Customer Satisfaction Surveys: How to Build a CSI and NPS Questionnaire

I have shipped a lot of surveys over the years, and most of the bad ones fail the same way. They ask one fuzzy question, get one fuzzy score, and leave you with nothing you can actually fix on Monday. A number that goes up and down with no explanation is not feedback, it is noise.

So let me walk you through the three tools people mix up: CSAT, CSI, and NPS. I will show you the exact questions I use, when to send them, how to do the math, and how to turn answers into a to-do list. This works whether you run an app, an ecommerce store, or a service business.

What CSAT, CSI, and NPS actually measure

These three get treated as synonyms, and they are not. Getting the difference straight is what makes a survey useful instead of decorative.CSAT (Customer Satisfaction Score) is the narrow one. You ask about a single interaction: a purchase, a support chat, an onboarding flow. Something like "How satisfied were you with your order?" on a 1-5 scale. It is great right after an event, but it only tells you about that one moment.CSI (Customer Satisfaction Index) is the diagnostic one. Instead of a single overall rating, you have people score several specific attributes, then you roll those into one tracked index. The whole point is that you can pinpoint what to fix rather than staring at one vague number.NPS (Net Promoter Score) is the loyalty headline. One question, 0-10, about whether they would recommend you. It is simple, it benchmarks well, and it is the number your leadership will actually remember. On its own it does not tell you why, which is why I pair it with CSI attributes and one open question.

How to write the questions

Here is the structure I use. Keep it short and attribute-specific, because a shorter survey that tells you what to fix beats a long one that gives you a shrug.Start with the NPS core question on a 0-10 scale: How likely are you to recommend us to a friend or colleague? That single line is the backbone of the whole thing.Then add CSI attribute ratings. Give people a clear scale for each attribute. A clean 5-point satisfaction scale reads: Fully satisfied, Satisfied, Neutral, Disappointed, Extremely disappointed. Rate the parts of the experience that matter for your business:

• Ease of ordering or checkout - how simple was it to buy or sign up
• Speed of service or delivery - did it happen as fast as they expected
• Product quality - did the thing do what they wanted
• Order accuracy - did they get exactly what they picked
• Staff or support friendliness - how the humans treated them
• Price-to-value - did it feel worth the money

Add one or two intent questions on a 5-point likelihood scale. "How likely are you to buy from us again in the next 30 days?" and "How likely are you to recommend us to friends?" catch loyalty from a different angle than the raw NPS number.Toss in one behavioral question to segment respondents, like which plan they are on or which item they ordered. That lets you slice scores by group later. Then finish with a single open-ended follow-up: What is the one thing we could do better? One open box, not five. That is where your best fixes come from.

When to send them

Timing decides your response rate and how honest the answers are. Match the survey to the moment.Send CSAT right after the event while it is fresh: post-purchase, at the end of a support chat, a day or two after delivery. Wait too long and people forget the details.Send NPS on a relationship cadence, not after every click. Once someone has had enough experience to form an opinion, roughly 30 days into a subscription or after a second or third purchase, then again on a quarterly rhythm so you can watch the trend.For apps, trigger it after a real success moment, not on first launch. For ecommerce, a few days post-delivery beats the confirmation page. For services, send after a completed project. And cap how often any one person gets asked, because over-surveying quietly kills your response rate.

How to calculate the scores

The math is easy once you see it laid out, so do not let it intimidate you.NPS. Bucket every 0-10 answer. Scores of 9 and 10 are promoters, 7 and 8 are passives, and 0 through 6 are detractors. Take the percentage of promoters, subtract the percentage of detractors, and ignore the passives. So if 50 percent are promoters and 20 percent are detractors, your NPS is 30. The scale runs from -100 to 100, and anything positive means more fans than critics.CSAT. Take the share of people who picked the top boxes (say "Fully satisfied" and "Satisfied") and divide by total responses. If 80 of 100 people landed in those top two, that is 80 percent CSAT.CSI. This is where the index earns its keep. Average the ratings per attribute, then combine them into one number you track over time. Watch each attribute and the blended index together. When the index moves, the attribute breakdown tells you which lever caused it, so you are never guessing.

How to turn results into action

A score you file away is wasted work. The value is in what you do next, and this is the step most teams skip.Start with your lowest attributes. Because CSI is attribute-specific, a weak "speed of service" score points you straight at fulfillment or staffing instead of a vague "do better." Pick the one or two attributes dragging your index down and fix those first.Read the open-ended answers for the pattern behind the numbers. Ten people saying checkout was confusing is a roadmap, not a complaint. Group the comments into themes and count them so you prioritize by volume, not by whoever wrote the angriest note.Close the loop with detractors. Reaching out to unhappy 0-6 responders recovers some of them and shows you are listening, which ties into retention and lifecycle marketing. Then track the index over time. Ship a change, watch whether the score moves next cycle, and keep what works. That feedback loop is the whole point of measuring satisfaction.