Sentiment

What is Sentiment?

Sentiment analysis turns text into insights about how people feel based on what they say or write. It looks at words and phrases to decide whether the overall emotion or attitude is positive, negative, or neutral.

For example:

• “The service was fantastic!” = positive sentiment
• “I’m really disappointed with this product.” = negative sentiment
• “It arrived on time.” = neutral sentiment

Why is Sentiment useful?

• Sentiment helps you understand customer feedback quickly without reading every transcript.
• It can show trends over time, like whether customers are becoming happier or more frustrated.
• It helps identify problem areas (e.g. lots of negative stated inquiries about delivery)

OI determines the average sentiment score from the customer’s side of the interaction. Sentiment is analyzed on only the section of the transcript where the customer says why they are contacting. This is a measurement of how positive, negative or neutral the customer is when initially contacting. It serves as a valuable “temperature check” as it relates to their reason for contact.

Note: As we analyze sentiment only on the stated reason for contacting, sentiment in OI does not reflect an agent’s performance and typically cannot be influenced by an agent. For a similar metric more suitable for measuring performance, see Predicted Satisfaction.
 

Interpreting Sentiment scores

This score is between -1 (negative sentiment) to +1 (positive sentiment), with zero as neutral sentiment.

Things to keep in mind

• Context matters: A score near zero might mean the text is neutral, or it could mean there’s both positive and negative language.
• Comparisons are most useful: Instead of focusing on one score, it’s often more meaningful to compare scores across time, agents, root causes or inquiries.
• Punctuation matters: Sentiment analysis is performed per sentence, meaning that sentence length matters. Longer sentences often result in neutral sentiment scores.
  • In chat or email interactions, when a customer writes a long block of text without punctuation (such as full stops or question marks), the sentiment model interprets it as one continuous sentence. Because longer sentences tend to balance positive and negative words, this can lead to a more neutral sentiment score. This applies to all sentiment analysis models.