Every one of the types of survey questions you’ll ever use belongs to one of two families. Closed-ended questions, where people pick from answers you provide. Open-ended questions, where they write whatever they want. Inside those two families live twelve working formats: dichotomous, multiple choice, checkbox, dropdown, Likert scale, rating scale, slider, NPS, semantic differential, ranking, matrix, and open text. Closed questions give you clean, countable data. Open questions give you the reasons behind the numbers. A good survey uses mostly the first family and a careful pinch of the second.
So you know the lens here: surveys are what I stare at all day, since I built Uplup to make them. And after years of watching what people create with it, I can tell you the most common mistake isn’t a bad question. It’s a mismatched format: an open text box where a scale belonged, or a 10-option multiple choice where a ranking would’ve said more. Honestly, matching the format to the question is the fun part of building a survey; get that right and the answers practically chart themselves.
So, this guide walks through each format with examples you can copy, the demographic questions everyone gets nervous about, how many points your scales should have, and the writing rules that keep your data honest.
Closed-ended types of survey questions
Closed-ended questions are the workhorses. Every answer lands in a bucket you set up in advance, so you can count it, chart it, and compare it instantly. They’re also fast to answer, and speed matters way more than most people think; I’ll show you the math further down.
1. Dichotomous questions
A dichotomous question has exactly two possible answers: yes/no, true/false, agree/disagree.
- “Have you purchased from us before?” Yes / No
- “Do you work remotely?” Yes / No
These are your screening and routing questions. They’re the natural branch points in a survey: answer yes, get the follow-up questions; answer no, skip ahead. But two answers is a blunt tool for feelings. “Was the event good?” as a yes/no throws away every shade of opinion, so save this format for facts, not feelings.
2. Multiple choice (single answer)
One question, several options, one selection allowed.
- “How did you hear about us?” Search engine / Social media / A friend / Podcast / Other
- “Which best describes your role?” Founder / Marketing / Engineering / Operations / Other
Use it for categorical facts where the options are known and mutually exclusive. Two craft rules: cover every realistic case (that’s what “Other” is for), and never let two options overlap. If “1-5 employees” and “5-10 employees” both exist, the 5-person company doesn’t know where to click, and you don’t know what their click meant.
3. Checkbox questions (multiple answers allowed)
The select-all-that-apply format.
- “Which of these tools do you currently use? Select all that apply.”
- “What topics would you like us to cover?”
Reach for it when answers aren’t mutually exclusive. One honest warning, though: checkbox data is messier to analyze than single-select. Percentages sum past 100, and “selected” doesn’t mean “cares deeply.” When you need intensity rather than membership, a ranking or scale question tells you more.
4. Dropdown questions
Multiple choice folded into a compact, often searchable menu.
- “Which country are you in?” (a searchable list of countries)
- “What year did you graduate?”
Dropdowns earn their spot when the option list is long: countries, states, industries, years. For short lists, keep the options visible instead. Hiding four choices behind a click adds friction without saving meaningful space, and people answer more accurately when they can see every choice at once.
5. Likert scale questions
The Likert scale measures agreement or frequency along a symmetric 5-point (sometimes 7-point) scale.
- “The onboarding process was easy to follow.” Strongly disagree / Disagree / Neutral / Agree / Strongly agree
This is the one for attitudes and opinions, and it’s the standard for a reason: respondents understand it instantly, and the results chart beautifully. Keep the scale balanced (equal options on each side of neutral) and label every point, not just the ends.
6. Rating scale questions
A numeric or star scale: rate this from 1 to 5, or 1 to 10.
- “How would you rate your support experience?” 1 to 5 stars
Perfect for quick satisfaction ratings where a full Likert statement would be overkill. Just define the endpoints (“1 = poor, 5 = excellent”), because an unanchored scale means different things to different people.
7. Slider questions
A rating scale with a draggable handle instead of fixed buttons, often 0 to 100.
- “How likely are you to attend next year?” (slide from Not at all to Definitely)
Sliders shine when you want fine-grained intensity, or just a more tactile moment in a long survey. The honest trade-off: they feel more precise than they are. The difference between one person’s 71 and another’s 74 is noise, not signal, so analyze slider data in bands. Also, the starting position biases answers; a handle that starts at 50 pulls responses toward 50.
8. NPS questions
Net Promoter Score is a specific, standardized rating question: “How likely are you to recommend us to a friend or colleague?” on a 0 to 10 scale. Scores of 9 to 10 are promoters, 7 to 8 are passives, 0 to 6 are detractors, and NPS = % promoters minus % detractors.
The whole point of NPS is benchmarking: loyalty over time, and against other companies, which only works because everyone asks it exactly the same way. So ask the standard question, then follow with one open-ended “What’s the main reason for your score?” That follow-up is routinely the most valuable text data in the whole survey.
9. Semantic differential questions
A scale strung between two opposite adjectives, with no labels in between.
