracts metrics based on your custom prompts

Webhook Sent

Complete call data and all metrics sent to your webhook URL

Process Data

Your system receives and processes the metrics

Setting Up Custom Metrics

Custom metrics are configured exclusively through the Dialgen Dashboard. There is no API endpoint for creating or managing custom metrics. All configuration must be done via the web interface.

Dashboard Configuration

Navigate to your agent settings
Go to the Metrics tab
Scroll to Custom Metrics section
Click Add Custom Metric
Configure:
- Metric Name: What to call this metric
- Data Type: Text, Number, Yes/No, or List
- AI Analysis Prompt: Instructions for AI extraction
- Required: Whether this metric must always have a value
- Description: Optional notes about this metric
Click Create Metric or Update Metric to save

Once configured, your custom metrics will be automatically applied to all calls for that agent.

Configuration Example

{
  "name": "Customer Satisfaction Score",
  "dataType": "number",
  "prompt": "Rate the customer's satisfaction level on a scale of 1-10 based on their tone, responses, and any explicit feedback given during the call.",
  "isRequired": true,
  "description": "Numerical satisfaction score for tracking customer happiness"
}

Data Types

Choose the appropriate data type for your metric:

Text (String)

Store textual responses or extracted information. Example:

{
  "name": "Product Interest",
  "dataType": "string",
  "prompt": "What products or services did the customer express interest in?"
}

API Response:

{
  "product_interest": "Enterprise plan and support services"
}

Number

Store numerical values like scores, ratings, or counts. Example:

{
  "name": "Sentiment Score",
  "dataType": "number",
  "prompt": "Rate the customer's sentiment on a scale of -10 (very negative) to +10 (very positive)",
  "isRequired": true
}

API Response:

{
  "sentiment_score": 7
}

Yes/No (Boolean)

Store true/false values for yes/no questions. Example:

{
  "name": "Callback Needed",
  "dataType": "boolean",
  "prompt": "Did the customer request a callback or follow-up call?"
}

API Response:

{
  "callback_needed": true
}

List (Array)

Store multiple values as a list. Example:

{
  "name": "Mentioned Issues",
  "dataType": "array",
  "prompt": "List all technical issues or problems mentioned by the customer during the call"
}

API Response:

{
  "mentioned_issues": [
    "Login timeout",
    "Mobile app crashes",
    "Password reset issues"
  ]
}

Default Metrics

Dialgen includes standard metrics for every call. These are always included unless explicitly disabled:

Metric	Type	Description
`summary`	string	Overall call summary
`intent`	string	Primary customer intent
`sentiment`	string	Customer sentiment (positive/neutral/negative)
`confidence`	number	AI confidence in the analysis (0-1)
`keyPoints`	array	Key discussion points
`actionItems`	array	Items requiring follow-up action
`callStatus`	string	Call status (SUCCESS, UNRESOLVED, ESCALATION)
`callbackRequired`	string	Whether callback is needed
`promisedDate`	string	Any promised date mentioned
`selfEvaluation`	array	Agent’s self-evaluation of performance
`followUpAction`	string	Recommended follow-up action
`ticketRaised`	string	Support ticket number if created

Disable Default Metrics

You can disable default metrics if you only want custom ones:

In the Metrics tab
Find Default Dialgen Metrics section
Toggle Enable switch to OFF

Disabling default metrics means standard analysis data won’t be generated. Only do this if you have custom metrics that cover the metrics you need.

Post-Call Webhooks

What are Post-Call Webhooks?

Post-call webhooks are HTTP POST requests sent to your server immediately after a call completes. They contain:

Complete call data
All default metrics
All custom metrics with their schema
Recording URL and transcript

The schema of your defined custom metrics is automatically included in every webhook payload so your system knows how to process each metric.

Configuration

Go to Metrics tab in agent settings
Find Post-Call Webhook section
Enter your webhook URL: https://api.yourdomain.com/webhook
Save

Once saved, all future calls will send metrics with the complete schema to this endpoint.

