# Insights

EARLY ACCESS
Important  
**Topics** and **success evaluation** are available only in the following data centers: **Europe (EU1)**, **Europe (EU2)**, **Germany (FR4)**, **Qatar**, **Sweden**, **United Arab Emirates**, and **United States**.

**Insights** enables you to analyze performance and view detected issues across your journeys, chatbots, AI agents, and human agents.

Get **AI-driven recommendations** to fix them, and track how conversation topics are performing.

Insights displays three types of information:

1. **Total interactions and success rate** - Count of customer-facing interactions and success percentage
2. **Top recommendations** - Detected issues and suggested fixes
3. **Topic performance** - Interaction topics with performance metrics

___

## Insights filters

These are global filters for both Insights and [Supporting metrics](https://www.infobip.com/docs/insights-and-analytics/insights/supporting-metrics). Use the filters at the top of the page to refine your data by:

- **Time period**
- **Journeys**
- **Chatbots**
- **Version** (if multiple versions of a flow exist)
- **Sender**

In each drop-down menu, you can search for specific journeys, chatbots, or senders.

___

## Total interactions and success rate

This section displays the total number of customer-facing interactions and the overall success rate based on applied filters.

### Total interactions [#total-interactions-total-interactions-and-success-rate]

**Total interactions** cover all interactions with **Automations** and **Inbox**. An interaction is an ongoing exchange of information between an end user and the platform, identified by an interaction ID. It starts in a Journey, Chatbot, or Inbox and has a defined end.

This metric includes both **conversational** (message exchanges with bots or human agents) and **non-conversational** (automated flows) interactions.

### Success rate [#success-rate-total-interactions-and-success-rate]

Success rate is the percentage of successfully resolved interactions compared to all interactions based on applied filters.

#### How success rate is calculated

Success rate is evaluated in two ways depending on whether the interaction is **conversational** or **non-conversational**:

1. **Conversational interaction** A conversational interaction is one in which at least one inbound message has been exchanged. Success is evaluated based on the transcription of all exchanged messages, driven by the outcome. This means whether the customer inquiry was addressed and resolved within the interaction. Evaluation of success is determined by AI technology, where the whole transcript is analyzed. Success is driven by outcome for the customer, not output.
2. **Non-conversational interaction** A non-conversational interaction is one in which no inbound message has been exchanged. Success is evaluated based on AI-driven analysis of the entire interaction setup to detect its purpose. Success is determined when the detected purpose was achieved (the customer completed the intended action during the interaction).

___

## Top recommendations

The **Top recommendations** section lists the top 10 recommendations ordered by impact (how many journeys that specific issue has affected so far). If you apply filters by Journey or Chatbot, you will see only specific recommendations for that journey or chatbot.

### Recommendation content [#recommendation-content-top-recommendations]

Each recommendation includes:

- **Problem description**
- **Suggested fix**
- **Affected element location** (when detected)
- **Impact metrics**
- Option to drill into **interaction examples** to see more details

Recommendations are specific to individual journeys or chatbots.

___

## Topic performance

The **Topic performance** section provides an overview of interaction topics and their performance metrics.

You can view the data in two tabs:

1. **Most common topics**
2. **Lowest performing topics**

Each tab includes a table with attributes that help you analyze topic performance in more detail.

Select **View all topics** to open the **All topics performance** page, where you can explore all topics detected in customer conversations, review their performance, and see how they are handled by automation and human agents.

Refer to the table below for more information about **topic attributes**.

Note  
**Non-conversational interactions** do not have assigned topics because **no messages** are exchanged.

### How topics are detected [#how-topics-are-detected-topic-performance]

A **topic** is a subject or issue raised by the customer during the interaction. One interaction may include multiple topics.

Topics are detected using AI technology that analyzes the **transcription** of all exchanged messages. The AI identifies what subjects or issues the customer raised and whether they were addressed and resolved within the interaction.

### Topic attributes [#topic-attributes-topic-performance]

The topics table contains the following attributes:

| Attribute | Description |
| --- | --- |
| **Topic** | Subject or issue raised by the customer during the interaction. One interaction may include multiple topics. Each topic is marked successful if the customer's issue was resolved, or unsuccessful if it was not. Select a topic to drill into its performance metrics and view interaction examples that contained that topic. |
| **Success rate** | Percentage of times this topic was successfully resolved when it appeared. |
| **Interactions** | Total number of interactions in which this topic has been detected. |
| **Percentage of total interactions** | Percentage of interactions that include this topic, out of all interactions. |
| **Containment rate** | Percentage of times this topic was handled by automation when it appeared. |
| **Transferred to human agent** | Percentage of times this topic was transferred to a human agent when it appeared. |
| **Average handling time** | Average time from topic start to resolution. |
| **CSAT score** | Percentage of positive survey responses from interactions where this topic was identified. |
| **Sentiment score** | Percentage of interactions with positive sentiment where this topic was identified. The sentiment score is based on analysis of customer messages. |
| **Performance** | Overall performance score based on success rate, containment, handling time, and satisfaction. |

___

## Data refresh and availability

Topics detection and assignment to interactions, success evaluation, and interaction problems and recommendation detection are updated **every 24 hours** at the same UTC time across the globe.

When you start a new journey or chatbot, recommendations will start appearing once the data set is filled with enough data for statistical evaluation of anomalies.

This means that first recommendations could start appearing a few days after the start.

___