> ## Documentation Index
> Fetch the complete documentation index at: https://help.verbu.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Reports
> Understand the numbers on Reports: involvement, resolution, customer experience and escalations, and how each conversation gets its outcome.
**Reports** in the menu on the left shows how your agents are doing. Choose the period and the agent at the top. To look at one agent, you find the same numbers under the agent in **Analyse** → **Performance**.
The numbers rest on one rule for each conversation: did the user take part, and did the user get what they came for? This page explains the rules, so you can read the numbers correctly.
## The four headline numbers
| Number | What it shows | How it is calculated |
| ----------------------------- | -------------------------------------------------------------- | -------------------------------------------------------------------------- |
| **Involvement rate** | How many conversations the user actually took part in | Involved conversations out of all conversations |
| **Resolution rate** | How many of the involved conversations ended resolved | Resolved conversations out of the involved conversations |
| **Customer experience score** | How happy users sounded, from 0 to 100 | The average of the satisfaction we estimate for each involved conversation |
| **Escalation rate** | How many of the involved conversations were passed to a person | Escalated conversations out of the involved conversations |
Resolution, customer experience and escalations are calculated on the involved conversations only. A conversation where the user said almost nothing counts neither for nor against the agent.
## When is a conversation involved?
A conversation is involved when the user said enough for a real conversation:
* **In chat, SMS and email:** the user sent at least one message.
* **On a call:** the user spoke at least twice, or said at least six words in total.
A call where the user only says "Hello?" or "Yes." is not involved. A call with one real question, such as "Do I need to be home when the worktop is delivered?", is involved.
## How conversations ended
Each conversation gets one outcome when it ends. The outcomes are collected in groups. The diagram **From conversation to result** shows the groups and the outcomes in each group:
| Outcome | Group | What it means |
| -------------------------- | -------------- | --------------------------------------------------------------------------------------------------------------------------- |
| **Resolved, confirmed** | Resolved | The user got an answer or help and said it helped. |
| **Handed to a person** | Resolved | A person took over the conversation. |
| **Set up for follow-up** | Resolved | The agent took a message from a user who wants to be contacted. You find it under **Messages**. |
| **Resolved, assumed** | Resolved | The user got an answer or help and did not come back about the same thing. |
| **Reversed** | Not resolved | The conversation was resolved, but the same user came back about the same thing within 7 days. The resolution did not hold. |
| **Asked for a person** | Not resolved | The user asked for a person, and no follow-up was arranged. |
| **Left without an answer** | Not resolved | The user left before what they came for was answered. |
| **No involvement** | No involvement | The user said too little for a real conversation. |
Conversations that do not have an outcome yet show as **Not evaluated**.
Click an outcome in the diagram to see exactly those conversations in **Inbox**, with the same agent and period. If the report leaves test conversations out, the inbox does too. A chip at the top shows the filter, and one click removes it.
A handover counts as resolved. Either a person took the conversation, or you have a message from a user who wants to be contacted. Both are a good result for you.
## How the outcome is decided
The outcome is decided when the conversation ends, in two steps:
1. **Fixed rules first.** If the conversation was transferred to a person, the outcome is **Handed to a person**. If the agent took a message, the outcome is **Set up for follow-up**. If the user did not take part, the outcome is **No involvement**. If the user asked for a person and a transfer was tried without success, the outcome is **Asked for a person**.
2. **Then the whole conversation is read.** We look at what the user asked for through the whole conversation, and whether the agent handled it. If the user left after getting their answer, the conversation is resolved. What the agent says at the end, such as "Are you still there?", does not change that.
Some answers need a quote or a colleague, such as a price that depends on measurements. Such a request counts as resolved when the agent gave the user the next concrete step. That can be an SMS with what the user needs to send, or a message to you. If the agent only said that it depends on measurements, it is not resolved.
If the user asked for a person on a call and we know the number, the conversation is marked so someone calls the user back. It then counts as **Set up for follow-up** and waits for you in **Inbox**.
## From conversation to result
The diagram shows where the conversations came in, whether the user took part, and which group they ended in. Follow a colour from left to right to see where the conversations from one channel ended. The last column splits **Resolved** and **Not resolved** into each outcome. Hover over an outcome to see what it means. Click a channel, a group or an outcome to open the conversations in **Inbox**.
## Over time
The chart shows involvement, resolution and customer experience day by day. Each value is an average of the last seven days, so a quiet day does not make the line jump. Switch to **Absolute numbers** to see the counts behind the rates.
## Why the experience was what it was
Each involved conversation gets one reason why the experience was good or could be better.
**What went well**
| Reason | What it means |
| -------------------------- | ------------------------------------------------------------------------------ |
| **Answered first time** | The agent gave the right answer the first time. |
| **Task completed** | The agent did what the user asked, such as a booking. |
| **Handled the unexpected** | The conversation went off the usual path, and the agent still handled it well. |
| **Warm and clear** | The agent was friendly and easy to understand. |
**What could be better**
This list shows only the reasons you can do something about. Click a reason to go to the page where you fix it.
| Reason | What it means | Where you fix it |
| ------------------------------ | --------------------------------------------------------- | --------------------------------------------------- |
| **No answer found** | The agent did not have the answer in its knowledge. | **Knowledge**: add the missing content. |
| **Wrong or misleading answer** | The agent gave an answer that was wrong or could mislead. | **Knowledge**: correct or remove the wrong content. |
| **No person available** | The user wanted a person, and nobody was available. | **Train** → **Escalate** under the agent. |
| **Too many questions** | The agent asked more questions than it needed. | **Train** → **Guidelines** under the agent. |
The links to **Escalate** and **Guidelines** show when you choose an agent at the top.
Some reasons do not show here, such as the agent repeating itself, mishearing the user or being slow. We fix those in the platform. We still keep them and use them to make the agents better.
## What to check
These are usually calls where the user hung up straight away or only said "Hello?", and chats where the user never wrote anything. They do not count in the resolution rate.
The outcome now reads the whole conversation and looks at what the user asked for. Before, a conversation could show as left without an answer only because the agent ended with a question. A handover to a person or a message now also counts as resolved.
Open the conversation in **Inbox**. Under **Conversation outcome** you find a short reason. Write to us with a link to the conversation, and we will look at the rule.