Good product survey questions change with the product's age. Before launch you are testing whether anyone wants the thing; after launch you are measuring satisfaction, usage, and friction. This page carries the full bank, grouped by stage, with the wording rules that keep answers honest and the questions worth skipping.
By the Wavefield Research team · Published Nov 3, 2026
Lists of new product survey questions tend to open with demographics, which is the least useful place to start. Pre-launch, one question has outearned the rest: how would you feel if you could no longer use the product? Sean Ellis proposed tracking the share who answer very disappointed, put the benchmark at 40 percent from his own startup benchmarking, and Superhuman's product-market fit engine turned it into a public, repeatable improvement loop. The question works because it asks about loss instead of enthusiasm, and people are more honest about what they would miss than about what they admire.
The other pre-launch questions triangulate the same thing: what would you use instead, and would you buy it at a named price. Treat stated intent gently. Identifying buyers before a product exists is its own craft, covered in the potential-customer questionnaire section of our market research surveys guide, and testing concepts head to head is what concept testing exists for.
Twelve questions cover nearly every product survey worth fielding. No single survey should run all twelve: pick the stage, take its group, add one open-end, and stop. Each note says what the question buys you and the mistake it usually ships with.
“How would you feel if you could no longer use [product]? (Very disappointed / Somewhat disappointed / Not disappointed)”
The product-market fit question. The share answering very disappointed is the number to track; the benchmark story behind it is in the section above. Ask it of people who have actually used the product, never of people reacting to a description.
“What would you use instead if [product] did not exist?”
The alternatives question, and the cheapest competitive research you will ever run. A respondent who names a spreadsheet is telling you something different from one who names your funded rival.
“How likely would you be to buy [product] at [price]? (Definitely would not / Probably would not / Might / Probably would / Definitely would)”
Purchase intent on a labeled five-point scale. Count only the top box or top two boxes, then discount: stated intent runs well ahead of behavior, because saying yes in a survey costs nothing.
“Overall, how satisfied are you with [product]? (Very dissatisfied to Very satisfied)”
Plain CSAT. Its job is the trend line, so lock the wording and the scale at wave one and resist every clever edit afterward.
“How likely are you to recommend [product] to a friend or colleague? (0 to 10)”
The NPS core. Keep the standard 0 to 10 scale and the recommend framing so your number stays comparable, and always pair it with an open-ended why.
“How often do you use [product]? (Daily / Weekly / Monthly / Less often)”
Frequency is your best free segmentation variable. Every other answer in the survey reads differently once you split it by daily users versus lapsed ones.
“What were you trying to get done when you first looked for something like [product]?”
The job-to-be-done question. Answers here tell you which market you are actually in, which is routinely a surprise to the roadmap.
“Which part of [product] would you miss least if we removed it?”
The inverse value question. People struggle to rank what they love and are strikingly fluent about what they would not miss. Roadmaps need both lists.
“How easy was it to [complete the core task]? (Very difficult to Very easy)”
Anchor it to one named task, asked right after the task happens. A general ease rating collected months later measures memory, and memory is generous.
“Where did you most recently get stuck or confused in [product]?”
Recency keeps this answerable. You are collecting bug-shaped anecdotes on purpose; three people naming the same screen outranks any opinion question.
“If you could change one thing about [product], what would it be?”
The single best open-end for roadmap input, because 'one thing' forces a priority instead of a wish list. Code the answers into themes before anyone quotes their favorite.
“What almost stopped you from buying or signing up?”
Asked of recent converts, this is objection research from people who overcame the objection, which makes them far easier to reach than the people who bounced.
The recommend question carries a whole discipline of its own, follow-ups and detractor emails included, on our NPS survey questions page, and the open-ends deserve the craft covered in open-ended survey questions.
Question wording is the cheapest quality lever in survey research and the most ignored. Pew Research Center's questionnaire guidance puts the first rule plainly: ask one question at a time, because double-barreled questions that bundle two concepts are hard for respondents to answer and impossible for you to interpret. The rest of the craft for product surveys fits in four cards.
How satisfied are you with the price and the quality is two questions wearing one question mark. The customer who loves the quality and resents the price has no honest answer available.
What did you do is more reliable testimony than what would you do. Lead with usage and recency questions, and let the prediction questions carry less of the decision.
A 1 to 5 scale with only the ends named makes respondents invent the middle. Very dissatisfied through very satisfied, all five written out, costs nothing and removes a whole class of noise.
Usage frequency from a daily-active user is a question wasted on respondent patience. Pull what you can from the product, and spend the questionnaire on the things only a human can report.
One sampling rule belongs beside the wording rules, because it breaks more product surveys than any phrasing does: a survey of your mailing list is a survey of people who liked you enough to stay on it. For decisions that need strangers, field to a matched panel and size the study first with the sample size calculator.
Good product survey questions match the product's stage. Before launch: the product-market fit question, the alternatives question, and labeled purchase intent. After launch: satisfaction, recommend likelihood, usage frequency, task-level ease, and one forced-priority open-end. Each question should have a named decision waiting on its answer; the pages of product survey questions and answers you find in search results mostly skip that test.
A five-question product survey that earns its length: how satisfied are you overall, how likely are you to recommend it (0 to 10), how often do you use it, what were you trying to get done when you found it, and if you could change one thing what would it be. That set covers the trend metrics, segmentation, the market you are really in, and one open roadmap input.
Five to ten for a standing feedback survey, up to fifteen for a one-off study with real sampling behind it, and one to three for anything triggered inside the product. Completion rate falls with every added question, and the people who finish a twenty-minute product survey are your superfans, which quietly biases the very scores you wanted.
How would you feel if you could no longer use this product, answered as very disappointed, somewhat disappointed, or not disappointed. Sean Ellis proposed tracking the percentage answering very disappointed, with 40 percent as the benchmark drawn from his startup benchmarking, and Superhuman later built a public improvement loop around the same question. It only works on people with real usage behind them.
Customer feedback survey questions follow the same rules as the bank above: one concept per question, labeled scale points rather than bare numbers, behavior questions before opinion questions, and an open-ended why next to any score you plan to act on. The commonest failure is asking about everything the team shipped instead of the few things a customer can actually judge.
The same bank works, scaled down: satisfaction, recommend likelihood, usage frequency, one ease question, and one open-end make a defensible class survey. The part graders actually reward is the method: say who you sampled, how many responded, and what the margin of error allows you to claim. A small honest sample beats a big convenience one.
Related: customer feedback surveys · survey question types · NPS survey questions · concept testing software
Brief the agent with your product and stage, and it programs the survey: questions, scales, logic, and the coded open-ends. $99 per study.
Build a product survey free