Customer Feedback Survey: 40 Questions, 5 Templates, and Where to Send It So People Actually Answer

A customer feedback survey is a short set of questions sent after a specific moment: a purchase, a support reply, a cancellation. Most use one scale question, CSAT 1 to 5 or NPS 0 to 10, plus one open follow-up asking why. Keep it to two questions and send it where the conversation already happened.

Everything below is the working version of that: a bank of 40 questions grouped by the moment they belong to, five templates you can copy, the two formulas written out, an honest read on which channel gets answers, and the follow-up workflow that decides whether the whole exercise was worth doing. Every external number in this post comes with the source we opened and the month we opened it. Where we could not verify a number, we say so and leave it out.

If you want to skip the theory, the Pinlyx surveys page shows what an NPS or CSAT send looks like when it runs from the same place your contacts and conversations already live.

What a customer feedback survey actually is

A customer feedback survey asks people who have already dealt with you what that was like. That single word, already, is what separates it from market research. You are not asking strangers what they might want. You are asking someone who bought, or asked for help, or canceled, about a thing that actually happened to them, while they still remember it.

That has two practical consequences and both of them shape everything else in this post.

First, the answer is only as good as the memory. A survey sent three weeks after a support ticket is not measuring that ticket. It is measuring a vague impression of your brand, mixed with whatever else happened that month. Send it while the memory is specific or do not send it.

Second, the answer belongs to a person you can identify. This is the part most survey tools quietly break. If the response arrives as an anonymous row in a dashboard, you have bought a number. If it arrives attached to a customer record, with their order, their last ticket and the conversation thread still open next to it, you have bought a to-do list. The number is the cheap half.

So a useful definition is narrower than the one you will find elsewhere: a customer feedback survey is a short, timed question attached to a named customer and a specific event, sent on a channel they already use, with somebody responsible for what comes back.

The four customer satisfaction survey types you will actually use

There are dozens of named metrics. In practice small teams use four, and mixing them up is the most common reason a feedback program produces numbers nobody trusts.

CSAT: the one to start with

CSAT asks how satisfied somebody was with one specific thing, on a five point scale, and the score is the share of answers in the top band. It is the easiest metric to collect honestly because the question is obvious to the person answering it. They just had an experience. You are asking about that experience. Nobody has to interpret anything.

Its weakness is that it is local. A great CSAT tells you your support replies feel good. It tells you nothing about whether your product is worth paying for next year. Teams with excellent CSAT and quiet cancellations are a recognized pattern, and the explanation is usually polite support wrapped around a product problem. Our CSAT glossary entry works through the formula, the confidence interval and the way neutral answers get dropped from the denominator by accident.

NPS (Net Promoter Score): the relationship number

NPS asks one question, how likely you are to recommend, on a zero to ten scale, and subtracts the share of low answers from the share of high ones. Bain, which created the metric, defines it plainly: promoters score 9 or 10, passives 7 or 8, detractors 0 to 6, and the score is "the percentage of customers who are promoters minus the percentage who are detractors" (Net Promoter System, checked September 2026).

The number is lossy by design. It throws away the difference between a 2 and a 6 and between a 7 and an 8. What it buys you is a single figure a whole company can hold in its head, and a clean definition of who to call. The NPS glossary entry has the variance math, which matters more than people expect: quarterly NPS moves on a few hundred responses are usually noise.

CES: the one for processes, not feelings

CES, the customer effort score, asks how hard the customer had to work. It is the right question when you are fixing a process rather than judging a person: a returns flow, an onboarding, a password reset. A low CSAT tells you somebody is unhappy. A high effort score tells you where.

An honesty note. CES is usually introduced with a set of dramatic figures about low-effort customers repurchasing at 94% and high-effort customers spreading negative word of mouth at 81%. Those figures are attributed to Gartner research and they appear on dozens of vendor blogs. We could not open a primary Gartner source for them in September 2026, so we are not going to repeat them as facts. Use CES because the mechanism is obvious, not because of a statistic you cannot check.

