> ## Documentation Index
> Fetch the complete documentation index at: https://docs.talkifai.dev/llms.txt
> Use this file to discover all available pages before exploring further.
# Post-Call Analysis
> Automatically analyze call quality, extract insights, and ensure compliance with customizable AI-powered evaluation forms.
## Overview
**Post-Call Analysis** automatically evaluates every call session using AI to:
* Assess call quality and agent performance
* Extract key insights and action items
* Ensure compliance with company standards
* Identify training opportunities
* Track metrics over time
**Best for:**
* Quality assurance teams
* Call center managers
* Compliance officers
* Training coordinators
***
## How It Works
```
Call ends
│
▼
Transcript saved to database
│
▼
Post-call analysis triggered automatically
│
▼
AI analyzes transcript using configured form
│
▼
Results saved and linked to call session
│
▼
Available in Analytics dashboard
```
***
## Step 1: Configure Analysis Model
### Choose Analysis Model
Go to **Agent Settings → Post-Call Analysis** and select:
| Model | Speed | Cost | Best For |
| ------------------------- | --------- | ---- | ---------------------------- |
| **GPT-5 Mini** ⭐ | Fast | \$\$ | Most use cases |
| **GPT-5 Nano** | Very fast | \$ | High volume, simple analysis |
| **Gemini 2.5 Flash** | Fast | \$\$ | Google ecosystem |
| **Gemini 2.5 Flash Lite** | Very fast | \$ | Cost-sensitive analysis |
**Recommended:** GPT-5 Mini offers the best balance of accuracy and cost for most quality assurance workflows.
### Update via API
```bash theme={null}
PUT /api/post-call-analysis/model/{agentId}
Authorization: Bearer YOUR_API_KEY
{
"model": "gpt_5_mini"
}
```
***
## Step 2: Create Analysis Form
### Form Fields
Each analysis form contains custom fields that define what the AI should evaluate:
**Field Types:**
| Type | Use Case | Example |
| ------------ | ---------------------- | ----------------------------------------------- |
| **Text** | Open-ended analysis | "Summarize the customer's issue" |
| **Selector** | Multiple choice rating | "Call outcome: Resolved / Partial / Unresolved" |
| **Boolean** | Yes/No questions | "Did the agent follow greeting protocol?" |
| **Number** | Numeric scores | "Rate professionalism 1-10" |
### Create Form via Studio
1. Go to **Agent Settings → Post-Call Analysis**
2. Click **Add Field**
3. Configure each field:
| Field | Description |
| --------------- | ------------------------------------ |
| **Name** | Internal identifier (snake\_case) |
| **Description** | What to evaluate |
| **Type** | Text, Selector, Boolean, or Number |
| **Options** | For Selector type: available choices |
| **Required** | Must this field be completed? |
| **Order** | Display order in results |
4. Click **Save Form**
### Example Form: Customer Support QA
```json theme={null}
[
{
"name": "call_summary",
"description": "Briefly summarize the customer's issue and how it was resolved",
"type": "text",
"optional": false,
"order": 1
},
{
"name": "call_outcome",
"description": "How was the call resolved?",
"type": "selector",
"options": [
"Fully Resolved",
"Partially Resolved",
"Unresolved - Escalated",
"Unresolved - Follow-up Required"
],
"optional": false,
"order": 2
},
{
"name": "agent_professionalism",
"description": "Rate the agent's professionalism and tone",
"type": "number",
"optional": false,
"order": 3
},
{
"name": "followed_protocol",
"description": "Did the agent follow the greeting and closing protocol?",
"type": "boolean",
"optional": false,
"order": 4
},
{
"name": "empathy_shown",
"description": "Did the agent demonstrate empathy for the customer's situation?",
"type": "boolean",
"optional": true,
"order": 5
},
{
"name": "improvement_areas",
"description": "Identify specific areas where the agent could improve",
"type": "text",
"optional": true,
"order": 6
}
]
```
### Create Form via API
```bash theme={null}
POST /api/post-call-analysis/{agentId}
Authorization: Bearer YOUR_API_KEY
{
"name": "Customer Support QA",
"description": "Quality assurance evaluation for support calls",
"type": "text",
"optional": false,
"order": 1
}
```
***
## Step 3: View Analysis Results
### Individual Call Results
1. Go to **Analytics → Call Logs**
2. Click on any completed call
3. View **Post-Call Analysis** tab
4. See AI-generated evaluation
**Example Result:**
```json theme={null}
{
"roomName": "agent_123_1705320000_xyz",
"analysisData": {
"call_summary": "Customer called about delayed order #ORD-12345. Agent checked system, identified shipping delay, and arranged expedited replacement. Customer satisfied with resolution.",
"call_outcome": "Fully Resolved",
"agent_professionalism": 9,
"followed_protocol": true,
"empathy_shown": true,
"improvement_areas": "Agent could have proactively offered compensation for the inconvenience."
