Social media terms

Sentiment analysis

Sentiment analysis is the process of sorting comments, mentions and messages by the feeling behind them, such as positive, negative or neutral.

What is sentiment analysis?

Sentiment analysis looks at what people write and judges the mood of it. A comment like "Love the new update, it saves me so much time" is positive. "Still waiting for a reply after two days" is negative. "Does this come in blue?" is neutral. Doing this by hand works for a few comments. For hundreds or thousands, teams use software that reads the text and labels it automatically.

On social media, sentiment analysis is often part of social listening and inbox tools. It helps you see how people feel about your brand, a product, a campaign or a topic, and how that feeling changes over time.

How sentiment analysis works

Modern tools use AI language models trained on large amounts of text. They read each message, look at the words, the order and the context, and give it a label or a score. A score might run from very negative to very positive, with neutral in the middle. Some tools go further and spot specific feelings, such as anger, joy or confusion, or tie the mood to a topic, like shipping, price or support.

Older tools counted good and bad words. They were fast but easy to fool. Today's models handle context better, though they still make mistakes, which is why a person should always check the results.

Once each message has a label, the tool adds them up. You see a split for a post, a week or a campaign, like the one below.

Mood of comments on a launch post

600 comments labeled by a sentiment tool

  • Positive55%
  • Neutral30%
  • Negative15%
PartShare
Positive55%
Neutral30%
Negative15%

Why sentiment analysis matters

  • Spot problems early: a jump in negative mentions can warn you about a bug, a delay or a bad ad.
  • Sort your inbox: angry or urgent messages can rise to the top so someone replies first.
  • Measure campaigns: see whether people liked a launch, beyond how many talked about it.
  • Compare with competitors: check how the mood around your brand stacks up.
  • Find fans: very positive comments point to customers who might share a review or join a program.

How to use sentiment analysis, step by step

  1. Decide what to track: your brand name, products, campaigns or industry topics.
  2. Collect the messages with a listening or inbox tool, so comments, mentions and messages from every channel land in one place.
  3. Label them with the tool, or with an AI assistant for a smaller batch.
  4. Review the trend each week, as well as single posts. A steady week matters more than one loud comment.
  5. Read a sample from each group to check the labels make sense for your audience.
  6. Act on what you find, such as fixing a common complaint or thanking happy customers.

For example, a food delivery app might see negative comments jump from about one in ten to one in four over two weeks. Reading a sample shows most of them mention late orders in one city. The team fixes the driver shortage there, posts an update and watches the share of negative comments fall again.

Share of negative comments by week

Before and after fixing late orders

0%12.5%25%11%Week 1Week 2Week 3Week 4Week 5Week 6
Negative share
Week 110%
Week 212%
Week 325%
Week 424%
Week 515%
Week 611%

How to measure sentiment

The simplest measure is the share of messages in each group: positive, neutral and negative. Some teams also track net sentiment, which is the positive share minus the negative share. Whatever you pick, use the same tool and the same rules every time, or your trend will jump for reasons that have nothing to do with your customers.

It also helps to split sentiment by topic. Overall mood may look fine while one topic, like returns, keeps getting worse.

Limits of sentiment analysis

Software can misread sarcasm. "Great, my order is late again" uses a positive word with a negative meaning. Slang, emoji, inside jokes, other languages and mixed messages are hard too. "The product is amazing but support is awful" holds two moods at once.

Treat sentiment scores as a guide and always have a person read the messages that matter most, especially complaints and anything that could become a crisis. Never reply to a customer based on a label alone.

Common mistakes

  • Trusting the labels without ever reading the messages behind them.
  • Panicking over one bad day instead of watching the weekly trend.
  • Counting only positive and negative and ignoring the neutral group, where many questions hide.
  • Changing tools and comparing the new numbers with the old ones.

Tools that help

Questions. Answers.

Something else? Write to hello@capybaroar.com.

What is an example of sentiment analysis?

A tool reads the comments on a product launch post and labels most of them positive, a few negative and the rest neutral. You then read the negative ones to see what went wrong.

How accurate is sentiment analysis?

It works well on clear, simple messages but can miss sarcasm, slang and mixed feelings. Check a sample by hand to judge how well it fits your audience.

Is sentiment analysis the same as social listening?

No. Social listening is gathering and studying what people say about a topic. Sentiment analysis is one part of it that judges the mood of those messages.

What is net sentiment?

Net sentiment is the share of positive messages minus the share of negative ones. It turns the split into one number you can track over time.