Glossary
Sentiment Analysis in Reputation Work
Sentiment analysis is the automated classification of text — articles, posts, comments, transcripts — by tone, typically into positive, negative, and neutral categories. Some tools layer on aspect-based sentiment, emotion detection, or intensity scoring. The core operation does not change: it evaluates how words sound, not what they mean for the entity being covered.
Sentiment analysis
Sentiment analysis is the part of the monitoring stack that scores tone. It reads text, calls it positive or negative or neutral, and rolls those scores up into a number leadership can put on a slide. For product reviews and customer-support tickets, this is fine. For reputation work, it is the metric I trust least.
The reason is structural, not implementation-detail. Tone and meaning come apart in exactly the cases that matter most — a measured wire-service report on regulatory scrutiny that should alarm legal, a sympathetic op-ed whose existence is itself the warning, a calm investigative piece that is quietly building the timeline a prosecutor will use later. The score reads them all wrong, in different directions, for the same reason: it is evaluating the surface of the words. A fuller account of where this breaks and what to use instead is in why sentiment analysis fails during a reputation crisis.
The clarification I would offer to anyone briefing an executive on a sentiment chart: sentiment is a useful question, just not the load-bearing one. Pair it with deviation from the entity's own baseline and a read of the opinion structure, and the chart starts telling you something. Run it alone, and it tells you the color of the words.
Key insight