- “How did the demo feel?” Complicated 1 2 3 4 5 Simple
- “Our pricing is:” Unfair 1 2 3 4 5 Fair
It’s built for brand and perception research, where you want gut reactions on specific dimensions rather than agreement with statements. Interestingly, it reads faster than Likert because there’s no statement to parse, and three or four adjective pairs will sketch a product’s personality better than you’d expect. The craft is choosing genuine opposites; “cheap” is not the opposite of “premium” in every respondent’s head.
10. Ranking questions
Respondents order options by preference or importance.
- “Rank these features by how important they are to you.”
Use it when you need priorities, not just preferences. A checkbox tells you people want all five features (of course they do); a ranking forces the trade-off, and trade-offs are where the truth lives. Keep lists to 5 or 6 items; ranking 10 things is real work and your completion rate will pay for it.
11. Matrix questions
A matrix is a grid: several rows (statements) sharing one column scale.
- Rows: “Ease of use,” “Value for money,” “Support quality.” Columns: Very dissatisfied to Very satisfied.
It compresses related scale questions into one compact block, which is a real space-saver. But use it sparingly: matrices are where surveys go to die on mobile screens, and big grids invite straight-lining, where a fatigued respondent picks the same column all the way down. Two or three rows, one matrix per survey, is the safe zone.
Open-ended survey questions
Only one format lives in this family, but placed well, it punches way above its weight.
12. Open text questions
A box and a prompt.
- “What almost stopped you from purchasing today?”
- “If you could change one thing about the product, what would it be?”
This is where you get the why behind the numbers, and the answers you didn’t know to list. The catch is cost: open questions take 5 to 10 times longer to answer and infinitely longer to analyze, and each one you add pushes some respondents to quit. My rule: one or two per survey, placed after the closed questions have built momentum, and almost never required. A required essay box on question two is the fastest abandonment trigger I know.
But there’s a middle path, and I love this pattern more than any other trick in survey building: the follow-up pattern. Ask closed (“How satisfied are you?”), then conditionally ask open (“Sorry to hear that. What went wrong?”) only to the people whose answer makes the follow-up relevant. You get depth exactly where depth exists, and everyone else sails straight through.
How many points should a scale have?
Five, most of the time. Every scale question forces this choice, and it’s less arbitrary than it looks, so here’s how I’d pick:
- 5 points is the default for a reason: everyone can hold “strongly disagree to strongly agree” in their head, and five buckets chart cleanly. When in doubt, five.
- 7 points adds resolution for engaged, thoughtful audiences (employee research, academic work). The extra granularity is real, but only if respondents are paying enough attention to use it.
- 10 or 11 points (0 to 10) exists mostly for NPS compatibility and slider-style intensity. Fine for benchmarked questions; overkill for “rate the venue.”
- Even-numbered scales (4 or 6 points) delete the neutral midpoint and force a lean. Use them deliberately when “neutral” would be an easy exit on a question where you need direction, and keep the midpoint when neutrality is a legitimate answer.
The rule underneath all of this: pick one convention per survey and stick to it. A survey that hops between 5-point, 7-point, and 10-point scales makes people relearn the ruler on every question, and comparability across your own questions quietly dies.
Demographic survey questions (and how to ask them politely)
Demographic questions (age, gender, location, income, ethnicity, education) are what turn one big average into groups you can actually compare. They’re also the questions people are most sensitive about, so the craft really matters here:
- Ask in ranges, not exact values. “What is your age? Under 18 / 18-24 / 25-34 / 35-44 / 45-54 / 55-64 / 65+” feels categorically different from “How old are you?” Same for income.
- Always include “Prefer not to say.” On every demographic question, no exceptions. A skipped sensitive question is fine; an abandoned survey is not.
- Make them optional and put them last. By the end, respondents have invested effort and are more willing to share. Leading with demographics is like opening a conversation with “How much money do you make?”
- Only ask what you’ll segment by. If you’re not going to cut the data by education level, don’t collect education level.
Writing rules that keep your data honest
So, the format is half the job. The wording is the other half. Four rules cover most of the damage:
- One question per question. “Was the venue clean and well organized?” is a double-barreled question: someone who found it clean but chaotic has no honest answer. Split it.
- Don’t lead. “How much did you enjoy our award-winning support?” has the answer built in. Neutral version: “How would you rate your support experience?”
- Balance your scales. Options running Excellent / Great / Good / Fair force a positive answer. Symmetric scales with a true midpoint keep the measurement honest.
- Match the option list to reality. Overlapping ranges, missing categories, and no “Other” all force false answers, and false answers are worse than no answers because they look like data.
And the rule above all of them: every question costs you completions. Here’s the math I always run. Closed questions take roughly 10 to 15 seconds; open ones take a minute or more. A 10-question closed survey is about a 2-minute ask, and short surveys reliably see completion rates around 80% or better. Stretch the same survey to 25 questions with three required essays and you’re asking for 8+ minutes, and completion commonly drops below half. If 500 people start both versions, that’s the difference between 400 finished responses and perhaps 200. The shortest survey that answers your actual research question is the best survey; everything else is decoration you’re paying for with data.
