Types of survey questions: seven formats and the job each one does best

The seven main types of survey questions, from yes/no to open-ended, with examples and one rule for choosing: pick the format by what you need to learn.

Instant Expert EditorialPublished Updated 6 min read

The main types of survey questions fall into four families: closed questions with fixed answers (yes/no, multiple choice, checkboxes), rating questions that measure degree (Likert, numeric and adjective scales), ranking questions that force a priority order, and open-ended questions answered in people's own words. Pick the format by what you need to know: a count or category, a degree of opinion, a priority, or something you cannot predict. Pick wrong and you get a tidy spreadsheet that answers a different question.

The types at a glance

TypeExampleUse it to learnWatch out for
Yes/no"In the past 30 days, did a tenant report a heating problem?"Whether something happenedPlain "Yes"/"No" invites agreement; spell out "Yes, this happened"
Multiple choice, one answer"How do tenants usually submit maintenance requests?"Which category someone falls intoOptions that overlap or miss common answers
Checkbox ("select all that apply")"Which of these tools do you use?"Several items at once, quicklyUndercounts compared with separate yes/no questions
Rating scale"How well does your current system work? Very well to Not at all well"Degree or intensityUnbalanced labels; everything rated high
Likert (agreement)"I trust our maintenance cost reports. Strongly disagree to Strongly agree"Attitudes across several statementsTendency to agree with statements
Ranking"Rank these three problems from most to least costly"Priority among optionsCannot show how far apart the ranks are
Open-ended"What almost stopped you from switching vendors?"Answers you cannot predictSkipped often and must be read and coded

Closed questions are for facts and recent behavior

Surveys lean on closed questions because, as the Nielsen Norman Group's article on open-ended vs. closed questions notes, they are much easier to answer and to analyze statistically.

Pew Research Center's guide to writing survey questions sets the rules: cover every reasonable answer, don't let options overlap, and keep most opinion questions to four or perhaps five choices (factual ones like industry can run longer). People filling in a survey alone tend to pick options near the top, so randomize lists with no natural order, and keep ordered scales like Excellent to Poor in order.

Checkboxes undercount, so ask yes or no for each item

"Select all that apply" is compact and less accurate than it looks. In a randomized Pew experiment with 4,581 U.S. adults, a separate yes/no question for each item produced endorsement rates that were, on average, 8 percentage points higher than a single checklist, and Pew concluded the forced-choice results were more accurate. People tick some of what applies and move on.

The rank order of items was nearly the same in both formats, so a checklist is fine when you only need the most common item. Use separate yes/no questions when each item's share matters to your decision, with construct-specific labels ("Yes, have done this" / "No, have not done this") to counter the tendency to say yes.

Rating scales measure how strongly people feel

GLG, an expert network that also runs surveys, lists several kinds of rating question: numeric (1-5 or 0-10), frequency ("Never" to "Always"), comparison ("Better" to "Worse") and Likert agreement scales. NN/g's survey best practices set the main rule: as many negative options as positive ones, plus a separate "Not applicable." The weakness of rating scales is that two items can tie at the top, and you can't tell which one people would pick. Likert questions have extra biases of their own, covered in Likert scale questions.

Ranking questions answer "which first"

GLG notes the limit of ranking: it can't show whether first and second were nearly tied, or whether fourth trailed far behind third. GLG suggests asking only for the top three (or bottom three), and asking both a rating and a ranking when you need intensity and priority together.

Keep ranking lists short. Ordering ten items on a phone is a small act of endurance, and GLG says that even among expert survey participants, 30% to 40% take surveys on their phones.

Open-ended questions catch what you didn't think to ask

Pew's election example shows why: 58% picked the economy from a list, but only 35% named it unprompted, and 43% of unprompted answers named something not on the list. The cost: Pew's analysis of its panel found open-ended questions skipped about 18% of the time on average, versus 1% to 2% for closed ones. Use one or two, where you can't predict the answers. Open-ended survey questions has examples and a simple coding method.

One survey, several types

Here is a hypothetical example. A founder building maintenance software for small property management firms needs to pick which problem to solve first. Their eight-question survey:

  1. Multiple choice: number of rental units managed, in non-overlapping ranges.
  2. Multiple choice: how tenants usually submit maintenance requests, with options from five earlier calls plus "Other."
  3. Yes/no series: "In the past 30 days, did any of these happen?" with "Yes, this happened / No, this did not happen" for each of four problems.
  4. Rating: "How well does your current process work for tracking repairs to completion?" Very well to Not at all well.
  5. Ranking: "Of the problems you said happened, rank the top two by how much time they cost you."
  6. Open-ended, optional: "What would you change first about how your firm handles maintenance?"
  7. Yes/no: "May we contact you for a 20-minute follow-up call?"
  8. Multiple choice: role, at the end, since Pew advises holding demographic questions until later unless they are needed for screening.

Question 3 shows how common each problem is, and question 5 shows which ones hurt most. Neither alone answers "which first."

Your next step

Next to each question in your draft, write what you need from it (a category, a count, a degree, a priority or an unknown) and check the format matches. Then read how to write survey questions and biased survey questions before you pretest.

When the survey leaves you with a "why," talk to people. Describe them, such as maintenance or procurement staff at small property management firms, and Instant Expert finds people who match; the directory page for procurement professionals in property management is one place to start.

The formats above have counterparts here. A call (15, 30, 45 or 60 minutes) is the open-ended question with follow-ups. A paid written or voice answer is a single open question the person answers on their own time. In both, the offer is what they receive, our fee is added on top (a $100 offer costs you $125), and you're charged only when they book or answer.