Webhook Payload with Schema

{
  "success": true,
  "callData": {
    "id": "call_789",
    "contactId": "contact_ABC",
    "agentId": "agent_XYZ",
    "batchId": "batch_456",
    "status": "COMPLETED",
    "startTime": "2025-11-16T10:15:30.000Z",
    "endTime": "2025-11-16T10:22:45.000Z",
    "duration": 435,
    "phoneNumber": "+14155552671",
    "contactName": "Sarah Anderson",
    "recordingUrl": "https://recordings.dialgen.ai/call_789.mp3",
    
    "transcription": [
      {
        "role": "assistant",
        "content": "Hello! Thanks for calling TechCorp. How can I assist you today?"
      },
      {
        "role": "user",
        "content": "Hi, I'm interested in your enterprise solution"
      },
      {
        "role": "assistant",
        "content": "Great! Let me tell you about our enterprise features..."
      }
    ],
    
    // Default Dialgen metrics
    "summary": "Customer called regarding enterprise solution pricing and features. Expressed strong interest in features like multi-team collaboration and advanced analytics. Requested pricing details.",
    "intent": "SALES_INQUIRY",
    "sentiment": "positive",
    "confidence": 0.94,
    "keyPoints": [
      "Interested in enterprise plan",
      "Wants multi-team collaboration",
      "Needs advanced analytics"
    ],
    "actionItems": [
      "Send enterprise pricing document",
      "Schedule product demo",
      "Provide ROI calculator"
    ],
    "callStatus": "SUCCESS",
    "callbackRequired": "false",
    "promisedDate": "2025-11-20",
    "followUpAction": "Send pricing and schedule demo",
    
    // Custom metrics with schema
    "customMetrics": {
      "lead_score": {
        "value": 92,
        "schema": {
          "type": "number",
          "name": "Lead Score",
          "description": "Lead quality rating on scale 1-100",
          "isRequired": true
        }
      },
      "use_case": {
        "value": "Multi-team project collaboration platform for software development teams",
        "schema": {
          "type": "string",
          "name": "Use Case",
          "description": "Primary use case mentioned by customer",
          "isRequired": true
        }
      },
      "budget_range": {
        "value": [
          "$5000-10000 annually"
        ],
        "schema": {
          "type": "array",
          "name": "Budget Range",
          "description": "Budget constraints mentioned",
          "isRequired": false
        }
      },
      "product_features_interest": {
        "value": [
          "Multi-team collaboration",
          "Advanced analytics dashboard",
          "API integration",
          "Single sign-on",
          "Audit logs"
        ],
        "schema": {
          "type": "array",
          "name": "Product Features Interest",
          "description": "Features customer expressed interest in",
          "isRequired": false
        }
      },
      "has_budget_approved": {
        "value": true,
        "schema": {
          "type": "boolean",
          "name": "Has Budget Approved",
          "description": "Whether customer has budget approval",
          "isRequired": false
        }
      },
      "competitor_mentioned": {
        "value": "Slack, Monday.com",
        "schema": {
          "type": "string",
          "name": "Competitor Mentioned",
          "description": "Competing solutions mentioned",
          "isRequired": false
        }
      }
    }
  },
  
  // Schema definition (sent once per webhook)
  "metricSchema": {
    "lead_score": {
      "type": "number",
      "name": "Lead Score",
      "description": "Lead quality rating on scale 1-100",
      "isRequired": true
    },
    "use_case": {
      "type": "string",
      "name": "Use Case",
      "description": "Primary use case mentioned by customer",
      "isRequired": true
    },
    "budget_range": {
      "type": "array",
      "name": "Budget Range",
      "description": "Budget constraints mentioned",
      "isRequired": false
    },
    "product_features_interest": {
      "type": "array",
      "name": "Product Features Interest",
      "description": "Features customer expressed interest in",
      "isRequired": false
    },
    "has_budget_approved": {
      "type": "boolean",
      "name": "Has Budget Approved",
      "description": "Whether customer has budget approval",
      "isRequired": false
    },
    "competitor_mentioned": {
      "type": "string",
      "name": "Competitor Mentioned",
      "description": "Competing solutions mentioned",
      "isRequired": false
    }
  }
}

The metricSchema field contains the complete definition of all custom metrics, including their type, description, and whether they’re required. This allows your system to properly validate and process the custom metric values.

