Voice of the customer questions: the short list that works

The most productive question in customer feedback is eight words long: what is the main reason for your score? Many surveys take the opposite approach: long questionnaires, several matrices and a response rate that declines over time. Good VoC survey design is mostly an exercise in deciding what can be removed. This article is the question bank for that exercise: what to ask, per situation, and what to strike.

Key takeaways:

• Two questions carry every VoC survey: a score chosen for the moment, and the open why. Add more only for a named, current decision.
• Length is measured in completion: across 21,863 surveys, one question completes at 85.7%, twenty questions at 68.7%.
• Five open questions earn a place, each with a distinct job: the reason, the instruction, the friction, the silence, the strength.
• Strike leading questions, double-barreled questions, jargon and matrix batteries. Each one measures the questionnaire instead of the customer.
• Never force an answer. Research links compulsory fields to more careless responses; an honest blank beats a resentful click.

Two questions carry every VoC survey

1. A score question, chosen to fit the moment.
2. The open why: "What is the main reason for your score?"

For many recurring surveys, this pair is enough. It is quick to answer and lets customers describe the issue in their own language. Modern text analytics can then group those answers without forcing the company's assumptions into a long list of closed questions.

The completion data backs the minimalism. Survicate's analysis of 21,863 surveys found one-question surveys completing at 85.7% while twenty-question surveys fall to 68.7%, with the steepest engagement losses in the 21-to-40-question range. Every question you add taxes the customers most willing to help you.

The score question, worded per situation

• For the relationship, use NPS: "How likely are you to recommend [company] to a friend or colleague?" (0 to 10). Ask it periodically, away from any single transaction. The exact wording variants and their trade-offs are covered in NPS survey questions.
• For a moment, use CSAT: "How satisfied were you with [your delivery / your visit / this conversation]?" Ask it within hours of the moment.
• For a process, use CES: "How easy was it to [return your item / submit your claim / change your plan]?" Ask it where effort is the risk.

The choice of metric matters more than small wording differences. Decide whether the touchpoint calls for a relationship, satisfaction or effort question before fine-tuning the sentence. The guide to transactional NPS explains the distinction in more detail.

Open questions that earn their place

A small set of open questions is useful, provided each has a clear job:

• "What is the main reason for your score?" The workhorse. If a survey keeps only one open question, keep this one.
• "What should we do differently next time?" Turns criticism into instruction; useful after service moments.
• "What almost stopped you?" The friction-finder, strong after a purchase or sign-up: it surfaces the hesitations of the customers who made it through, which are the barriers of the ones who did not.
• "Was there anything you wanted to raise with us and didn't?" The silence question. Customers who wanted to complain and never did are typically the unhappiest segment in the base; this question is how you find them while they are still customers.
• "What did we get exactly right?" Positive drivers are coaching and marketing material; asking only about problems teaches customers to only report problems.

Use one, at most two, per survey. The bank is a menu rather than a checklist.

Questions to strike

• Leading questions. "How much did you enjoy our new store?" instructs the answer. My own four-leaf clover piece calls this the self-fulfilling prophecy: phrase a question so that it almost guarantees a type of answer, and the survey measures the question.
• Double-barreled questions. "Were our staff friendly and knowledgeable?" cannot be answered by someone who found them friendly and clueless. Pew Research Center's questionnaire guidance is blunt about these: questions evaluating two concepts at once produce answers that cannot be interpreted.
• Internal-jargon questions. Customers do not know what your "omnichannel journey" is, and should not have to.
• Matrix batteries. Ten aspects on five-point scales create work for customers and often produce superficial answers. Use them only when each aspect feeds a real decision.

Two further rules from the survey-methods world deserve adoption in VoC:

• Keep wording stable over time. Pew's standard for trend research applies directly to your NPS: change the wording and you have started a new trend line, whatever the dashboard says.
• Never force an answer. A 2025 study in Psychological Reports found that forced answering increases careless responding, and careless responses measurably distort results. A skipped question is information; a coerced answer is noise wearing information's clothes. Make the open field optional, always.

