Skip to main content
What you’ll learn: How to create precise AI-powered questions that automatically analyze your calls and extract valuable business insights from every conversation.
Critical: Make your insights questions as precise as possible. Don’t just ask “if the customer is interested” - ask very specifically what you need to know. Precise questions generate accurate, actionable insights.

​
What Are Insights?

Insights are intelligent questions that you feed into the Vodex system. After every call, Vodex’s AI automatically generates answers to these questions based on the conversation content, providing you with structured data and valuable business intelligence.

​
Key Benefits

  • Automated Analysis: No manual call review needed
  • Structured Data: Get consistent, queryable results
  • Business Intelligence: Track trends, conversion rates, and customer behavior
  • Webhook Integration: Automatically receive insights via API
  • Campaign Optimization: Use data to improve future campaigns

​
How Insights Work

1

Create Insights Questions

Before Running Campaigns
  • Define specific questions you want answered
  • Set up insights at the campaign level
  • Configure via API or dashboard interface
2

Run Your Campaign

During Calls
  • AI listens to the entire conversation
  • Records all relevant information
  • Prepares data for analysis
3

AI Analysis

After Each Call
  • AI analyzes the complete conversation
  • Answers your predefined questions
  • Generates structured insights data
4

Receive Results

Get Your Insights
  • Access via dashboard
  • Receive via webhook (if configured)
  • Use data for business decisions
Important: Insights must be created before you run the campaign. You cannot add insights to calls that have already completed.

​
Insights Configuration

​
Three Required Fields

Every insight requires these three components:
  • Name
  • Type
  • Question
Programming VariableThe technical identifier for your insight, used in webhooks and API responses.Rules:
  • No spaces allowed
  • No dashes allowed
  • Only underscores for separation
  • Use descriptive, clear names
Examples:
✅ customer_interested
✅ payment_timeline
✅ product_requirements
✅ appointment_confirmed

❌ customer interested (spaces)
❌ customer-interested (dashes)
❌ interested (too vague)

​
Insights Examples

​
Example 1: Purchase Intent

Customer Purchase Agreement

Name: customer_interestedType: True/FalseQuestion: “Did the customer agree verbally that they are going to purchase the product? Return True or False. If the customer does not have anything definitive or seems uncertain, return N/A”Use Case: Track conversion rates and sales pipeline

​
Example 2: Payment Timeline

Payment Schedule

Name: num_daysType: NumberQuestion: “In how many days did the customer say they are going to make the payment? Return only the number of days mentioned. If no specific timeline was discussed or the customer was vague, return N/A”Use Case: Cash flow planning and follow-up scheduling

​
Example 3: Customer Requirements

Product Requirements

Name: requirementsType: TextQuestion: “What are the customer’s specific requirements for the product? Include any features, specifications, or conditions they mentioned. If no specific requirements were discussed, return N/A”Use Case: Product customization and service delivery

​
Example 4: Appointment Confirmation

Appointment Booking

Name: appointment_confirmedType: True/FalseQuestion: “Did the customer confirm a specific appointment date and time? Return True only if both date and time were agreed upon. If only tentative or partial scheduling occurred, return False. If no appointment was discussed, return N/A”Use Case: Schedule management and confirmation tracking

​
Example 5: Budget Information

Customer Budget

Name: customer_budgetType: NumberQuestion: “What budget amount did the customer mention they are willing to spend? Return only the numerical value without currency symbols. If they mentioned a range, return the higher number. If no budget was discussed, return N/A”Use Case: Pricing strategy and proposal preparation

​
Advanced Insights Strategies

​
Precision Guidelines

Instead of vague questions, define exact conditions:❌ Vague: “Is the customer happy?”✅ Specific: “Did the customer express satisfaction with our service using positive language such as ‘great’, ‘excellent’, ‘satisfied’, or similar positive terms? Return True only for clearly positive feedback, False for negative feedback, and N/A for neutral or unclear responses.”
Always provide instructions for unclear situations:✅ Good Example: “What is the customer’s company size? Return the number of employees if mentioned. If they say ‘small business’ return 10, if ‘medium business’ return 100, if ‘large business’ return 500. If no size information is provided, return N/A.”
Be explicit about what constitutes different responses:✅ Clear Criteria: “Did the customer agree to a follow-up call? Return True only if they agreed to a specific time or said ‘yes’ to scheduling. Return False if they declined or said they would call back. Return N/A if follow-up was not discussed.”

