Analytics

Analytics lets you review customer messages and your assistant's answers in aggregate. You can see on which days and hours messages increase, which channels they come from and how the answers are distributed.

This information helps you plan your team's working hours and identify your assistant's knowledge gaps.

Selecting the Date Range

Before starting the analysis, select the period you want to review. You can use the ready-made date options or set a start and end date.

The history you can view depends on your account's analysis allowance. When you select a range older than the allowed period, the data may be narrowed to this limit. In such a case, take the on-screen notice into account.

On Starter accounts the analysis window is the last 90 days. On accounts opened earlier and continuing with Free rights, this period is the last 7 days. Other older accounts may have different rights. The analysis window is not the same as the deletion period of past records.

Message Counts and Daily Change

In the daily message chart you can review incoming and outgoing messages:

• Incoming messages: Messages sent by customers.
• Outgoing messages: Answers and other messages sent to customers.
• Total messages: The sum of incoming and outgoing messages.

The message count is not the same as the number of customers or conversations. A customer can send more than one message in the same conversation.

The daily customer message chart is kept separate from test conversations. If there is test data in the period you selected, test console messages are shown in a separate chart. Do not use the test chart as a measure of the real demand coming from customers.

Hourly Message Density

The hourly chart shows at which hours of the day messages concentrate in the period you selected. You can use it to plan team working hours and task distribution during busy periods.

This chart aggregates the messages within the period by hour. It is not a calendar where each day is shown separately hour by hour.

Answer Types and Response Time

Answer types show the distribution of messages by whether they were created by AI, an agent, an automated action or the system.

Response time indicators are calculated from answers whose time can be measured:

• Average response time: The average of the measured response times.
• Median response time: The value in the middle when the times are sorted. It provides a comparison in cases where a few very long answers raise the average.

A fast answer does not mean the question was answered correctly or completely. To assess answer quality, also review the relevant conversations.

AI Resolution Rate and Human Handoff

The AI Resolution Rate on screen is calculated based on the current human-support status of customers who have messages in the selected period. It shows the proportion of customer records marked as not requiring human support.

This rate is not the proportion of conversations where the customer confirmed their issue was resolved. When evaluating, review the conversation content, customer feedback and the requests in the Human Handoff section together.

Channel Distribution

Channel distribution shows how messages are distributed among the connected channels. This way you can see which channel your customers reach you through most.

An increase in the message count on a channel does not mean the same proportion of new customers arrived. Existing customers sending more messages can also increase this number.

Customer Behavior and Sales Stages

The customer behavior section includes indicators such as the number of distinct customers, the amount of messages per conversation and customers who contact again.

Sales stages show which stages the customer records are in. Contact information indicators help you track records that have an email or phone.

When interpreting these indicators, consider how up to date the customer records are. For example, a record's sales stage does not by itself show the amount of payment collected.

Sentiment Analysis and Customer Ratings

The sentiment tone of messages and the ratings customers give offer different information:

• Sentiment analysis: Helps you understand the positive, neutral or negative tone in a message. There can also be mixed or not-yet-classified records.
• Customer ratings: Shows the star ratings customers give. If rating data exists, you can review the average rating and the distribution of ratings.

A negative message may not mean the customer is dissatisfied with the assistant's answer. The customer may be describing a delivery or product issue they experienced. For a detailed review, use the Sentiment Analysis section.

Frequently Discussed Topics and Knowledge Gaps

You can review the topics customers frequently discuss and the phrasings they use from the Sentiment Analysis section. These help you understand which topics need explanation.

For questions that cannot be answered or that have missing information, open the Knowledge Gaps section. A frequently occurring phrase and a question that cannot be answered are not the same thing; evaluate the two pieces of information together.

For example, if customers frequently ask about delivery time, review your delivery descriptions. If the assistant cannot answer these questions, add the relevant information or FAQ record, then try it in the AI Test section.

When Is the Data Updated?

Analytics data is retrieved when the screen is opened and when the date selection changes. To see the most current status, you can reload the screen. To follow new messages in real time, use the Messages section.

An empty chart or a value that appears as zero does not by itself indicate that the assistant is not working. First check the selected date, the messages in the relevant period and whether the data the indicator needs is available.

Accessing Analytics Data via API

Accounts with API access can retrieve summary data from GET /api/v1/analytics with the analytics:read permission.

This response includes message, customer and session counts; an API usage summary and the daily message distribution. It is not the detailed equivalent of all the charts in the panel. The date range can be specified with the start_date and end_date parameters.

Related Topics