Agent Dashboard


15 key metrics to track Chatbot and Human Agents performance


Human Agents + Chatbot, it is crucial to track key metrics, continuously improve chatbot performance and free up Human Agents from mundane tasks.

This section we would focus on 15 key metrics which will help us track the performance of Chatbot and Human Agent. Then we focus on how we can use the detail Chatbot failure report to improve the bot's performance.

Tip: Sign-up today for a FREE AmplifyReach Catalyst Platform account and try out how you can free up human agents from mundane tasks. SIGN UP FREE

Bot Performance Report


Bot Performance Reports helps to look at total number of request severed by the bot successfully and all the different reasons why the chat was transferred to an human agent.

Tip: This graph is measured in “number of request” and not “number of chats”. A single chat session can server multiple request to a customer.

Tip: A chat session for a web channel ends when customer / bot / human agent closes the chat.

  • Successfully Handled - Successful Requests handled by the bot.
  • Trigger Failure - Request failed due to Triggers failure. Trigger failure can occur due to timeout, input validation condition check with-in trigger and HTTP failure.
  • Input Validation Error - Bot is not able to understand the input entered by the customer or the validation check under the Input Parameter node for the input type fails.
  • Intent Identification Failure - Intents nodes help the bot to identify customer intent by parsing the English statement. When Bot is unable to understand what customers is trying to say, it could be because of the missing training example or the bot is not designed to handle those request. In such cases the bot would handover the chat to human agent.
  • Info Not Available - When customer says that I don’t have a particular information handy or available, those failures are captured here.

Tip: The details on all the different reasons why a chat got transferred to an human agent with all the details regarding chat can be found under “Chatbot Failure Report”

Bot vs Agent Chats


Comparison between chats handled between Bot and Human Agent. Human Agents handles chat only when it is transferred from Bot due to some failure.

Chat Request Over Time


The below graph provides the agents with a quick analysis of the chat traffic spread across the date range selected as part of the filter.

Device Distribution


This provides the distribution of chat traffic with respect to Web, Mobile and Tablet.

Chat By Departments


Departments are mapped with “Purpose of conversation” under the conversation builder. Departments can added through the Live Chat Software settings.

This provides a snapshot of chat traffic per department for Bot, Human Agent and Total.

  • Bot - Total chats successfully handled by Bot.
  • Agent - Total chats that were transferred and handled by human agents.
  • Total - Total chats handled for a particular department.

Chats Handled By Request Type


Request type are identified by the intent name provided as part of the intent node or from the value of the top level menu shown through a UI Control. Read more on naming the request type here.

This provides a snapshot of chat traffic per request type for Bot, Human Agent and Total.

Closed and Handover Chats


This provides a snapshot of chat closed by Visitor, Agent and total Handover chats.

Proactive vs User Initiated Chats


This provides a comparison snapshot of chats initiated Proactively vs User Initiated. Proactive chats are where chat proactively gets launched depending on parameters set under Live Chat Software.

Average Handling Time


Average Handle Time (AHT) is a metric that measures how long it takes bot / agent to complete a chat session.

Bot Average Handle Time (AHT) is calculated for all successful request handled by Bot and doesn’t consider the request that were handovered to Human Agents.

Human Agent Average Handle Time (AHT) is calculated from the time the Agent accepts the chat till the chat is closed either by Visitor / Agent.

Average Handling Time by Department


This provides a snapshot of Average Handling Time (AHT) broken down per department for Bot, Human Agent and Overall for a chat session.

Average Handling Time by Request Type


This provides a snapshot of Average Handling Time (AHT) broken down per request type for Bot, Human Agent and Overall.

Visitor Wait Time Report


This provides a snapshot of wait time between chat handovered by bot and Human Agent accepting the chat. This also provides the longest wait time for a customer in the date range selected.

Chat Handled by Agents


This provides a snapshot of number of chats handled per human agent.

Average Chat Time by Agents (Mins)


This provides a snapshot of Average Chat Time in Minutes per human agent.

Chat Metrics


A quick snapshot of chat metrics including:

  • Total Chats - This is with respect to the selected date range
  • Total Active Bot Chats - Total number of active chats between bot and customers
  • Total Active Agent Chats - Total number of active chats between human agents and customers
  • Handover Pending Chats - Total chats which are transferred by bot and waiting for a human agent to accept

Improve Chatbot Perforance with detailed Chatbot Failure Report


Chat bot failure report is one of the crucial report for improving Bot performance and reducing the number of handover happening to Human Agents. This is a detail snapshot of information providing all the details regarding the failure and why the Bot transferred a chat session to Human agent. This report can be downloaded for further analysis in formats like CSV and PDF.