Processing Webhooks

Node.js Example

const express = require('express');
const app = express();

app.use(express.json());

app.post('/webhook', async (req, res) => {
  const { callData, metricSchema } = req.body;
  
  try {
    // Extract custom metrics
    const metrics = callData.customMetrics || {};
    
    console.log('Call Analysis:');
    console.log(`- Lead Score: ${metrics.lead_score?.value}`);
    console.log(`- Use Case: ${metrics.use_case?.value}`);
    console.log(`- Has Budget: ${metrics.has_budget_approved?.value}`);
    console.log(`- Features Interest: ${metrics.product_features_interest?.value?.join(", ")}`);
    
    // Use metricSchema to validate and process metrics
    Object.entries(metricSchema).forEach(([key, schema]) => {
      const metricValue = metrics[key]?.value;
      const isRequired = schema.isRequired;
      
      if (isRequired && !metricValue) {
        console.warn(`Warning: Required metric "${schema.name}" is missing`);
      }
    });
    
    // Update your CRM with extracted metrics
    await crm.updateContact({
      id: callData.contactId,
      lastCallDate: callData.startTime,
      leadScore: metrics.lead_score?.value,
      useCase: metrics.use_case?.value,
      budgetApproved: metrics.has_budget_approved?.value,
      featuresInterested: metrics.product_features_interest?.value
    });
    
    // Create sales opportunity if lead score is high
    if (metrics.lead_score?.value >= 80) {
      await crm.createOpportunity({
        contactId: callData.contactId,
        title: `Enterprise Sales: ${metrics.use_case?.value}`,
        value: 50000, // Estimated based on budget range
        stage: 'DISCOVERY'
      });
    }
    
    // Schedule follow-up if callback requested (if you had such metric)
    if (metrics.has_budget_approved?.value) {
      await scheduling.createTask({
        contactId: callData.contactId,
        type: 'PRODUCT_DEMO',
        dueDate: '2025-11-20',
        priority: 'HIGH'
      });
    }
    
    res.status(200).json({ success: true });
  } catch (error) {
    console.error('Webhook processing error:', error);
    res.status(200).json({ success: true }); // Still return 200 to prevent retries
  }
});

app.listen(3000);

Python Example

from flask import Flask, request, jsonify
from datetime import datetime, timedelta

app = Flask(__name__)

@app.route('/webhook', methods=['POST'])
def webhook():
    data = request.json
    call_data = data.get('callData', {})
    metrics = call_data.get('customMetrics', {})
    metric_schema = data.get('metricSchema', {})
    
    print('Call Analysis:')
    print(f"- Lead Score: {metrics.get('lead_score', {}).get('value')}")
    print(f"- Use Case: {metrics.get('use_case', {}).get('value')}")
    print(f"- Has Budget: {metrics.get('has_budget_approved', {}).get('value')}")
    features = metrics.get('product_features_interest', {}).get('value', [])
    print(f"- Features Interest: {', '.join(features)}")
    
    # Validate required metrics using schema
    for key, schema in metric_schema.items():
        metric_value = metrics.get(key, {}).get('value')
        is_required = schema.get('isRequired', False)
        
        if is_required and not metric_value:
            print(f"Warning: Required metric '{schema.get('name')}' is missing")
    
    # Update CRM
    crm.update_contact({
        'id': call_data.get('contactId'),
        'last_call': call_data.get('startTime'),
        'lead_score': metrics.get('lead_score', {}).get('value'),
        'use_case': metrics.get('use_case', {}).get('value'),
        'budget_approved': metrics.get('has_budget_approved', {}).get('value'),
        'features_interested': metrics.get('product_features_interest', {}).get('value')
    })
    
