By Apercu team ·
Thematic Analysis vs Sentiment Analysis: Which One Do You Need?
Thematic analysis shows what people talk about. Sentiment analysis shows how they feel. Learn the difference, how sentiment scores work and when to use each.
Quick answer: thematic analysis finds what people talk about, such as delivery, pricing or onboarding, and how often. Sentiment analysis measures how they feel, usually as positive, negative or neutral. Themes tell you what to fix. Sentiment tells you how bad it is. For most feedback projects, start with themes, because "34 percent of customers mention slow delivery" is actionable and "sentiment is 0.2" is not.
The difference in one example
Take three comments:
- "Delivery took three weeks, really disappointing."
- "Support sorted my refund in an hour, brilliant."
- "Love the product but delivery was slow."
| Method | Output |
|---|---|
| Sentiment analysis | 1 negative, 1 positive, 1 mixed |
| Thematic analysis | Delivery speed (2 comments), Customer support (1), Product quality (1) |
| Both combined (aspect-based) | Delivery speed: negative. Customer support: positive. Product: positive. |
Sentiment alone says feedback is "mostly mixed". Themes say delivery is the problem.
Side-by-side comparison
| Thematic analysis | Sentiment analysis | |
|---|---|---|
| Question it answers | What are people talking about? | How do people feel? |
| Output | Named themes with counts and examples | Labels or scores, such as positive, negative or -1 to 1 |
| Best for | Prioritizing fixes, understanding reasons, research | Tracking mood over time, flagging angry messages |
| Actionability | High: each theme points to an owner | Low on its own: needs a topic to act on |
| Typical risks | Themes too broad or too many | Sarcasm, mixed comments, domain words like "sick" or "killer" |
How sentiment scores are calculated
If you are evaluating sentiment tools, it helps to know what the score means:
- Lexicon-based scoring adds up positive and negative words from a dictionary. It is simple and transparent, but misses context and sarcasm.
- Machine learning classifiers are trained on labeled examples and predict positive, negative or neutral. They handle context better, but depend on training data that matches your domain.
- Large language models can label sentiment and explain why, with good accuracy on mixed comments, at a higher cost per comment.
Scores are usually reported as a share of positive and negative comments, an average on a -1 to 1 scale, or a net sentiment score: the percent positive minus the percent negative. Whatever the method, check a sample of labeled comments by hand before trusting the numbers.
How to choose a sentiment analysis tool
Bring these questions to any vendor demo:
- Does it score sentiment per topic or per comment? Per-topic (aspect-based) sentiment is far more useful for mixed comments.
- How does it handle mixed and neutral comments? Ask for examples from your own data.
- Can you see why a comment got its label? Unexplained scores are hard to defend.
- How accurate is it on your domain? Test 100 of your own comments and compare with a human reading.
- Which languages are supported, and is quality the same in each?
When to use thematic analysis
- You need to know what to fix first
- You are analyzing open-ended survey questions, reviews or support tickets
- You want counts and quotes for a report or a roadmap discussion
- You are doing qualitative research and need a code frame
When to use sentiment analysis
- You monitor large, continuous streams such as social mentions or chat
- You want an early warning when mood shifts
- You need to route angry messages to a human quickly
How to get sentiment insight from a thematic analysis
Many questions already carry the sentiment. Answers to "What could we improve?" are complaints, and answers to "What do you like most?" are praise. When the question is neutral, split the data first:
- NPS: analyze detractors, passives and promoters separately. See NPS comment analysis.
- Reviews: split by star rating, or by likes and dislikes on sites like G2. See how to extract themes from G2 reviews.
- Surveys: split by a satisfaction score if you asked one.
Comparing themes across those groups shows what drives positive and negative experiences, which is often what people want from sentiment analysis in the first place.
Where Apercu fits
Apercu does thematic analysis: it groups responses into named themes with counts, percentages and examples. It does not score sentiment. If you need both, split your data by score or rating as described above, or use a dedicated sentiment tool alongside it.
FAQ
Is thematic analysis the same as sentiment analysis?
No. Thematic analysis identifies the topics in text and how often they come up. Sentiment analysis classifies the emotional tone of text as positive, negative or neutral.
Which is better for customer feedback?
Thematic analysis is usually more useful because it shows what to fix. Sentiment adds intensity and helps with monitoring. The most useful view combines them: sentiment per theme.
What is aspect-based sentiment analysis?
It scores sentiment separately for each topic in a comment. "Great product, slow delivery" becomes positive for product and negative for delivery, instead of one mixed score.
How do you measure a sentiment score?
Classify each comment as positive, negative or neutral (or score it from -1 to 1), then report the share of each class or a net score: percent positive minus percent negative. Validate the classifier on a sample of your own comments first.