What is sentiment analysis?
Also called social media sentiment, opinion mining.
Sentiment analysis is sorting comments, mentions and reviews by the feeling they express, usually positive, negative or neutral, so you can see how people feel about a brand, product or topic and how that changes over time.
Why it matters
Volume alone can mislead: a week with twice the usual mentions might be good news or a complaint going round. Sentiment adds the missing context, and a sudden turn negative is often the earliest warning that something needs fixing.
Example
A hypothetical restaurant reads the 80 comments and mentions it received in a month and tags 52 as positive, 20 as neutral and 8 as negative. Positive sentiment is 52 ÷ 80 × 100 = 65%. Six of the eight negative comments mention slow delivery.
How to calculate it
Software does this automatically at scale, but the logic is the same as doing it by hand.
- Collect the mentions, comments or reviews for a set period.
- Label each one positive, negative or neutral. Leave out spam and anything unrelated.
- Divide the number in each group by the total and multiply by 100.
- Optionally work out net sentiment: (positive minus negative) ÷ total × 100. In the example above, (52 − 8) ÷ 80 × 100 = 55.
- Read the negative group in full and note the recurring reasons. That list is more useful than the percentage.
What kinds of tools do it
Sentiment analysis is usually a feature of social listening tools, some customer service and review management tools, and text analysis software. They use language models or word lists to guess the feeling in each piece of text, then chart the results over time.
Automated sentiment works best on large volumes of straightforward text. It struggles with sarcasm, slang, emoji, mixed languages and short replies such as 'wow' that can be read either way, so spot-check a sample by hand before trusting the chart.
Many small businesses don't need a dedicated tool. If you receive a few dozen comments and reviews a month, reading and tagging them yourself is quicker and more accurate than any automated score, and you will notice details software misses.
Common mistakes
Watch out for these:
- Trusting automatic labels without checking. One misread sarcastic comment won't matter; hundreds will.
- Reporting a sentiment score without the reasons behind it.
- Counting your own replies, or a single angry customer's ten comments, as many separate voices.
- Assuming silence is positive. People who are unhappy often just leave.
Using the results
Sentiment is most useful as a trend and a list of reasons. Check it monthly, compare it with the month before, and look for causes: a new product, a price change, a slow week of replies. If one complaint keeps appearing, fixing it in the business usually does more than any post could.
Common questions
Can a small business do sentiment analysis without software?
Yes. Once a month, copy your comments, mentions and reviews into a spreadsheet, add a column for positive, negative or neutral and another for the topic. It takes under an hour for most small accounts and the topics column is where the value is.
How accurate is automated sentiment analysis?
It varies by tool, language and type of text, and no tool reads tone perfectly. Treat automated scores as a trend indicator and check a sample of labels yourself.
What should I do with negative sentiment?
Group the negative mentions by reason, fix what you can (slow delivery, unclear prices, rude replies), and reply publicly and politely where it helps. Then check whether that reason appears less often the following month.