    # Create opportunity if high lead score
    lead_score = metrics.get('lead_score', {}).get('value', 0)
    if lead_score >= 80:
        crm.create_opportunity({
            'contact_id': call_data.get('contactId'),
            'title': f"Enterprise Sales: {metrics.get('use_case', {}).get('value')}",
            'value': 50000,
            'stage': 'DISCOVERY'
        })
    
    # Schedule demo if budget is approved
    if metrics.get('has_budget_approved', {}).get('value'):
        scheduling.create_task({
            'contact_id': call_data.get('contactId'),
            'type': 'PRODUCT_DEMO',
            'due_date': (datetime.now() + timedelta(days=4)).isoformat(),
            'priority': 'HIGH'
        })
    
    return jsonify({'success': True}), 200

if __name__ == '__main__':
    app.run(port=3000)

Best Practices

Writing Effective Prompts

❌ Vague:

Extract the customer's feedback

✅ Specific:

Did the customer express satisfaction with the service? If yes, rate their satisfaction level 1-10. If no, explain why they were dissatisfied.

Choose Right Data Types

Metric Type	Best Data Type
Scores or ratings	Number
Yes/No decisions	Boolean
Multiple values	Array
Descriptions or feedback	String

Required Fields

Mark metrics as “Required” only when:

The metric must always have a value
Missing values would cause problems in your system
You want to validate quality

Testing Metrics

Create test agents with your metrics
Run test calls with various scenarios
Review webhook payloads to verify accuracy
Adjust prompts based on results
Deploy when confident

Security

✓ Validate webhook signatures if possible
✓ Use HTTPS URLs only
✓ Implement rate limiting
✓ Log all webhook activity
✓ Handle errors gracefully

Common Use Cases

Lead Scoring

{
  "name": "Lead Score",
  "dataType": "number",
  "prompt": "Rate the lead quality on a scale of 1-100 based on their interest level, budget clarity, and buying timeline.",
  "isRequired": true
}

Sentiment Tracking

{
  "name": "Customer Emotion",
  "dataType": "string",
  "prompt": "Describe the customer's emotional state during the call (happy, frustrated, neutral, angry, confused)"
}

Issue Documentation

{
  "name": "Technical Issues",
  "dataType": "array",
  "prompt": "List each technical issue or bug the customer reported"
}

Follow-up Actions

{
  "name": "Required Follow-ups",
  "dataType": "array",
  "prompt": "What specific actions must be taken before the next customer contact?"
}

Troubleshooting

Webhook Not Received

Check:

URL is correct and publicly accessible
Server is responding with 2xx status code
Firewall isn’t blocking incoming requests
Check Dialgen logs for errors

Metrics Empty or Incorrect

Solutions:

Review and clarify your AI prompt
Provide examples in the prompt
Use simpler, more direct language
Test with different call scenarios

High CPU Usage

Optimization:

Reduce number of custom metrics
Simplify AI prompts
Disable default metrics if not needed
Increase webhook processing interval

Retrieving Metrics via API

In addition to webhooks, you can retrieve custom metrics by calling the get-call-metric endpoint. The endpoint will return the complete schema definition for all custom metrics configured for that agent.