The question set per lifecycle stage

The same short-survey discipline applies across the customer lifecycle; only the score type and the one optional extra change:

• Early relationship (first order, onboarding): CSAT on the moment, the open why, and if anything extra, "What almost stopped you?", because fresh customers still remember the barriers.
• Steady state (recurring transactions): the transactional score fitting each touchpoint plus the why, nothing else, on a contact policy that respects how often the same customer appears.
• Renewal or repurchase window: relationship NPS with the why, asked away from any single transaction, plus the silence question ("anything you wanted to raise and didn't?") in the markets or segments where churn runs high.
• After cancellation: one question, openly framed: "What is the main reason you left?" Keep it in the confirmation flow, expect brutal honesty, and route these verbatims with priority, because exit answers are the cheapest churn research that exists.

The lifecycle view also exposes double-asking: a customer hitting renewal week and a delivery survey in the same fortnight should get the renewal question only. The relationship moment outranks the transaction.

Before you ask anything

The question is the last of four choices, and the easiest. As I wrote in the four-leaf clover of effective surveys, a touchpoint is defined by what you ask, whom you ask, at which moment and through which channel, and the thinking comes before the survey. The full methodology, from sampling to contact policy, lives in the voice of customer survey guide; the templates that hold it all together are in the voice of the customer template.

The test for every candidate question is the same: what will we do with the answer? Measuring is knowing, but knowing without doing is worthless.

Wording that lifts answer quality

Beyond choosing the right questions, five small wording habits raise the quality of what comes back:

• Name the moment. "How satisfied were you with yesterday's delivery?" beats "How satisfied are you with our service?" The specific memory produces the specific answer.
• Ask for the main reason, singular. "What is the main reason for your score?" pulls the priority; "any comments?" pulls whatever surfaced first.
• Use the customer's vocabulary. If customers say "the app", the question says "the app", not "our digital platform".
• Signal the effort honestly. "Two questions, thirty seconds" in the invitation lifts starts and, more importantly, finishes. Overpromising brevity and delivering a matrix battery is how response rates die.
• Write for a phone screen. Most feedback is typed with thumbs. A question that needs re-reading on mobile will be skipped or answered carelessly.

For multi-market brands, one more rule: translate the meaning, not the words, and keep the translation stable with the same discipline as the original. A CES question that reads naturally in English and stilted in Dutch is measuring translation quality in one market and effort in the other. Have a native speaker own each language version, and never A/B-test wording in one market while trending it in another.

A question audit you can run this week

Take your current surveys and score each question against four checks:

1. Use: can someone name the decision this answer feeds? If not, cut it.
2. Singularity: does it ask exactly one thing? Split or cut.
3. Neutrality: could a dissatisfied customer answer it as easily as a satisfied one? Rewrite or cut.
4. Language: would a customer use these words to a friend? Translate or cut.

Most surveys shrink by half under this audit, and the response rate usually thanks you within a quarter.

Add a fifth check for surveys that have run a while: stability. Has this question's wording, position or scale changed since launch? If yes, mark the trend line at the change date and stop comparing across it. The audit is also the right moment to bury questions added "temporarily" for decisions long since taken; every survey more than a year old contains at least one.

Frequently asked questions

How many questions should a voice of the customer survey have?

Use two as the default: a score and the reason for it. Add a third only when a specific, current decision requires it. Every extra question should have a named use.

What is the best follow-up question after NPS?

"What is the main reason for your score?" It works identically for promoters and detractors, invites the customer's own priorities, and feeds straight into topic and key driver analysis.

Should every survey include an open text question?

Usually, yes. Without an open field, you know the rating but have to guess what influenced it. Make the question optional and keep the survey short; forcing it demonstrably degrades the data.

Are closed questions ever the right choice?

For specific decisions, yes: a yes/no on a proposed opening hour, a multiple choice on preferred channel. Use them per decision, temporarily, and retire them once the decision is made.

Should VoC questions be personalised?

Where you can, yes: naming the store, the product or the moment lifts both response rate and precision. The metadata you attach matters as much as the wording, because it decides what the analysis can later compare.

Why the question set belongs on a leadership agenda

Survey wording looks like a detail to delegate, and mostly it is. Two aspects are not. The question set decides the total survey pressure on your customer base, which is a brand asset one enthusiastic department can quietly spend for everyone. And wording stability decides whether your NPS trend, the one in the board pack, is comparable across years at all. Both deserve an owner with the authority to say no, and a leadership habit of asking two questions a year: who approved the current question set, and when did the wording last change?

A practical default

Start with a score suited to the moment and one open question asking for the reason. Add nothing until a team can explain what it will do with the answer. That discipline protects the customer's time and makes the resulting feedback easier to act on. The best survey is the one your busiest customer finishes without noticing the effort.