Open feedback: where the actual answer is

Scale questions tell you that something moved. Only text tells you what. The single highest-value output of most feedback programs is not the score, it is the list of named people who said something specific, arriving fast enough that somebody can reply to them the same day.

This is why nearly every template below is a scale question plus one open follow-up. The scale gets you the trend. The text gets you the fix.

When to send it, and the timing windows that matter

Timing is the cheapest lever in this whole post. The same question, sent at two different moments, produces different response rates and different answers, and the mechanism is simply whether the person still remembers.

Four moments are worth surveying, and they want different metrics.

• After a purchase. Send within 24 to 48 hours of delivery, not of payment. What you want to know is whether the thing that arrived matched the page that sold it, and until it arrives they cannot tell you. For a digital product, send after first use rather than after checkout.
• After a support conversation. Send immediately after the ticket closes, in the same thread if the channel allows it. Same day is the rule. If the case needed a follow-up, send after the follow-up, otherwise you are asking about half a resolution.
• Thirty days into a subscription. This is the onboarding checkpoint and it is the one most teams skip. At day 30 the customer either got to the outcome they signed up for or they did not, and they know which. The answer is still cheap to act on because they have not yet decided to leave.
• After cancellation. Send within a day, and keep it to one reason question plus one open box. You are not trying to win them back in a survey. You are trying to find out which of the six repeatable reasons this was.

There is one sourced timing number worth knowing, and it is about the day rather than the hour. SurveyMonkey analyzed 100,000 surveys and reports that customer surveys sent on a Monday "received 10% more responses than average, and surveys sent out on Fridays received 13% fewer responses than average" (SurveyMonkey, checked September 2026). Treat that as a tiebreak for a scheduled quarterly send, not as a reason to delay a post-support survey until Monday. Event-triggered surveys should follow the event.

Beyond that, be suspicious of precise timing claims. We looked for a credible cross-industry study on the optimal delay between an interaction and a survey and did not find one that publishes its method. The defensible rule is the one that follows from how memory works: closer is better, and same-channel is better than a new channel.

The 40 question bank, grouped by moment

Pick one or two of these per send. The temptation to use eight is the single most reliable way to get zero answers, and the next section explains why with numbers. Each question is tagged with its answer format: scale, yes/no, choice or open text.

Post-purchase (8 questions)

• How satisfied are you with your order? (scale, 1 to 5)
• How easy was it to place your order? (scale, 1 to 5)
• Did your order arrive when we said it would? (yes/no)
• Was the item what you expected from the product page? (yes/no)
• How likely are you to buy from us again? (scale, 0 to 10)
• How would you rate the packaging and the condition it arrived in? (scale, 1 to 5)
• What almost stopped you from ordering? (open text)
• Is there anything we could have done better? (open text)

Post-support (8 questions)

• How satisfied are you with the help you received? (scale, 1 to 5)
• Was your issue resolved? (yes/no)
• How much effort did you have to put in to get this sorted? (scale, 1 to 5, where 1 is very little)
• Did you have to explain your problem more than once? (yes/no)
• How would you rate how quickly we first replied? (scale, 1 to 5)
• Did we answer in language you could understand? (yes/no)
• What would have made this easier? (open text)
• If anything is still outstanding, what is it? (open text)

Product and feature (8 questions)

• How satisfied are you with the product overall? (scale, 1 to 5)
• How often do you use it? (choice: daily, weekly, monthly, rarely, never)
• How disappointed would you be if you could no longer use it? (choice: very, somewhat, not at all)
• How easy is it to do the thing you came here to do? (scale, 1 to 5)
• Which of these would you use most? (choice, from a short list of real options)
• What is the one thing that would make this more useful to you? (open text)
• Is there anything that regularly gets in your way? (open text)
• What did you use before us, if anything? (open text)