There are four high level reason for bot failure to handle a request

  • Trigger Failure - Request failed due to Triggers failure. Trigger failure can occur due to timeout, input validation condition check with-in trigger and HTTP failure.
  • Input Validation Error - Bot is not able to understand the input entered by the customer or the validation check under the Input Parameter node for the input type fails.
  • Intent Identification Failure - Intents nodes help the bot to identify customer intent by parsing the English statement. When Bot is unable to understand what customers is trying to say, it could be because of the missing training example or the bot is not designed to handle those request. In such cases the bot would handover the chat to human agent.
  • Info Not Available - When customer says that I don’t have a particular information handy or available, those failures are captured here.

The Bot failure reports consists of:

  • Date - Date for the conversation
  • Chat Id - The complete chat transcript can be viewed using this chat id by going to Navigation Bar → Chat → Chat list
  • Type - This is a filter to club the failures into three categories
    • Intent
    • Input Variable - Captures issues for Input validation and Info not available
    • Trigger
  • Description - This provides the sub-type for the failure
    Type Description Error
    Input Variable Invalid variable value Bot is not able to understand the input entered by the customer or the validation check under the Input Parameter node for the input type fails.
    Input Variable Handover - Invalid variable value retry limit reached Max retry limit is reached and the handover is initiated by Bot to Human Agent.
    Intent Intent detection failed Intents nodes help the bot to identify customer intent by parsing the English statement. When Bot is unable to understand what customers is trying to say, it could be because of the missing training example or the bot is not designed to handle those request.
    Intent Handover - Intent detection failed and retry limit reached Max retry limit is reached and the handover is initiated by Bot to Human Agent.
    Trigger Handover - Trigger status failure Successful status check for the trigger fails
    Trigger Handover - Trigger timeout The HTTP API / Service URL that the trigger is trying to consume timeout
    Trigger Handover - Trigger validation failure Validation provide under the trigger fails
    Input Variable Handover - Info not available When customer says that I don’t have a particular information handy or available, those failures are captured here.
    Trigger Chat Closed - Trigger status failure Chat closed indicates that Human Agent support is turned off. Successful status check for the trigger fails.
    Trigger Chat Closed - Trigger timeout Chat closed indicates that Human Agent support is turned off. The HTTP API / Service URL that the trigger is trying to consume timeout.
    Trigger Chat Closed - Trigger validation failure Chat closed indicates that Human Agent support is turned off. Validation provide under the trigger fails.
    Input Variable Chat Closed - Invalid variable value retry limit reached Max retry limit is reached and the chat is closed as Human Agent support is off.
    Intent Chat Closed - Intent detection failed and retry limit reached Max retry limit is reached and the chat is closed as Human Agent support is off.
    Input Variable Chat Closed - Info not available When customer says that I don’t have a particular information handy or available, those failures are captured here. In this particular case the chat is closed by bot as Human Agent support is off.
  • Failed On - This provides the information on the conversation node where the failure occurred depending on the Type of failure.
    • Intent - As we were unable to identify the customer's intent, this would be blank
    • Input variable - Name of the variable that we were unable to validate or information was not available with customer.
    • Trigger - Name of the trigger that failed
  • User Message - The message provided by the customer after which the failure occurred. Here we will able to review that the customer had entered. Depending on the type of failure, we can use this data to improve the workflow
    • Intent - We can look at the customer message to determine if the customer was expressing same intent in different manner where we could update the intent training example. Customer may also have asked for a different service request which is not yet implemented.
    • Input Variable - This will help us look into validation problems or how to handle if a particular value is not available with customer.
    • Trigger - Thi will be blank in case of trigger.
  • Number of Retries - Every failure, even if it doesn’t result into handover is captured as per retry count.
    • Example: If we are asking customer to enter policy number in a certain format. The customer enters wrong twice but 3rd time he / she is able to enter in correct format. Then in that case there would be 2 entries in the report, one entry with retry count 1 and another with retry count 2.
  • Request Type - Request type for which the failure occurred. This field would be empty if the Bot failed to identify the intent.
  • Department Name - Name of the department for which the request was being served.
  • Conversation Name - Name of the conversation for which the request was being served.
  • Deployment ID - This help to trace which deployment did this failure belong too. This is helpful mainly if you have multiple deployments.

Date range filter for dashboard


Select the date range and click on “Search” button to populate the different analytical information. All the graphs except the Active chat metrics are with respect to the below date range selected.