What is sentiment analysis?

Sentiment analysis, explained

Sentiment analysis reads text and estimates its tone. Early methods used word lists; modern ones use language models that take context into account and can handle phrases like “not bad at all”.

Its big advantage over surveys is coverage. Only some customers answer a CSAT survey, but every conversation has text. Sentiment can therefore show how customers feel across all conversations, including those who left without rating.

Support teams use it in three ways: in real time, to prioritise or escalate a conversation that is turning negative; in reporting, to see which topics or pages produce frustration; and in quality review, to pick conversations worth reading.

It has real limits. Sarcasm, very short messages, mixed feelings and non-English slang are hard to classify. A customer can also be negative about a delayed parcel while being perfectly happy with the support they received. Treat sentiment as a pointer to conversations, not a verdict on them.

How to measure it

Spot-check classifications by reading a sample, especially in each language you support.

Common mistake

How SuperBot relates

SuperBot's analytics include sentiment alongside resolution rate, CSAT and knowledge gaps.

• SuperBot features

Related terms

• Customer satisfaction score (CSAT)Customer satisfaction score (CSAT) measures how satisfied customers are with a specific interaction, usually from a one-question survey such as “How satisfied were you with this conversation?” on a 1–5 scale. CSAT is the percentage of respondents who choose the top ratings, typically 4 or 5, out of everyone who responded.
• Intent detectionIntent detection is identifying what a customer is trying to achieve from their message, such as tracking an order, requesting a refund or asking about pricing, and labelling it as a category. Chatbots use it to choose a response or route a conversation, and support teams use intent reports to see what customers ask about most.
• Human handoff (escalation)Human handoff is the point where an automated system such as a chatbot passes a conversation to a person, ideally with the full conversation history and context attached. Good handoff rules trigger on low confidence, sensitive topics, explicit requests for a human or signs of frustration, and tell the customer what happens next and how long it will take.
• Conversational AIConversational AI is the set of technologies that let software understand and respond to human language in a back-and-forth dialogue, by text or voice. It combines natural language understanding, dialogue management and response generation, and today is usually powered by large language models. Chatbots, voice assistants and AI support agents are applications of it.