Onboarding and the first week (6 questions)

• How easy was it to get set up? (scale, 1 to 5)
• Did you manage to do the thing you signed up to do? (yes/no)
• How clear were the setup instructions? (scale, 1 to 5)
• How likely are you to recommend us so far? (scale, 0 to 10)
• Where did you get stuck, if anywhere? (open text)
• What were you hoping to do that you have not managed yet? (open text)

Churn and exit (5 questions)

• What is the main reason you are leaving? (choice: price, missing feature, not using it, switched to something else, one-off need, other)
• How satisfied were you overall? (scale, 1 to 5)
• Would you consider coming back? (yes/no)
• Is there anything we could have done to keep you? (open text)
• What are you using instead, if anything? (open text)

Open-ended follow-ups (5 questions)

These are the second question, the one that appears after somebody picks a number. Pick one and keep it.

• What is the main reason for your score? (open text)
• What is the one thing we should change first? (open text)
• What nearly made you give us a lower score? (open text)
• What should we keep exactly as it is? (open text)
• Is there anything else you want to tell us? (open text)

Note what is missing from all 40. There is no question asking people to rate your brand values, no five-point grid of adjectives, and nothing that asks the customer to do your segmentation for you. If the answer will not change something you do next month, it does not belong in a survey a customer has to read.

How to ask so people actually answer

Response rate is mostly decided before anyone reads the question. Length, channel and timing do more work than wording. But wording decides whether the answers mean anything, and there is good evidence on that.

One question first, then an optional follow-up

The structure that works is a single scale question that can be answered in one tap, followed by an open box that is genuinely optional. The first question is the survey. The second is a bonus from the people who care.

The reason is arithmetic. SurveyMonkey's analysis of roughly 100,000 surveys reports that respondents spend about 75 seconds on the first question, 40 seconds on the second, and roughly 30 seconds each on questions three through ten, with completion rates "dropping anywhere from 5% to 20%" once a survey passes seven or eight minutes (SurveyMonkey, checked September 2026). A separate SurveyMonkey study, covering 100,000 surveys deployed between January 2009 and September 2010, found the sharpest increase in abandonment happens with each additional question up to about 15 questions (SurveyMonkey, checked September 2026). That second study is old, and we are citing it for the shape of the curve rather than for a precise figure.

Nielsen Norman Group reached the same place from usability research back in February 2004, recommending short surveys and noting that in 13 of 14 case studies the single recommendation question predicted loyalty as well as a longer questionnaire (NN/g, checked September 2026). The advice is old enough to have been proven by everyone since.

Plain language, and no leading questions

Pew Research Center runs wording experiments on identical populations, and the size of the effects should make anybody careful about how they phrase a satisfaction question. In one experiment, 68% supported military action, but adding a clause about potential casualties dropped support to 43%. In another, 51% favored "making it legal for doctors to give terminally ill patients the means to end their lives" while only 44% favored "making it legal for doctors to assist terminally ill patients in committing suicide", which is the same policy in different words (Pew Research Center, checked September 2026).

Two more findings from the same source translate directly to customer surveys. Question order changes answers: satisfaction questions asked after a related question produced 88% dissatisfied against 78% when asked cold. And open versus closed changes what you hear: 58% picked the economy as the top issue when it was offered in a list, but only 35% volunteered it when asked in an open box, while 43% gave an unlisted answer in the open version against 8% in the closed one.

The practical rules that fall out of this:

• Do not put your preferred answer in the question. "How much did you enjoy our fast delivery?" is not a question, it is a press release with a scale attached.
• Ask about one thing. "How satisfied are you with our price and support?" cannot be answered by anyone who feels differently about the two.
• Use the customer's words, not your internal ones. Nobody outside your company knows what your onboarding flow is called.
• Balance the scale. Label both ends, and make the midpoint genuinely neutral.
• Offer the open box after the number, never before. Asking for text first makes the whole thing look like homework.
• Do not ask a question whose answer you will ignore. Every ignored question costs you responses on the ones that matter.

