Skip to main content
Post-Call Analysis (PCA) extracts structured fields from call transcripts after each call ends. The Analysis results view in Monitor lets you see those extracted values across all calls, filter by specific field values, and route downstream workflows based on what was extracted. This page covers the results view and how to work with extracted data. For configuring the extraction fields themselves, see the agent’s Post-Call Analysis settings. Access the results view at Monitor > Post-Call Analysis.

​
What you see in the results view

The results view shows a table of sessions with one row per call. Each column corresponds to a configured PCA field for the selected agent. Common fields include:
  • Intent — the caller’s primary goal (e.g., booking, cancellation, inquiry, complaint)
  • Sentiment — caller sentiment at end of call (e.g., positive, neutral, negative)
  • Resolution status — whether the call was resolved (resolved, escalated, abandoned)
  • Extracted entities — specific data points collected during the call (e.g., appointment_date, product_name, complaint_category)
Any field you defined in the agent’s PCA configuration appears as a column.

​
Filtering by field value

Use the Filter controls at the top of the results table to narrow by extracted field values. Example filters:
  • Show only calls where intent = booking
  • Show only calls where sentiment = negative and resolution_status = escalated
  • Show only calls from a specific date range where appointment_date was extracted
Filters are additive (AND logic). Applying multiple filters shows only sessions that match all conditions.
Combine date range and field value filters to answer operational questions like “How many complaints did we receive in the last 7 days?” or “What was the booking rate for last Monday’s campaign?”

​
Exporting analysis results

Click Export to download the filtered result set as a CSV file. Each row is a session; columns are the session metadata (session ID, date, duration, agent, phone number) plus all configured PCA fields. For API access: 
GET /v1/sessions?agent_id=agent_xxx&pca_field=intent&pca_value=booking&start=2024-01-01
The response includes full session metadata and extracted PCA fields for each matching session.

​
Using analysis results to route workflows

PCA results integrate with webhook delivery and downstream automation. When a call ends and analysis completes, a call.analysis_ready webhook fires containing the extracted fields. Use this to trigger workflows based on specific values. Example: Send a WhatsApp follow-up for booking calls 
// Webhook payload (call.analysis_ready)
{
  "event": "call.analysis_ready",
  "session_id": "sess_xxxxxxxx",
  "phone_number": "+91987654321",
  "analysis": {
    "intent": "booking",
    "appointment_date": "2024-03-20",
    "sentiment": "positive"
  }
}
In your webhook handler, check analysis.intent === "booking" and enqueue a WhatsApp confirmation message with analysis.appointment_date. Example: Escalate negative-sentiment calls to a human queue Check analysis.sentiment === "negative" and analysis.resolution_status === "unresolved" to trigger a CRM ticket or push the session to a supervisor review queue.

​
Accuracy and reliability

Post-call analysis uses the LLM to extract fields from the transcript. Extraction accuracy depends on:
  • Transcript quality: if the transcription had errors, the extracted fields may be wrong. Check Session History to compare the raw transcript against the extracted values.
  • Field definition clarity: vague extraction prompts produce inconsistent values. Define fields with clear instructions and explicit value sets (e.g., “Classify as one of: booking, cancellation, inquiry, complaint”).
  • Call content: if the caller did not discuss a particular topic, the field will be empty or marked as not applicable. This is expected behavior, not an error.
Do not use PCA results as the sole input for high-stakes automated actions (refund triggers, account changes) without human review. LLM extraction can produce incorrect values, especially for callers with unclear speech or unusual conversation patterns.