Sentiment analysis for brands: why tone is the wrong question

Most sentiment tools tell you whether a sentence sounds happy. A brand needs to know whether it is good or bad news. How to measure the second, and how to read the numbers.

By Commently · Published November 2, 2026 · 4 min read

What social media sentiment analysis is

Sentiment analysis classifies a post or a comment as positive, neutral or negative. Brands use it to answer questions like: how do people feel about the launch, did the campaign land, is a complaint spreading?

The method decides whether the answer is useful. Classic sentiment models score the tone of the text. A brand needs a different question answered: is this good or bad for us?

Tone versus brand sentiment

A tone model gets these wrong in the direction that matters: sarcasm counts as praise, comparisons are split down the middle, and praise for a creator inflates your positive share.

Where sentiment numbers go wrong

• Irrelevant mentions. If a mention is not about your brand, its sentiment does not belong in your numbers. Check relevance first.
• The account's own replies. A brand that thanks every commenter fills its comment section with its own positive messages. Leave the post owner's replies out.
• Small samples. Eight positive comments out of ten is not "80% positive" in any useful sense. Commently's campaign health score leaves sentiment out entirely when a campaign has fewer than ten analysed comments.
• Languages and slang. Translating first and scoring second loses sarcasm and slang. Read each comment in the language it was written in.
• Overwritten corrections. When your team corrects a label, the next AI run must not undo it. Otherwise the team stops correcting, and the error stays.

Beyond positive and negative: category and priority

Sentiment gives you a direction. Two more labels tell you what to do:

• Category, from your own list: delivery, price, product quality, customer service, the campaign itself. "Negative" becomes "negative about delivery", and that has an owner.
• Priority, from critical to low. One comment about an allergic reaction matters more than a hundred about the packaging colour.

Without categories and priorities, a sentiment dashboard tells you something is wrong. With them, it tells you what, and how urgently.

How to read sentiment

• Read shares, not counts. A bigger campaign gets more negative comments simply because it gets more comments.
• Compare with your own baseline. A negative share that is normal for an airline can be alarming for a skincare brand.
• Read the comments behind the number. Open the ten most negative comments in a category before you report on it.
• Separate your posts from your creators'. Audiences react to the creator as much as to the product.

Choosing a sentiment analysis tool

• Does it score sentiment for the brand, or the tone of the text?
• Does it check relevance before it counts?
• Does it read comments, and on which platforms?
• Can you define your own categories?
• Does it assign a priority?
• Do your team's corrections survive the next run?
• Which languages does it read natively?

How Commently does it

Every mention is first checked for whether it is actually about your brand. Then mentions, and every comment under your and your creators' posts on Instagram, TikTok, YouTube and X, are read by AI for three things: whether they are good or bad for the brand, which of your categories they belong to, and how urgent they are. Your team's corrections are never overwritten. Commently Agent summarises what people are saying and what to do next, on request, for the filters you chose. Comment and sentiment analysis in Commently