Mobile first, because that is where it opens

Assume the survey will be read on a phone, in a notification, with one thumb. That means the scale has to be tappable rather than a dropdown, the question has to fit above the fold without scrolling, and the open box has to be skippable. If your survey needs a pinch-zoom, the response rate question has already been answered.

Five templates you can copy and paste

Each of these is one scale question, one follow-up and a thank-you. Replace the bracketed parts. The thank-you matters more than it looks: it is the only moment the customer finds out whether anyone is on the other end.

1. Post-purchase CSAT

Subject or opening line: Quick question about your order

Question: How satisfied are you with your order? (1 = not at all, 5 = very satisfied)

Follow-up: What is the main reason for your score?

Thank-you: Thanks. A real person reads every one of these, and if anything went wrong with this order just reply here and we will sort it.

2. Post-support CSAT

Subject or opening line: Did we actually fix it?

Question: How satisfied are you with the help you received? (1 = not at all, 5 = very satisfied)

Follow-up: If anything is still outstanding, what is it?

Thank-you: Thanks. If you scored us 1 or 2, someone will be in touch today rather than next week.

3. Quarterly NPS

Subject or opening line: One question, 10 seconds

Question: How likely are you to recommend [company] to a colleague? (0 = not at all likely, 10 = extremely likely)

Follow-up: What is the main reason for your score?

Thank-you: Thank you. We read every comment, and the ones that come with a low score get answered first.

4. Product feedback

Subject or opening line: What should we build next?

Question: How easy is it to do the thing you came here to do? (1 = very hard, 5 = very easy)

Follow-up: What is the one thing that would make this more useful to you?

Thank-you: Thanks. We cannot promise to build everything, but we can promise this gets read by the people who decide.

5. Exit survey

Subject or opening line: Before you go, one question

Question: What is the main reason you are leaving? (price, missing feature, not using it, switched to something else, one-off need, other)

Follow-up: Is there anything we could have done to keep you?

Thank-you: Thanks for telling us. No sales follow-up, we just wanted to know. The door is open if you come back.

That last line is not decoration. An exit survey that triggers a retention call is a trap, people learn about it quickly, and after that your cancellation reasons all read "other".

Where to send it: email, the chat you were already in, or a public link

Channel is the second biggest lever after length, and it is the one teams pick by habit rather than by evidence.

SurveyMonkey publishes response rates by collector type from its own 2025 platform data. The spread is large: email invitations at 49.17%, web links at 29.95%, in-app mobile surveys at 34.37%, SMS at 18.54%, and website popups at 3.65%, with customer feedback surveys as a category averaging 31.81% (SurveyMonkey, checked September 2026). Read those as one platform's own customers rather than as a law of nature, and note the page does not state a dataset size. The ranking is still informative: an invitation sent to somebody who already knows you beats an interception of a stranger by an order of magnitude.

Why in-conversation wins for support

The mechanism is boring and strong. The customer has just been talking to you in a thread. The survey arrives in the same thread, from the same name, about the thing they were just discussing. There is no new app to open, no context to reconstruct, and no doubt about who is asking.

Compare that to the normal setup, where a support agent closes a ticket in one tool and a survey platform sends an email from a no-reply address a day later, referencing a ticket number the customer never knew. Every step of that is a place to lose somebody.

The same logic explains why chat ratings work: the rating request appears at the end of a conversation the person is still in. If you run a live chat widget, asking for a rating as the chat ends catches people while they are still there, which is also why that number belongs next to reply and resolution times rather than in a separate report. Our post on live chat benchmarks goes through what those operational numbers actually mean.

The messaging channel caveats

Sending a survey into WhatsApp or Telegram is effective and it is not unconstrained.

