Research methods
Product research survey: what to ask, how to write it and who to send it to
How to design a product research survey: when a survey fits, example questions by goal, question-writing rules, a worked example, costs and the interviews to pair it with.
A product research survey asks the people you want to serve about their needs, the way they solve the problem today, which features matter most and what they would pay. It is useful for one kind of question: how many people think or feel a certain way. It is poor at explaining why. So the best product surveys come after a few conversations, which tell you what to ask, and before a few more, which explain the answers.
This guide is about surveys that inform what to build. If you already have users and want to know whether the product has found its market, see the product-market fit survey instead.
Check that a survey fits your question
The Nielsen Norman Group describes surveys as the one common research method that is both quantitative and attitudinal: good for measuring how many people hold a view, not for observing what they do. Its two examples make the line clear. "What percentage of users feel confident their tax return is correct?" suits a survey. "Why are people abandoning their carts?" is a question about behavior and reasons, better answered by watching people use the product.
Before writing questions, write the sentence you want to be able to say afterwards, with a number in it. For example: "Most shop owners who track inventory in spreadsheets rank supplier reordering above sales reports." If you cannot write that sentence, start with interviews.
Questions by research goal
| What you want to learn | Example question | Watch out for |
|---|---|---|
| How people handle the problem today | "How do you currently reorder stock from suppliers?" (list built from interviews, plus "Other") | Answer lists invented without talking to anyone first |
| How often the problem occurs | "In the last month, how many times did you run out of a best-selling item?" | Asking about a vague period, or asking people to predict the future |
| Which features matter most | Show a few features at a time and ask which matters most and least (a MaxDiff exercise) | Rating every feature "very important" on a 1-5 scale |
| Reaction to a concept | Short description, then "How well does this fit how you work?" | Describing the concept in glowing terms |
| Price | Questions about what they pay today and at what price the idea would seem too expensive | Asking "Would you pay $X?" and trusting the yes |
| Who they are | Role, business size, current tools | Putting these first; Pew advises against asking demographics near the start unless needed for screening |
Price questions deserve extra care: saying yes to a price in a survey costs the respondent nothing. Researching willingness to pay covers better approaches.
Write questions people can answer accurately
Pew Research Center's questionnaire guidance contains most of the rules that matter:
- Ask one thing at a time. A question about two things at once ("How satisfied are you with the speed and accuracy of the app?") gives answers you cannot interpret.
- Keep answer lists short and complete. Pew suggests four or perhaps five choices for most questions, with options that do not overlap and that cover every reasonable answer.
- Build lists from real answers. Pew notes that researchers sometimes run an open-ended pilot first, then turn the most common answers into the options for the closed-ended version.
- Avoid agree-disagree statements. People tend to agree with statements, a pattern Pew calls acquiescence bias. Offer a choice between alternatives instead.
- Randomize lists without a natural order. People completing surveys themselves tend to pick options near the top.
- Pretest. Pew calls pretesting an essential step, especially for new questions. Watch three or four people take the survey and ask what they thought each question meant.
A worked example
A hypothetical company is building inventory software for independent bike shops. It starts with six calls to shop owners and learns that reordering from suppliers causes more trouble than tracking stock. That changes the survey.
The survey goes to 150 shop owners. It asks how they reorder today (options taken from the calls), how many times in the past month they ran out of a top seller, a MaxDiff exercise across eight possible features, and what they pay for current tools. The result is a ranked feature list and a count of how common stock-outs are. The team then calls five owners whose answers surprised them, to understand the reasons.
Who to send it to, and what it costs
Send the survey to people who match your target customer, including people who are not your users. Existing customers are the easiest to reach but will not tell you why others chose something else.
For consumer products, a panel is the quickest route. Pollfish lists responses from $0.95 each for a short survey, with add-ons in its calculator: $0.50 per response for one screening question, $1.00 for a MaxDiff or conjoint exercise and $0.50 for a price-sensitivity module. It says final pricing varies with length and targeting.
For business products with small, specialized audiences, panels often cannot reach enough of the right people. A survey sent to your own list plus a set of interviews may serve you better. Validating feature requests covers how to test demand for a specific feature.
Follow the numbers with conversations
A survey might tell you that most respondents rank one feature first. It does not tell you what they would stop using to adopt it, or who signs off on the purchase. Plan a few calls with respondents, or with people like them, before you commit to building anything.
Instant Expert finds people who match a description you write, such as "product managers at B2B software companies who ran a pricing change in the last year." You review the people it finds, it sends your invitations and you pay only for calls that get booked. The directory page for product professionals in enterprise SaaS shows the kind of people you can reach.