To measure customer sentiment, collect what customers say about you (survey answers, reviews, support tickets, call transcripts and social posts), label each piece as positive, negative or neutral and how strongly, then turn those labels into a customer sentiment score you track over time. Most teams combine two kinds of measure: survey scores such as CSAT, NPS and customer effort, and text and speech analysis that reads the feedback customers give in their own words.
This guide explains what customer sentiment is, the ways to measure it, how to calculate a sentiment score, and how to turn the results into better customer experience. The examples are for consumer brands, but the method works for any business with customers who talk back.

What is customer sentiment?
Customer sentiment is how customers feel about your business, products or service: positive, negative or neutral, and how strongly. It is the emotional tone behind what they say, whether they are pleased, frustrated, disappointed or delighted.
You find it in the feedback customers already give you:
- Live chat and messaging conversations
- Call recordings and transcripts
- Emails and support tickets
- Reviews and ratings
- Survey responses, especially free-text answers
- Social media mentions
Sentiment is different from satisfaction. A satisfaction survey asks a question and gets a number. Sentiment is read from what customers choose to say, often without being asked, which is why it catches problems that surveys miss.
Two ways to measure customer sentiment
| Approach | What it uses | Strength | Limit |
|---|---|---|---|
| Survey scores | CSAT, NPS, customer effort score | Simple, comparable over time | Only hears from people who answer, and only on the question asked |
| Feedback analysis | Tickets, chats, calls, reviews, social posts | Covers every customer who writes or calls, in their own words | Needs tools or careful manual tagging |
Qualtrics’ customer sentiment guide describes survey scores such as CSAT, NPS and CES as lagging measures, and sentiment read from feedback in near real time as a leading indicator of problems. In practice, the surveys tell you the level, and the analysis tells you why.
How to quantify customer sentiment
Customer sentiment can seem like a feeling that is hard to put in numbers. It becomes measurable once you split it into two parts.
Polarity is the direction:
- Positive: the customer is happy or satisfied
- Negative: the customer is unhappy or disappointed
- Neutral: no strong view either way
Intensity is how strongly they feel:
- “I love this product” is strongly positive
- “It’s okay” is mildly positive
- “I hate this” is strongly negative
Combine the two and every piece of feedback gets a score. Qualtrics gives an example intensity scale that runs from +2 for very positive, through +1 for mildly positive and 0 for neutral, down to -2 for very negative. Help desk tools use similar bands. Zendesk’s intelligent triage, part of its Copilot add-on, for example labels each ticket very positive, positive, neutral, negative or very negative, and can report those labels against CSAT ratings.
How to calculate a customer sentiment score
There are two common formulas. Pick one and keep it.
Net sentiment score
Net sentiment = (positive mentions – negative mentions) / total mentions x 100
Example: in a month you receive 1,000 pieces of feedback. 520 are positive, 280 are neutral and 200 are negative. Net sentiment is (520 – 200) / 1,000 x 100 = 32. The score runs from -100 (all negative) to +100 (all positive).
Average intensity score
Give each piece of feedback a value from -2 to +2 and take the average. If 300 comments score +2, 200 score +1, 250 score 0, 150 score -1 and 100 score -2, the total is 600 + 200 + 0 – 150 – 200 = 450, and the average across 1,000 comments is +0.45.
Whichever you use, report it by source and topic as well as overall. A single brand-wide number hides the detail: sentiment on delivery may be falling while sentiment on the product is rising.