On WhatsApp, Meta's documentation is explicit: "When a WhatsApp user messages you or calls you, a 24-hour timer called a customer service window starts", and "When the window closes, you can only send pre-approved template messages" (Meta, checked September 2026). In practice that means a WhatsApp survey is something you send to someone you were just talking to, inside the day. It is not a broadcast tool, and anyone selling you one is selling you a policy problem. Our customer service window entry covers the mechanics, and the WhatsApp CRM page covers what a compliant setup looks like.

Telegram is looser about windows but not about behavior: an unsolicited survey to people who never talked to you is spam regardless of what the API permits. The rule that keeps you safe on every channel is the same rule that gets you answers, which is to only survey people who recently interacted with you.

How to calculate CSAT, NPS and CES

Three formulas, and one mistake each that quietly inflates the result.

CSAT

CSAT % = (responses in the top band / all responses) x 100

On a five point scale the top band is normally 4 and 5. If 120 people answer and 84 pick 4 or 5, CSAT is 84 / 120 x 100 = 70%.

The mistake: dropping neutral answers from the denominator. If you count only the people who picked 1, 2, 4 or 5 and ignore the 3s, the score rises without a single customer changing their mind. Keep every response in the denominator.

NPS survey template

NPS = %Promoters - %Detractors

Promoters answer 9 or 10. Passives answer 7 or 8. Detractors answer 0 through 6. Passives count in the denominator and contribute nothing to the result. If 200 people answer, 90 are promoters and 40 are detractors, then 45% - 20% = +25.

Two mistakes. The first is excluding passives from the denominator, which is the same inflation trick as above and can move the score by ten points or more without a single customer changing their mind. The second is writing the result with a percent sign. NPS is the difference between two percentages and runs from -100 to +100, so it is "+25", not "25%". That matters because a percent sign invites people to compare it against CSAT, which is a real percentage measuring something else.

CES

CES = average of all effort ratings, or the share of low-effort answers, depending on which convention you pick.

The mistake here is inconsistency. Some teams score 1 as low effort and some score 1 as high effort, and if you change convention halfway through a year your trend line is meaningless. Pick one, write it down, and label both ends of the scale in the question itself.

What a good score looks like, with sources

Short answer: benchmarks vary enormously by industry, by country, by how you asked and by who you asked, and most published league tables are marketing. Compare your score against your own score last quarter, on the same population and the same question.

That said, here are three published reference points you can actually open and check, as of September 2026.

Note how wide the NPS range is. An industry scoring 26 and an industry scoring 68 are not eight times apart in quality, they are different businesses with different customers and different scale habits. It is also well established in survey methodology that respondents in different countries use the extremes of a numeric scale differently, which means a multi-market NPS moves when your market mix moves, before anything about your service changes.

The one benchmark that is genuinely yours is direction. Same question, same population, same channel, four quarters in a row. If that is going up, it is going up.

Closing the loop, which is the part that pays for the survey

A feedback program that produces a chart and no conversations has kept all of the cost and thrown away the benefit. Closing the loop is the whole return.

Here is a workflow that fits a small team.

1. Define the trigger. A detractor is NPS 0 to 6, or CSAT 1 to 2. Decide this once and automate it. Nobody should be reading a spreadsheet to find unhappy people.
2. Name the owner before the first send. One person, by name, not "the support team". Unowned feedback is a queue that fills up.
3. Reply within 24 hours. Not to fix it. To acknowledge it. The gap between a complaint and any human response is what turns a bad experience into a story people tell.
4. Reply on the channel they answered on. If they rated you from a WhatsApp thread, answer in the WhatsApp thread. Starting a new email chain restarts the relationship from zero.
5. Write the outcome on the contact record. Next time this person talks to you, whoever picks it up should see that they scored you a 3 in July and what happened next.
6. Review weekly, for thirty minutes. Read every open text answer from the week out loud. Count the reasons. Fix the one that appeared most. That meeting is worth more than the dashboard.