​
Multi-Layered Insights

Create complementary insights that provide deeper analysis: 
Primary: customer_interested (True/False)
Secondary: interest_level (Number 1-10)
Supporting: objections_mentioned (Text)
Timeline: decision_timeframe (Number - days)

​
Implementation Methods

​
Method 1: Campaign-Level Configuration

1

Access Campaign Settings

Navigate to Your Campaign
  • Open your campaign in the Vodex dashboard
  • Go to the Insights section
  • Click “Add New Insight”
2

Configure Insights

Set Up Each Insight
  • Enter the Name (programming variable)
  • Select the Type (Text/Number/True-False)
  • Write the precise Question
  • Save the configuration
3

Test and Validate

Before Going Live
  • Run test calls to verify insights work
  • Check that questions are answered accurately
  • Adjust questions if needed

​
Method 2: API Integration

{

    ... other params
  "campaign_id": "your_campaign_id",
  "insights": [
    {
      "name": "customer_interested",
      "type": "boolean",
      "question": "Did the customer agree verbally that they are going to purchase the product? Return True or False. If the customer does not commit definitively, return N/A"
    },
    {
      "name": "payment_days",
      "type": "number", 
      "question": "In how many days did the customer say they will make payment? Return only the number. If no timeline mentioned, return N/A"
    },
    {
      "name": "requirements",
      "type": "text",
      "question": "What specific product requirements did the customer mention? List all features and specifications discussed. If none mentioned, return N/A"
    }
  ]
}

​
Best Practices

​
Question Writing Guidelines

Precision is Key: The more specific your question, the more accurate and useful your insights will be. Take time to craft detailed, unambiguous questions.
AspectBest PracticeExample
ClarityUse simple, direct language”Did the customer confirm…” not “Was there confirmation…”
SpecificityDefine exact conditions”Return True only if they said ‘yes’ or ‘I agree‘“
FallbacksAlways include N/A instructions”If unclear or not discussed, return N/A”
ContextProvide relevant background”Based on the sales conversation…”

​
Data Type Selection

Best for:
  • Customer feedback and comments
  • Product requirements and specifications
  • Objections and concerns
  • Open-ended responses
Example: “What objections did the customer raise about the pricing?”
Best for:
  • Quantities and amounts
  • Timelines and dates
  • Ratings and scores
  • Measurable values
Example: “How many users will be using the system?”
Best for:
  • Yes/no decisions
  • Confirmation of actions
  • Binary states
  • Agreement/disagreement
Example: “Did the customer agree to the proposed terms?”

​
Insights in Webhooks

When you configure post-call insights webhooks, the insights data is included in the questions array: 
{
  "call_uuid": "unique-call-id",
  "questions": [
    {"key": "your_insight_name", "value": "extracted_value"},
    {"key": "another_insight", "value": true},
    {"key": "numeric_insight", "value": 45}
  ],
  // ... other call data
}

​
Troubleshooting Insights

​
Common Issues

ProblemCauseSolution
Insights not appearingCreated after campaign runCreate insights before running campaigns
Inaccurate responsesVague questionsMake questions more specific and detailed
N/A responsesInformation not discussed in callNormal behavior - indicates topic wasn’t covered
Wrong data typeType mismatch with expected answerVerify question matches selected data type

​
Quality Assurance

1

Test Questions

Validate Your Insights
  • Run sample calls with known outcomes
  • Verify AI answers match expected results
  • Refine questions based on accuracy
2

Monitor Results

Track Insight Quality
  • Review insights from first few calls
  • Look for patterns in N/A responses
  • Adjust questions if needed
3

Iterate and Improve

Continuous Optimization
  • Refine questions based on results
  • Add new insights as needs evolve
  • Remove insights that aren’t valuable

​
Next Steps

To get started with Insights:
  1. Plan your questions - Identify what information you need from calls
  2. Write precise questions - Follow the specificity guidelines above
  3. Create insights - Set up insights before running campaigns
  4. Test thoroughly - Validate accuracy with sample calls
  5. Monitor and optimize - Continuously improve question quality
Ready to extract valuable insights? Well-crafted insights questions transform your call data into actionable business intelligence, helping you make data-driven decisions and optimize your campaigns.
Need help with webhooks? Check out our Advanced Settings guide to learn how to receive insights data automatically via webhooks.
⌘I