You want customer satisfaction survey questions that people actually answer and you can act on. Ask one rating about a specific moment, add one open follow-up, and keep the scale the same every time.
This page sits within the wider topic of customer experience metrics, and it focuses on the wording of the questions themselves.
| Item | Value |
|---|---|
| Average US customer satisfaction score | 74.4% |
| Suggested CSAT survey length | Two or three questions |
| Standard CSAT rating scale | 5-point scale |
| CSAT questions that name the overall company (Survicate sample) | 32.2% |
| Generic rate when a question names a concrete event or object | 4.5% |
Figures come from SurveyMonkey’s CSAT guide and Survicate’s analysis of CSAT surveys. They can change, so check the sources before you rely on them.
What a customer satisfaction survey should measure
A customer satisfaction survey asks people how well a product, a service or a recent experience met their needs. SurveyMonkey names CSAT, NPS and CES as the three core Voice of the Customer metrics and says they are best tracked together, not alone, in its CSAT guide.
A CSAT survey opens with a multiple-choice question on a 5-point scale, running from very unsatisfied through neutral to very satisfied, as SurveyMonkey explains. The score is the share of happy customers, not an average of the numbers.
To work out the CSAT score, add up the very satisfied (5) and satisfied (4) answers, divide by every response received, and multiply by 100, following the method in the same guide. The result is a percentage you can track from one survey round to the next.
Qualaroo points out that Net Promoter Score is often deployed alongside CSAT programs, but it is a distinct loyalty metric rather than a satisfaction question, in its survey design guide. For a side-by-side comparison, read NPS vs CSAT vs CES.
Why a specific moment beats a brand-wide rating
Survicate studied 835 English-only CSAT surveys across 269 companies. It found that 32.2% of the CSAT questions name the overall company rather than a specific interaction, product feature or touchpoint, according to its write-up.
Small words push a question toward the vague. Of the questions using terms such as “overall” or “in general”, 51.4% fail to name an actual event, feature or interaction, while prompts that target a tangible object or event see the generic rate drop to just 4.5%, per the same analysis.

Survicate suggests a simple test for every question in a program: ask what the team will do differently if the score is low, as set out in its analysis. If nobody can answer, the question probably points at a relationship instead of an experience.
- Name something the customer touched: a delivery, a support call, a checkout or a feature.
- Pick a topic one team owns, so a weak result has a clear destination.
- Rate the experience against expectations rather than mood, because Survicate advises measuring experience against expectations rather than happiness.
Customer satisfaction survey questions you can adapt
SurveyMonkey’s list of CSAT questions groups examples by industry, which shows how a rating can be pinned to a single process. Swap the bracketed parts for your own product or touchpoint.
- Healthcare rating: “How satisfied are you with the appointment scheduling process?”
- Financial services rating: “How satisfied are you with our banking application process?”
- Improvement follow-up: “What is one thing we could do to improve your experience with [company/service]?”
- Open comment: “Is there anything else you’d like to share about your experience with [company/event/service/purchase]?”
The same page advises keeping the scale consistent and adding a simple follow-up to understand what shaped each rating, as SurveyMonkey’s UK guide puts it. Salesforce adds that a neutral follow-up such as “Why or why not?” feels natural whichever way the customer answered, in its survey guide.
Brevity protects response rates. SurveyMonkey’s guidance is to stop at two or three questions: one rating plus a follow-up.
- Choose one moment in the journey, such as a resolved ticket or a delivery.
- Ask a single rating question that names that moment.
- Add one open follow-up, kept neutral.
- Send it soon after the event, while the memory is fresh.
Mistakes that spoil the answers
Most weak surveys fail on wording or length rather than on the platform. Three problems turn up again and again.
- Biased phrasing. Salesforce says that to get accurate answers you should avoid questions that express a bias, and it lists “How much do you love this product?” as an example to avoid, in its guide.
- Surveys that run long. The longer and more challenging a survey is, the more likely people are to exit partway through and leave questions unanswered, per Salesforce.
- Mixed scales. Survicate warns that mixing scales across touchpoints creates fragmented data no team can retrofit or normalise later, in its analysis.
Delivery matters too. Salesforce notes that mailed surveys have the lowest response rates, which could result in nonresponse bias, so phone, in-person or online prompts are often the better fit, as its guide explains.
How do you read the scores you collect?
SurveyMonkey calls the overall US customer satisfaction score of 74.4% a good starting point, and says the American Customer Satisfaction Index keeps a quarterly list of industry benchmarks, in its guide. Treat that figure as context, because your own trend over time is the more useful comparison.
Asking directly also guards against wishful thinking. A SurveyMonkey UK report says half of customer experience professionals believe satisfaction has improved over the last six months, but only 18% of consumers agree, and 53% of people say the situation has actually got worse, according to its CSAT question guide.
If you also run a recommend question, the good NPS score benchmark page explains how to read that separate number. Keep it apart from your satisfaction rating when you report results.
Frequently asked questions
How many questions should a satisfaction survey include?
Show answer
Keep to a pair or a trio: the rating first, then a single open follow-up. SurveyMonkey’s guidance sets the ceiling at two or three questions in total, as described in its CSAT guide.
Should the survey ask about the company overall?
Show answer
Only when you want a loose reading of the relationship. A rating tied to a call, a delivery or a feature gives a team something concrete to fix, which is the point Survicate makes in its CSAT wording study.
Is NPS the same as a satisfaction question?
Show answer
No. Qualaroo describes it as a separate loyalty measure that is often run next to CSAT, so the two should be asked and reported separately, per its survey guide.
How do you turn the answers into a percentage?
Show answer
Count the responses at the top two points of the scale, divide that count by all responses, then multiply by 100. The method is laid out in SurveyMonkey’s guide.