A starter question bank you can copy
Formats are abstract until they’re questions, so here are ready-to-use examples grouped by the three most common survey jobs. Steal freely and reword to your context.
Customer satisfaction (post-purchase or post-support):
- “How would you rate your experience today?” (1-5 rating)
- “How easy was it to get your issue resolved?” (Likert: Very difficult to Very easy)
- “How likely are you to recommend us to a friend or colleague?” (NPS, 0-10)
- “What’s the main reason for your score?” (open text, optional)
- “Did our team resolve your issue on the first contact?” (Yes / No)
Product feedback:
- “How often do you use [product]?” (multiple choice: Daily / Weekly / Monthly / Rarely)
- “Rank these upcoming features by how useful they’d be to you.” (ranking, 5 items)
- “How disappointed would you be if you could no longer use [product]?” (multiple choice: Very / Somewhat / Not at all)
- “The product feels:” Slow 1-5 Fast, Confusing 1-5 Intuitive (semantic differential pairs)
- “If you could change one thing, what would it be?” (open text, optional)
Event feedback:
- “How would you rate the event overall?” (1-5 rating)
- “Rate the following:” Speakers / Venue / Schedule (matrix, one 5-point scale, three rows)
- “Which sessions did you attend?” (checkbox)
- “How likely are you to attend next year?” (slider or 0-10)
- “What should we do differently next time?” (open text, optional)
Notice the shape repeating across all three: mostly closed, one scale benchmark, one trade-off question, and exactly one optional open text at the end. That’s not me being tidy; that’s the completion-rate math wearing a template.
Building each format in Uplup
All twelve formats in this guide exist as ready-made field types in Uplup’s form builder: multiple choice, checkboxes, dropdown (searchable, with a full country list built in), Likert scale (5-point by default with editable labels), star rating, slider, NPS (0 to 10 with the standard detractor/passive/promoter coloring), ranking, matrix (you set the rows and columns), and short or long open text. And all of them are on the free plan!
The craft rules from this guide map straight to choices you make in the builder: any question can be required or optional (demographics: optional), “Other” options are one toggle on choice fields, and conditional logic handles the follow-up pattern, showing the open “what went wrong?” box only when the satisfaction score comes in low. Answers land in a results view with per-question charts, and you can export to CSV when you’d rather do the slicing yourself. Also, the survey templates gallery has prebuilt customer satisfaction, feedback, and research surveys if you’d rather start from a working example and edit.
Frequently asked questions
The ones I get asked most, with the answer up front every time.
What are the main types of survey questions?
Twelve formats across two families. Closed-ended: dichotomous (yes/no), multiple choice, checkbox, dropdown, Likert scale, rating scale, slider, NPS, semantic differential, ranking, and matrix. Open-ended: free text questions. The closed ones give you countable data; the open ones give you the explanations.
What is the difference between open-ended and closed-ended questions?
Closed-ended questions supply the answer options and the respondent selects; open-ended questions let the respondent write anything. Closed answers are fast to give and easy to chart. Open answers are slower, but they’re where the surprises live, and surprises are exactly what predefined options can’t catch.
What survey questions should I ask?
Start from the decision the survey serves, then work backwards: for each question, name the action you’d take differently based on its answer. If no action changes, cut the question; that one habit shrinks most drafts fast. Most decision-focused surveys land at 8 to 12 closed questions plus one open follow-up.
How many questions should a survey have?
Under 10 questions and roughly 2 to 3 minutes is the reliable zone for strong completion rates. Longer surveys can work when the audience is invested (employees, paying customers), but every added question, and especially every required open question, drops completion.
What is a good example of an unbiased survey question?
“How would you rate the checkout process?” with a balanced scale from Very difficult to Very easy. It names the topic without praising it, offers symmetric options, and lets a negative answer feel as legitimate as a positive one. The biased version of the same question would be “How easy was our new streamlined checkout?”
Should survey questions be required?
No, most of them shouldn’t be. Require the few questions your analysis can’t live without and leave the rest optional, especially demographics and open text. A forced answer to a question someone wanted to skip is frequently a junk answer, and junk that looks like data is worse than a blank.
Is a 5-point or 7-point scale better?
For most business surveys, 5-point: it’s universally understood and charts cleanly. Use 7-point when the audience is engaged enough to use the extra resolution (employee or academic research), and an even-numbered 4 or 6-point scale when you deliberately want to remove the neutral midpoint and force a lean.
Do this next
Run the one-question audit on the last survey you sent (or the draft still on your desk): for each question, write down the action its answer would change. Cut everything without an action, then check each survivor’s format against this list. That usually shrinks a survey by a third and sharpens everything that’s left. And if you’re starting fresh, build the survey in Uplup with the follow-up pattern wired in; logic included, it’s maybe a 15-minute build. Send it today, while the question you actually want answered is still fresh in your head.