Get Call Metrics Endpoint

curl --location 'https://sa.dialgen.ai/api/v1/call/get-metric?callId=call_789&userId=user_456&agentId=agent_XYZ&contactId=contact_ABC' \
  -H 'Authorization: Bearer YOUR_API_KEY'

Response with Custom Metric Schema

{
  "success": true,
  "callData": {
    "id": "call_789",
    "status": "COMPLETED",
    "duration": 435,
    "recordingUrl": "https://recordings.dialgen.ai/call_789.mp3",
    
    // Default metrics...
    "summary": "Customer inquiry about enterprise solution...",
    "intent": "SALES_INQUIRY",
    "sentiment": "positive",
    
    // Custom metrics with schema
    "customMetrics": {
      "lead_score": {
        "value": 92,
        "schema": {
          "type": "number",
          "name": "Lead Score",
          "description": "Lead quality rating on scale 1-100",
          "isRequired": true
        }
      },
      "use_case": {
        "value": "Multi-team project collaboration",
        "schema": {
          "type": "string",
          "name": "Use Case",
          "description": "Primary use case mentioned by customer",
          "isRequired": true
        }
      },
      "product_features_interest": {
        "value": ["Multi-team collaboration", "Advanced analytics", "API integration"],
        "schema": {
          "type": "array",
          "name": "Product Features Interest",
          "description": "Features customer expressed interest in",
          "isRequired": false
        }
      },
      "has_budget_approved": {
        "value": true,
        "schema": {
          "type": "boolean",
          "name": "Has Budget Approved",
          "description": "Whether customer has budget approval",
          "isRequired": false
        }
      }
    }
  },
  
  "metricSchema": {
    "lead_score": {
      "type": "number",
      "name": "Lead Score",
      "description": "Lead quality rating on scale 1-100",
      "isRequired": true
    },
    "use_case": {
      "type": "string",
      "name": "Use Case",
      "description": "Primary use case mentioned by customer",
      "isRequired": true
    },
    "product_features_interest": {
      "type": "array",
      "name": "Product Features Interest",
      "description": "Features customer expressed interest in",
      "isRequired": false
    },
    "has_budget_approved": {
      "type": "boolean",
      "name": "Has Budget Approved",
      "description": "Whether customer has budget approval",
      "isRequired": false
    }
  }
}

Both the get-call-metric endpoint and the onCallComplete webhook return the complete metric schema, allowing your application to properly understand and validate each custom metric’s structure.

Next Steps

Agent Configuration

Learn how to configure agents

API Reference

Complete API documentation

Call Recordings

Access call data and recordings

​What are Custom Metrics?

​Key Features

AI-Powered Extraction

Multiple Data Types

Post-Call Webhooks

Default Metrics

​How It Works

​Setting Up Custom Metrics

​Dashboard Configuration

​Configuration Example

​Data Types

​Text (String)

​Number

​Yes/No (Boolean)

​List (Array)

​Default Metrics

​Disable Default Metrics

​Post-Call Webhooks

​What are Post-Call Webhooks?

​Configuration

​Webhook Payload with Schema

​Processing Webhooks

​Node.js Example

​Python Example

​Best Practices

​Writing Effective Prompts

​Choose Right Data Types

​Required Fields

​Testing Metrics

​Security

​Common Use Cases

​Lead Scoring

​Sentiment Tracking

​Issue Documentation

​Follow-up Actions

​Troubleshooting

​Webhook Not Received

​Metrics Empty or Incorrect

​High CPU Usage

​Retrieving Metrics via API

​Get Call Metrics Endpoint

​Response with Custom Metric Schema

​Next Steps

Agent Configuration

API Reference

Call Recordings

What are Custom Metrics?

Key Features

How It Works

Setting Up Custom Metrics

Dashboard Configuration

Configuration Example

Data Types

Text (String)

Number

Yes/No (Boolean)

List (Array)

Default Metrics

Disable Default Metrics

Post-Call Webhooks

What are Post-Call Webhooks?

Configuration

Webhook Payload with Schema

Processing Webhooks

Node.js Example

Python Example

Best Practices

Writing Effective Prompts

Choose Right Data Types

Required Fields

Testing Metrics

Security

Common Use Cases

Lead Scoring

Sentiment Tracking

Issue Documentation

Follow-up Actions

Troubleshooting

Webhook Not Received

Metrics Empty or Incorrect

High CPU Usage

Retrieving Metrics via API

Get Call Metrics Endpoint

Response with Custom Metric Schema

Next Steps