What to actually say to a detractor

Keep it short, do not defend anything, and ask one question. Something like: "Thanks for the honest score. I would rather understand it than average it. What went wrong?" Then stop typing and wait.

Three things not to do. Do not open with an apology for a problem you have not identified yet, because it reads as a script. Do not offer a discount in the first message, because it converts a quality conversation into a price negotiation. And do not ask them to change their score, ever. The moment a customer learns that the score has consequences for you, the score stops measuring anything.

Promoters are a workflow too

Nearly every article stops at detractors. A 9 or a 10 from somebody with an open deal is a referral request, a review request and a case study in one record, and the window for asking is roughly the week after they gave you the score. The reason this gets skipped is that promoters do not create urgency. Put it on the same weekly review and it stops getting skipped.

The connection between feedback and revenue is not the score, it is this follow-up. Detractor conversations are early warning for churn, and how fast you answer them is the same operational muscle as first response time. If your feedback and your conversations live in two different products, both of those get slower.

Running surveys from your inbox with Pinlyx

This is the part we build, so here is exactly what it does and what it does not.

Surveys in Pinlyx run from the same workspace as your contacts and conversations, which means a response lands on the contact record rather than in a separate tool, and the follow-up happens in the thread that is already open. Surveys are available in Pinlyx; see the pricing page for which plan you need.

1. Surveys, then new survey. Pick the type: NPS on a 0 to 10 scale, or CSAT on a 1 to 5 scale.
2. Write the three pieces of text. The question, the follow-up question and the thank-you message. These are the three fields from every template above.
3. Pick the language. English, Turkish, Spanish or Russian, so the answer page renders in the language your customer reads.
4. Turn on the detractor task. With the option enabled, an NPS score of 6 or below, or a CSAT score of 2 or below, opens a high-priority task in the CRM automatically. That is step one of the closing-the-loop workflow above, done without anybody watching a dashboard.
5. Set it active, then Send survey. Choose the channel: email, Telegram, WhatsApp or a shareable link.
6. Choose the people. Contacts are picked from a search box, up to 500 per batch. Send.
7. Read the results. The results screen shows the score, the distribution across the bands, the trend over time, and the list of responses with a channel column and a detractor badge on the ones that need a reply.

Answers are collected on a public answer page, so the person answering does not need an account or a login. Because the response is tied to the contact, the score sits next to the conversation history in the unified inbox and next to any tickets in the support desk, which is what makes a same-thread reply possible.

Now the honest limits, because you will hit them in week one:

• There is no tag or pipeline-stage audience picker. You choose recipients from a search box, one batch at a time, capped at 500 contacts. For a scheduled quarterly NPS across a large base, that is manual work.
• There is no Instagram survey channel. Email, Telegram, WhatsApp and link only.
• WhatsApp sending needs a connected WhatsApp account, which is on the paid plans, and it obeys the 24 hour customer service window described earlier. If the customer has not messaged you in the last day, use email or a link instead.
• The support desk takes tickets from live chat, the support mailbox and manual entry. Conversations on other channels do not create tickets by themselves.

If you want to try the flow end to end, create a free account, connect a channel, and send a CSAT survey to five contacts you spoke to this week. Five answers will teach you more about your question wording than a month of planning.

If the follow-up volume is more than one person can carry, an AI sales agent can draft the first reply and leave the sending to you, which keeps the loop closed without pretending a machine apologised.

When a survey is the wrong tool

Three cases where you should not run one, and one where you should buy a different product.

You have fewer than about 50 relevant customers per period. At realistic response rates that is a handful of answers, and a handful of answers cannot support a percentage. Call five of them instead. A thirty minute conversation with a real customer beats a score with an interval so wide it covers every possible reality.

You already know what is wrong. If your delivery is late and everybody has told you so, a survey confirming that delivery is late is a way of being busy. Fix it, then measure whether the fix worked.

You are not going to reply to anyone. A survey is a promise that somebody is listening. If nobody is, you are teaching customers that telling you things does not work, which is worse than never asking.