},
"createdAt": "2024-01-15T10:35:00Z"
}
```
### Aggregate Analytics
Go to **Analytics → Quality Scores** to see:
* **Average scores** by metric
* **Trends over time**
* **Agent performance comparison**
* **Common improvement areas**
* **Compliance rates**
***
## Advanced Features
### Drag-and-Drop Field Ordering
**In Studio:**
1. Go to **Agent Settings → Post-Call Analysis**
2. Drag fields to reorder
3. Changes save automatically
4. Order updates for all future analyses
### Multiple Analysis Forms
Create different forms for different call types:
**Example Setup:**
```
Sales Agent
├── Lead Qualification Form
├── Demo Call Form
└── Follow-up Form
Support Agent
├── Technical Issue Form
├── Billing Inquiry Form
└── Escalation Form
```
### Custom Analysis Models
For specialized needs, use custom models:
**Use Cases:**
* Industry-specific compliance (HIPAA, PCI-DSS)
* Multi-language analysis
* Sentiment analysis for specific demographics
* Custom scoring algorithms
Contact support for custom model setup.
***
## API Reference
### List Analysis Forms
```bash theme={null}
GET /api/post-call-analysis/{agentId}
```
**Response:**
```json theme={null}
{
"success": true,
"fields": [
{
"id": "field_abc123",
"name": "call_summary",
"description": "Summarize the customer's issue",
"type": "text",
"optional": false,
"options": [],
"order": 1,
"createdAt": "2024-01-10T09:00:00Z"
}
]
}
```
### Create Field
```bash theme={null}
POST /api/post-call-analysis/{agentId}
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json
{
"name": "customer_satisfaction",
"description": "Rate overall customer satisfaction",
"type": "number",
"optional": false,
"order": 1
}
```
### Update Field
```bash theme={null}
PUT /api/post-call-analysis/{agentId}
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json
{
"fieldId": "field_abc123",
"name": "customer_satisfaction_score",
"description": "Rate customer satisfaction (1-10)",
"type": "number",
"optional": false,
"options": [],
"order": 1
}
```
### Delete Field
```bash theme={null}
DELETE /api/post-call-analysis/{agentId}?fieldId={fieldId}
Authorization: Bearer YOUR_API_KEY
```
### Get Analysis Results
```bash theme={null}
GET /api/post-call-analysis/results/{agentId}
Authorization: Bearer YOUR_API_KEY
Query Parameters:
- roomName: Filter by specific call
- startDate: Filter by date range
- endDate: Filter by date range
- limit: Number of results (default: 20)
```
**Response:**
```json theme={null}
{
"success": true,
"results": [
{
"id": "result_xyz789",
"roomName": "agent_123_1705320000_xyz",
"analysisData": {
"call_summary": "...",
"call_outcome": "Fully Resolved",
"agent_professionalism": 9
},
"createdAt": "2024-01-15T10:35:00Z"
}
]
}
```
***
## Use Cases
Automatically evaluate every call for compliance with company standards. Identify agents needing coaching.
Ensure regulated industries (healthcare, finance) meet compliance requirements. Generate audit reports.
Identify common mistakes and improvement areas. Create targeted training programs.
Extract recurring issues, sentiment trends, and customer pain points from call data.
Track agent performance over time. Set benchmarks and measure improvement.
Flag problematic calls for review. Identify potential legal or PR risks early.
***
## Best Practices
### Writing Effective Analysis Fields
❌ "Rate the call quality"
✅ "Did the agent actively listen and acknowledge the customer's concerns?"
❌ "Rate professionalism 1-10" (undefined)
✅ "Rate professionalism: 1=Rude, 5=Neutral, 10=Exceptionally polite"
Limit to 5-10 key metrics. Too many fields increase analysis time and cost.
Always have a text field for "Improvement areas" or "Action items" to get actionable feedback.
### Analysis Frequency
| Use Case | Recommended Frequency |
| ----------------------- | --------------------------- |
| **Quality Assurance** | 100% of calls |
| **Compliance** | 100% of calls |
| **Training** | 10-20% random sample |
| **Customer Insights** | 100% of calls |
| **Performance Reviews** | All calls + monthly summary |
### Cost Optimization
**Tips:**
* Use **GPT-5 Nano** for high-volume simple analysis
* Analyze only **completed calls** (skip abandoned/failed)
* Limit text fields to **200 words max**
* Batch analysis during off-peak hours (coming soon)
***
## Troubleshooting
**Problem:** Analysis typically takes 10-30 seconds per call. If longer:
**Solutions:**
* Reduce number of fields
* Use faster model (Nano or Flash Lite)
* Shorten text field descriptions
* Check if transcript is very long (>30 min call)
**Problem:** Field descriptions are too vague.
**Solution:** Be more specific:
* ❌ "Rate the agent"
* ✅ "Rate how well the agent explained the solution in simple terms"
**Problem:** Optional fields may be skipped if AI can't determine answer.
**Solution:**
* Make critical fields required (optional: false)
* Improve field descriptions
* Add examples to descriptions
**Solutions:**
* Switch to cheaper model (Nano or Flash Lite)
* Analyze sample instead of 100% of calls
* Reduce number of text fields (most expensive)
* Shorten analysis prompts
***
## Next Steps
Set up your first quality assurance evaluation form.
See aggregated quality scores and trends.
Download analysis results for external reporting.