Five ways to measure customer sentiment
1. Use sentiment analysis software
Sentiment analysis software reads written feedback, such as reviews, emails, chats and support tickets, and labels the tone of each one using natural language processing and machine learning. Many help desks now include it, and standalone tools pull in reviews and social posts as well.
How to use it well:
- Feed it every channel, not just one, so the score reflects all customers
- Check a sample of its labels by hand each month, because sarcasm, slang and mixed messages trip it up
- Track the trend by week, and look at the tickets behind any sudden drop
For a brand with a lean support team, this is the only practical way to read thousands of messages a month.
2. Send surveys and feedback forms
Surveys give you numbers you can compare over time. The three that matter:
- CSAT asks how satisfied the customer was with one interaction, usually on a 1 to 5 scale. It is typically reported as the share who chose 4 or 5, which is how Zendesk maps its CSAT scale.
- Customer effort score (CES) asks how easy it was to get the issue resolved.
- NPS asks how likely the customer is to recommend you, covered below.
Always add one free-text question. The comment is where the sentiment is. A 4.5 out of 5 average looks strong, but if the comments keep saying “slow reply”, you know exactly what to fix.
3. Monitor social media and reviews
Customers talk about brands on Instagram, Facebook, X, TikTok, Reddit and review sites, often without tagging you. Social listening tools search for mentions of your brand, products and competitors, and classify the tone of each one.
The calculation is the net sentiment formula above: if 80 of 100 mentions are positive and 10 are negative, net sentiment is +70. Watch it closely around product launches, price changes and shipping delays, when it moves fastest. There is no universal benchmark here, because what is normal differs by industry and platform. Compare against your own history.
4. Net Promoter Score (NPS)
NPS measures loyalty with one question. As Bain’s Net Promoter System defines it, customers answer “How likely are you to recommend us to a friend or colleague?” on a scale of 0 to 10:
- Promoters (9 to 10): loyal and likely to recommend
- Passives (7 to 8): satisfied but open to competitors
- Detractors (0 to 6): unhappy and likely to say so
NPS = % promoters – % detractors, which gives a score from -100 to +100. NPS is a relationship measure, so run it every quarter or so rather than after every contact.
5. Text analytics, topic modeling and call analysis
Topic modeling groups feedback by subject, such as “delivery time”, “product quality” or “refunds”, and then measures the sentiment within each group. That turns a pile of comments into a ranked list of what customers are happy or unhappy about.
Speech analytics does the same for phone calls. It transcribes calls and flags negative tone, silence, interruptions and phrases like “cancel” or “complaint”, so a quality team can listen to the calls that matter instead of a random sample.
Example: a brand runs topic modeling on a month of reviews and finds that most negative comments mention late delivery. It changes carrier on its slowest routes, and tracks whether delivery sentiment recovers over the next two months.
How to set up a customer sentiment program
- List your sources. Tickets, chats, calls, surveys, reviews and social mentions.
- Choose one score and one formula. Net sentiment or average intensity, applied the same way to every source.
- Tag by topic. Use the same topic list across channels so you can compare.
- Set a baseline. Measure for a month or two before setting targets.
- Review weekly, act monthly. Look at the biggest negative topics every week, and assign each one an owner.
- Close the loop. Tell customers when their feedback led to a change, and watch whether sentiment on that topic improves.
How to improve your customer sentiment score
- Fix the top negative topic first. For consumer brands it is often delivery, refunds or reaching a person.
- Respond fast to negative feedback, in public on reviews and social, and privately to resolve the issue.
- Coach agents on low-sentiment contacts. Pair sentiment data with quality assurance reviews so coaching targets how customers felt, not just whether the script was followed.
- Make it easy to reach a person when automation cannot solve the problem.
- Share wins. Positive comments that name agents are good for morale and show what good looks like.
Customer sentiment analysis with Atidiv
Atidiv runs outsourced customer support for consumer brands in the US, UK and Australia across phone, chat, email, messaging and social. Dedicated QA analysts review tickets and calls against scorecards customized to your brand standards, with weekly coaching, and our NPS services run from survey design to analytical dashboards.
- 98% average QA score and a 4.8 average CSAT
- 200,000+ customer experiences delivered
- Support in 11+ languages
- 30 to 80% of repeat queries automated, with QA guardrails

Explore our customer experience service, including social media support for brands that need mentions and messages answered. To talk it through, contact us.
Frequently asked questions
What is customer sentiment?
Customer sentiment is how customers feel about a brand, product, service or experience: positive, negative or neutral, and how strongly. It shows up in reviews, surveys, support conversations, calls and social posts.
What is customer sentiment analysis?
Customer sentiment analysis is the process of collecting that feedback and labeling its tone, usually with natural language processing software, to find patterns. Instead of reading isolated comments, you see which topics customers feel strongly about and whether that is improving.
How do you calculate a customer sentiment score?
The simplest method is net sentiment: positive mentions minus negative mentions, divided by total mentions, times 100. Another is to score each piece of feedback from -2 to +2 and take the average. Use one method consistently.
How do I know if customers are unhappy if they do not complain directly?
Many unhappy customers never complain directly. Monitor reviews, social media and survey comments, and run sentiment analysis on support tickets and chats to catch negative tone early.
I do not have time to read every review. What can I do?
Use sentiment analysis software, which reads and labels feedback automatically, groups it by topic and summarizes the trend. Check a small sample of its labels by hand each month.
How often should I measure customer sentiment?
Track it weekly from tickets and social mentions, and review it monthly. Check it after big events too, such as a product launch, a price change, a new campaign or a change in your support team.
Can I measure customer sentiment on a small budget?
Yes. Start with a one-question survey with a comment box, a spreadsheet of tagged reviews and your help desk’s built-in reporting. Add dedicated tools once volume makes manual tagging impractical.
Ayushi leads Customer Experience services at Atidiv with a strategic/operations-focused mindset. Her primary objective is to increase how well businesses deliver service and retain customers. She evaluates customers' journeys through marketing impact, performance metrics, and gaps to develop improved systems and processes. With a reputation for curiosity and structured thought processes.