And if you need a real survey platform, buy one. Complex branching logic, randomized question order, panel sampling, academic-grade weighting, multi-language questionnaires with dozens of questions: that is what SurveyMonkey and Qualtrics are for, and we are not pretending to compete with them. What we do is narrower on purpose. One question, sent into a conversation, attached to a customer record, with a task waiting on the other side. If your problem is measurement at scale, use a survey platform. If your problem is that feedback arrives detached from the person who gave it, that is the one we built for.

Ten mistakes that kill response rates

1. Asking eight questions. Each extra question up to about 15 adds abandonment, per the SurveyMonkey study above. One question plus one optional follow-up is not a compromise, it is the design.
2. Sending from a no-reply address. You asked for a conversation from an address that cannot receive one. People notice.
3. Surveying too late. A week after the event you are measuring a mood, not an experience.
4. Surveying the same people over and over. Set a rule, for example nobody gets surveyed twice in 90 days, and enforce it across teams. Marketing and support both surveying the same customer in the same week is common and invisible.
5. Sending only to happy customers. Surveying after successful cases only, or only people who logged in recently, guarantees a flattering number and hides the people about to leave.
6. Leading questions. Pew's experiments show wording alone can move an answer by 25 points. Yours will move too.
7. Making the open box mandatory. You will get "n/a" and lose the people who would have given you a number.
8. No thank-you screen. The last thing a respondent sees decides whether they answer the next one.
9. Not replying to bad scores. This is the expensive one. A detractor who hears nothing has learned something about you that no survey will ever measure.
10. Reporting the score without the response rate. A CSAT of 92% on a 4% response rate is a statement about four percent of your customers, and probably the friendliest four percent.

There is a broader expectation shift behind several of these. Zendesk's CX Trends 2026 reports that 88% of customers expect faster response times than they did a year ago, 74% now expect service to be available 24/7, and 63% say their demand for transparency has risen compared to a year earlier (Zendesk, checked September 2026). It is a vendor's own survey of its market, so weigh it accordingly, but the direction is not controversial: asking for feedback and then going quiet reads worse every year.

Privacy and consent, in plain terms

This section is informational and it is not legal advice. Get a lawyer for the specifics of your business.

Survey answers are personal data as soon as they are attached to a person, which in this whole post they are. That puts them under the same rules as the rest of your customer data.

Under the GDPR and its UK equivalent, you need a lawful basis before you process personal data, and the regulator is direct about the sequence: "You must determine your lawful basis before you start using the personal information and you must document it", and you must include your lawful basis in your privacy information (ICO, checked September 2026). The ICO lists seven bases, and for surveying your own existing customers, legitimate interests or contract are the ones usually in play rather than consent. Consent is what you need for marketing that rides along with the survey, which is an argument for not attaching marketing to your survey.

In Turkey the equivalent framework is the Law No. 6698 on the Protection of Personal Data, administered by the Personal Data Protection Authority (checked September 2026), and the same instincts apply: know why you are processing, tell people, and keep it to what you said.

Whatever jurisdiction you are in, four practices cover most of the risk and all of the decency:

• Say what happens to the answer. One line under the question: who reads it, whether it is linked to their account, whether anyone will reply.
• Be honest about anonymity. If the response is attached to a contact record, it is not anonymous, and calling it anonymous is the fastest way to lose trust. Identified feedback is more useful precisely because you can act on it. Say that instead.
• Do not use a satisfaction survey as a marketing list. Answering a CSAT question is not opting in to a newsletter.
• Keep answers as long as you need them and not longer. A retention period you can state is better than an archive nobody has looked at since 2021.

One more, which is ethics rather than law: if somebody tells you something in a survey and you quote it publicly, ask first. Testimonials harvested from feedback forms without permission are a reliable way to make your best customers stop answering.

