# Post Analysis (Reference: https://docs.iqra.bot/build/operations/analysis)

Voice conversations are unstructured. A 10-minute call is just a blob of text. **Post Analysis** uses an LLM to process the transcript immediately after the call ends, converting it into structured business data.

The Analysis Pipeline [#the-analysis-pipeline]

<Mermaid
  chart="graph LR
    Call[Call Ends] --> Transcript[Transcript]
    Transcript --> LLM{LLM Analysis}
    
    LLM --> Summary[Generate Summary]
    LLM --> Tag[Apply Tags]
    LLM --> Extract[Extract JSON Data]
    
    Extract --> Webhook[Send to Webhook]
    
    style LLM fill:#2563eb,stroke:#333,stroke-width:2px,color:#fff
    style Webhook fill:#16a34a,stroke:#333,stroke-width:2px,color:#fff"
/>

Creating a Template [#creating-a-template]

You define your analysis rules in a **Template**. This allows you to reuse the same extraction logic across multiple campaigns.

Navigate to **Business Dashboard** -> **Post Analysis** -> **Add Template**.

1. Summary [#1-summary]

The simplest form of analysis.

* **Prompt:** You can guide the AI.
  * *Default:* "Summarize the conversation."
  * *Custom:* "Summarize the user's complaints specifically focusing on technical issues."

2. Conversation Tags [#2-conversation-tags]

Classify calls into buckets for reporting. Iqra AI supports **Hierarchical Tags** (up to 5 levels deep).

* **Top Level:** e.g., "Sales", "Support".
* **Child Level:** e.g., "Sales" > "Interested", "Sales" > "Not Interested".

**Configuration:**

* **Description:** Tell the AI *when* to apply this tag. (e.g., "Apply this if the user asks for pricing").
* **Multi-Select:** Can a call have multiple tags from the same level?

3. Data Extraction (JSON) [#3-data-extraction-json]

This is the most powerful feature. You define a **Schema**, and the AI fills it out.

**Supported Data Types:**

* **String:** Text values (e.g., Names, Feedback).
* **Number:** Amounts, Quantities.
* **Boolean:** True/False flags (e.g., `is_urgent`).
* **Datetime:** Standardized timestamps (e.g., `2023-10-27T10:00:00Z`).
* **Enum (Options):** Restricts the AI to a specific list (e.g., `[Low, Medium, High]`).

Conditional Logic [#conditional-logic]

You can create dynamic extraction rules. Why extract a "Cancellation Reason" if the user didn't cancel?

* **Logic:** *If* `user_cancelled` (Boolean) is `True`, *Then* extract `cancellation_reason` (String).
* **Benefit:** Keeps your data clean and reduces hallucinations.

***

Using the Data [#using-the-data]

Once you have created a template, you must **Assign it**.

1. Go to your **Inbound** or **Outbound Campaign**.
2. Navigate to the **Post Analysis** tab.
3. Select your Template.

Viewing Results [#viewing-results]

* **Dashboard:** Open the **Conversations Tab** and click on a call details view. You will see the analysis card.
* **API/Webhook:** The analysis result is sent in the `Call Ended` webhook payload.

```json title="Example Webhook Payload"
{
  "session_id": "...",
  "summary": "User called about login issues. Reset password successfully.",
  "tags": ["Support", "Resolved"],
  "extracted_data": {
    "issue_type": "Login",
    "customer_mood": "Frustrated",
    "ticket_id": 12345
  }
}
```