Sentiment Analysis

Sentiment Analysis shows the positive, neutral or negative tone of analyzed customer messages. It helps you find negative messages and review the relevant conversation, and notice recurring issues.

The rates here are separate from the satisfaction scores customers give. The negative language a customer uses while describing an issue does not mean they found the assistant's answer bad.

What Do the Sentiment Labels Mean?

• Positive: Messages containing thanks, satisfaction or a positive approach.
• Neutral: Questions and information requests that carry no clear positive or negative emotion.
• Negative: Messages containing negative expressions such as complaint, disappointment or anger.

Labels are an AI assessment. Short phrases, irony or statements that depend on earlier messages in the conversation can be misinterpreted. Before taking an action, read the message within the whole conversation.

How Is It Used?

If the list has multiple pages, you can navigate with the previous and next page buttons. When the whole conversation cannot be loaded, only the selected message and, if any, the relevant AI answer may be shown.

General Distribution and Daily Change

The general distribution shows the proportions of messages with a sentiment label in the selected period. The daily chart lets you see how these messages change by day.

The rates are based on messages, not on the number of distinct customers. For example, five negative messages from the same customer do not mean five separate customers.

When comparing one period with another, also consider the message count. In a period with few messages, a few new messages can noticeably change the percentages.

Topics and Frequently Used Phrases

In this section, when data exists, you can see the topics customers frequently discuss and the phrases they use. Sentiment information by topic also helps you review which areas have more positive or negative messages.

The topic and phrase lists are not directly a list of the most frequently asked questions. The same topic can be raised with different questions or complaints. Before adding information, read the relevant messages.

How Should You Evaluate the Results?

Do not track negative messages merely as a number; investigate their common causes. The issue may be missing information, a misunderstood answer or a process in your business.

• If information is missing, update the relevant training content or FAQ record.
• If the answer is not clear enough, simplify the explanation and try it in the AI Test section.
• If the conversation requires human support, follow it from the Human Handoff section.

For example, it can be expected that the message "My product still hasn't arrived" is negative. Here, only changing the language of the answer may not be enough; the customer's order status may need to be checked.

The suggestions on screen provide general guidance. When deciding about a customer, rely on the content of the conversation and, if any, the direct rating score.

If There Is No Analysis Data

Messages for which a sentiment assessment has been created are reflected on this screen; you do not need to enable a separate analysis setting to view them.

If there are no analyzed messages in the period you selected, the screen may remain empty. This does not mean that customers think negatively or that satisfaction is zero. Check the date selection and whether there are customer messages in the relevant period.

History View Limit

The period you can view depends on your account's analysis allowance. On Starter accounts the last 90 days are covered. On accounts opened earlier and continuing with Free rights, the last 7 days are covered; other older accounts may have different rights. Selecting a longer period on screen does not mean data older than the account allows will be shown.

Related Topics