To try it: open a free account on Pinlyx, connect Telegram and import your contacts, then run one CSAT survey on last week's conversations and reply to whoever scores you lowest. The free plan is the Telegram channel plus the CRM core and needs no card; surveys are available in Pinlyx and the pricing page says which plan they need. Create a free account, no credit card required.

Frequently asked questions

What is a customer feedback survey?

A short set of questions sent to an existing customer after a specific event, such as a purchase, a support conversation or a cancellation, to find out how that event went and why. Most use one scale question plus one open follow-up. The useful ones are attached to a named customer, so the answer becomes a follow-up rather than a statistic.

What questions should I ask in a customer feedback survey?

One scale question about the specific moment, then one open question asking why. For a purchase, ask how satisfied they are with the order. For support, ask how satisfied they are with the help they received. For the relationship, ask how likely they are to recommend you. The 40 question bank above is grouped by moment so you can pick the pair that fits.

How many questions should a customer satisfaction survey have?

One or two. SurveyMonkey's analysis of around 100,000 surveys found the steepest rise in abandonment comes with each question added up to about 15, and that completion rates fall by 5% to 20% once a survey passes seven or eight minutes (SurveyMonkey, checked September 2026). One scale question plus one optional open box is the structure that survives contact with a phone.

What is a good CSAT score?

There is no credible universal number, because the answer depends on the industry, the country, the scale you used and which points you counted as satisfied. Two dated reference points you can check: the UK Customer Satisfaction Index sat at 78.3 out of 100 in July 2026 across 277 organizations (UKCSI, checked September 2026), and Retently's 2026 benchmarks put NPS across industries between 26 and 68 (Retently, checked September 2026). Judge yourself against your own trend on the same question and population.

What is the difference between CSAT and NPS?

CSAT asks about one interaction and is answered right after it, so it is a fast operational signal. NPS asks about the whole relationship on a 0 to 10 scale and is answered whenever you send it, so it is a slower strategic one. They are not interchangeable: a company with high CSAT and low NPS usually has good support wrapped around a product problem. CSAT is a real percentage; NPS is the difference between two percentages and runs from -100 to +100.

When should I send a customer feedback survey?

As close to the event as you can. Within 24 to 48 hours of delivery for a purchase, immediately after a support ticket closes, at day 30 for a new subscription, and within a day of a cancellation. For scheduled sends, SurveyMonkey found customer surveys sent on Mondays got 10% more responses than average and Friday sends got 13% fewer (SurveyMonkey, checked September 2026), but never delay an event-triggered survey to catch a better weekday.

Can I send a survey on WhatsApp?

Yes, to people you were recently talking to. Meta's rules open a 24 hour customer service window when a customer messages you, and once it closes you can only send pre-approved templates (Meta, checked September 2026). So a WhatsApp survey is a same-day follow-up to a conversation, not a mass send. In Pinlyx, WhatsApp sending requires a connected WhatsApp account on a paid plan and respects that window; for anything outside it, use email or a shareable link.

How do I increase survey response rates?

In order of effect: ask fewer questions, send sooner, and send on the channel the conversation already happened on. Then make the first question tappable on a phone, write a subject line that says how long it will take, never make the open box mandatory, and reply to the people who answer. The last one compounds, because a customer who got a reply last time answers again.

Should survey responses be anonymous?

Only if you genuinely cannot act on them otherwise. Anonymity slightly increases candor and removes almost all of the value, because you cannot call the person who told you something was broken. Identified feedback with a clear statement about who reads it and what happens next is the better trade for most small businesses. What you must not do is call a response anonymous when it is attached to a contact record.

Related reading

• Live chat conversion benchmarks, including what chat CSAT figures actually measure.
• Omnichannel messaging benchmarks, on where customer conversations are actually happening.
• Telegram versus WhatsApp for business, if you are deciding which channel to survey on.
• How to choose a CRM in 2026, for the data model questions behind attaching a score to a person.
• Customer support with Pinlyx, for how feedback fits alongside tickets and reply times.