> ## Documentation Index > Fetch the complete documentation index at: https://docs.callkaro.ai/llms.txt > Use this file to discover all available pages before exploring further. # Post-Call Variables > Extract structured data from conversations ## Overview Post-call variables are the data that you want to extract from the call. Define what data to extract, and AI automatically extracts it from every call. *** ## Post-Call Variables Post-call variables are information fields that AI extracts from the conversation after the call ends. This data can be used in post-call functions and analytics. ### Variable Types **Purpose:** Extract free-form text responses **Use Cases:** * Call summaries * Customer feedback * Address or location information * Notes or comments **Example:** ``` Name: detailed_call_summary Description: Detailed summary of the call covering main points discussed, customer concerns, and action items ``` **AI Behavior:** Extracts and returns free-text response based on your description **Purpose:** Extract one option from multiple choices **Use Cases:** * Customer type (new/existing/lead) * Product interest category * Service tier preference * Priority level **Example:** ``` Name: customer_type Description: What type of customer is the user? Options: new_customer, existing_customer, returning_customer, lead ``` **AI Behavior:** Analyzes conversation and selects the most appropriate option List all possible options in the description for accurate AI extraction. **Purpose:** Extract yes/no or true/false answers **Use Cases:** * Appointment booked (yes/no) * Interest expressed (true/false) * Follow-up requested * Problem resolved **Example:** ``` Name: appointment_booked Description: Did the user book an appointment during the call? ``` **AI Behavior:** Returns `true` or `false` based on conversation analysis **Purpose:** Extract numeric values **Use Cases:** * Customer age * Order quantity * Number of attendees * Budget amount **Example:** ``` Name: user_age Description: What is the user's age mentioned during the conversation? ``` **AI Behavior:** Extracts and returns numeric value from conversation *** ## Creating Post-Call Variables Click the **"Add Variable"** dropdown button Choose the appropriate type: Text, Selector, Boolean, or Number **Name:** Unique identifier (use underscores, e.g., `customer_interest`) **Description:** Clear instruction for AI on what to extract The description is critical - it tells the AI exactly what information to look for and how to extract it. Click Submit to add the variable ### Best Practices for Descriptions **Bad:** "Get customer info" **Good:** "Extract the customer's full name as mentioned during the conversation. If multiple names are mentioned, use the one they introduce themselves with." Always list all possible values: ``` Description: What is the customer's preferred contact method? Options: email, phone, whatsapp, sms ``` Help AI understand edge cases: ``` Description: Did the customer book an appointment? Return true only if a specific date and time was confirmed. Return false if they said "maybe" or "I'll think about it." ``` Frame as yes/no questions: * ✅ "Did the customer express interest in the product?" * ✅ "Was the call issue resolved?" * ❌ "Customer interest" (too vague) *** ## Using Post-Call Variables ### In Post-Call Functions Access extracted variables in your post-call functions: **Basic Mode:** ```json theme={null} { "customer_name": "{{customer_name}}", "interest_level": "{{interest_level}}", "appointment_booked": "{{appointment_booked}}" } ``` **Advanced Mode (JavaScript):** ```javascript theme={null} async function processCall(context) { // Access post-call variables const customerName = context.post_call.customer_name; const interestLevel = context.post_call.interest_level; const appointmentBooked = context.post_call.appointment_booked; // Use in your logic if (appointmentBooked) { // Send confirmation email await axios.post('https://api.example.com/send-email', { to: customerName, type: 'appointment_confirmation' }); } } ``` *** ## Post-Call Model Choose which AI model processes your post-call analysis. * **o4-mini** (Default) - Fast and cost-effective * **gpt-4.1-mini** - Balanced performance * **gpt-4.1** - Higher accuracy * **gpt-5-mini** - Latest model * **gpt-4o** - OpenAI's flagship model For most use cases, **o4-mini** provides excellent accuracy at the best price. *** *** ## Conversion Reason Define what constitutes a "conversion" for your use case. **Purpose:** Tell the AI what successful outcome looks like so it can mark calls as converted. **Examples:** **E-commerce:** ``` Customer places an order or commits to purchasing a product. Mark as converted if they provide payment information or confirm the order. ``` **Appointment Booking:** ``` Customer books a confirmed appointment with a specific date and time. Do not mark as converted for tentative or "maybe" responses. ``` **Lead Qualification:** ``` Customer expresses strong interest and agrees to a follow-up call with the sales team. They must provide their contact information and confirm a callback time. ``` Be specific about what qualifies as a conversion vs. what doesn't. This ensures accurate conversion tracking. *** ## Best Practices Begin with 3-5 core post-call variables. Add more as you identify needs from actual call data. * Use `snake_case` for names * Be descriptive: `customer_interest_level` not `interest` * Avoid abbreviations unless universally clear After running calls: 1. Check if variables are being extracted correctly 2. Refine descriptions if AI misunderstands 3. Add new variables as patterns emerge 4. Remove unused variables Use the Test Call feature to verify: * Variables are extracted as expected * Descriptions are clear enough for AI * Data types match expectations *** ## Examples by Industry ### E-commerce **Post-Call Variables:** ``` order_amount (Number): Total order amount in rupees product_category (Selector): Which product category did customer order from? Options: electronics, clothing, home, beauty delivery_preference (Selector): Preferred delivery option? Options: standard, express, same_day ``` *** ### Healthcare/Medical **Post-Call Variables:** ``` appointment_type (Selector): Type of appointment? Options: consultation, follow_up, lab_test, emergency urgency_level (Selector): How urgent is the appointment? Options: routine, urgent, emergency insurance_verified (Boolean): Was insurance information verified? ``` *** ### Real Estate **Post-Call Variables:** ``` property_type (Selector): Type of property interested in? Options: apartment, villa, plot, commercial budget_range (Text): Budget range mentioned by customer location_preference (Text): Preferred locations or areas mentioned viewing_scheduled (Boolean): Was a property viewing scheduled? ``` *** ## Troubleshooting **Possible Causes:** * Description is too vague * Information wasn't discussed in call * Variable type doesn't match data **Solutions:** * Make description more specific * Add examples of what to look for * Check call transcript to verify info was mentioned **Issue:** Too strict criteria **Solution:** Clarify what counts as true: ``` Bad: "Did customer book appointment?" Good: "Did customer book appointment? Return true if a specific date and time was confirmed, even if tentative." ``` **Issue:** Options not clearly defined **Solution:** Provide examples for each option: ``` Description: Customer's experience level? Options: - beginner (first time using, needs guidance) - intermediate (some experience, needs occasional help) - expert (very familiar, minimal assistance needed) ``` *** ## Getting Help Use Test Call to verify variables work as expected Our team can help optimize your post-call variable configuration for your specific use case