The short answer: treat "5 minutes to complete, around 15 questions" as your upper limit, then cut every question that doesn't directly serve the goal. The reasoning is simple — the more questions you add, the more both your response rate and your response quality drop. "While we're at it, let's ask everything" is the single biggest way to ruin your data.
That said, "15 questions" is only a starting point. The right number shifts with your survey's purpose, how motivated your respondents are, and the channel you distribute through. This guide walks through 5 steps to set the question count for your own survey, each with the "this is where many people fail" pitfall to watch for.
Why more questions means worse results
The link between question count and response rate/quality has been demonstrated repeatedly across studies. In the web-survey experiment by Galesic & Bosnjak (2009), the longer the survey appeared, the lower the start and completion rates — and the later a question sat, the sloppier the answers became (response times shrank and people stopped reading the options carefully).
Two problems show up at once.
- A quantity problem (drop-off): long surveys don't get finished. Mid-survey abandonment rises and your sample shrinks
- A quality problem (cutting corners): the further in, the more straightlining you get — "just pick the middle," "same option all the way down"
In other words, the more questions you add, the less and the rougher the data you collect. That's the mechanism behind why "ask everything" backfires. So how many questions should you have? Decide it in five steps.
Step 1: Work backward from the decision the survey informs
Before you settle on a question count, write one sentence: "What will we decide based on the results of this survey?" Keep only the questions that decision actually needs.
- ❌ "We want to understand our customers broadly" → questions multiply without limit
- ⭕ "We need to decide whether the next feature ships as Option A or Option B" → the necessary questions narrow down fast
This is where many people fail: when the goal stops at "we want to know," the question count grows without bound — every "this would be interesting to have" question makes it in. Ask of each question, one at a time, "Would the answer to this change what we do?" and cut the ones that wouldn't. This alone cuts most surveys in half.
Step 2: Sort questions into must-have, nice-to-have, and unnecessary
Take your list of candidate questions and split it into three buckets.
- Must-have: questions that feed directly into the Step 1 decision
- Nice-to-have: questions that deepen the analysis but aren't required to make the decision
- Unnecessary: "just in case" or "we might use this someday" questions → cut them
Build the survey from the must-haves first, and only add a few nice-to-haves if you have time to spare. Keep that order.
This is where many people fail: reflexively packing in a pile of demographic questions (age, gender, occupation, and so on). Any attribute you won't cross-tabulate just complicates the analysis. Narrow it down to only the dimensions you'll actually cross-tabulate. Too many demographics is a classic cause of drop-off.
Step 3: Cap the length by completion time (the 5-minute, 15-question rule of thumb)
It's more accurate to think in completion time than in raw question count, because question types vary widely in how long they take.
- Single-answer (SA): about 10–15 seconds per question
- Open-ended: about 30–60 seconds per question (heavy burden)
- Matrix question: one grid carries the burden of several questions
As a rule of thumb, design to an upper limit of 5 minutes to complete — roughly 15 questions in simple-question terms. If you go over, head back to Step 2 to cut, or split the survey into two.
This is where many people fail: counting a matrix question as "one question." A 5-item by 7-point matrix feels like five questions' worth of effort. Leaning heavily on matrices invites cognitive load and straightlining. For more, see The design and pitfalls of matrix questions.
Step 4: Adjust for distribution channel and device
The same 15 questions can call for a different length depending on who answers and where.
- Mostly mobile respondents → go even shorter. Scrolling and small screens cause people to drop off sooner
- Panel surveys (incentivized, used to answering) → can run a bit longer and still get completed
- Requests to your own customers (high motivation) → moderate length is tolerated
- Cold lists, first contact → keep it as short as possible
Deutskens et al. (2004) showed experimentally that respondent motivation and incentive design affect both response rate and quality. Setting one uniform length while ignoring "who you're asking" is risky.
This is where many people fail: designing on a desktop, judging it fine, and never checking how it looks and feels on a phone. Most survey responses today come from mobile. Tune the length for mobile-first. For more, see The mobile survey design guide.
Step 5: Shrink the perceived length with intros, progress, and branching
When you genuinely can't cut more questions, you can still reduce the feeling that a survey is long.
- State the time required in the intro: telling people "about 3 minutes, 10 questions" up front lets them brace for it, which reduces drop-off
- Show a progress bar: seeing "almost there" lifts completion rates
- Serve questions conditionally with branching: instead of showing everyone every question, show relevant questions only to the people they apply to. This erases the felt burden of "questions that have nothing to do with me"
This is where many people fail: showing every respondent every question. With branching, you can dramatically reduce the perceived question count per person. For writing intros, see The survey intro and closing message guide; for designing branches, see The complete guide to survey branching logic.
From the editors — the 3 things that really move the needle on question count
From continuously tracking industry cases and the voices of practitioners, here are the three things about question count that always pay off.
1. Every "just in case" question you cut buys you more completed responses
Deciding to cut one question is almost the same decision as adding one completed response. Questions added "just in case" or "we might use this someday" almost always go unused while doing nothing but driving drop-off. The courage to cut ends up growing your data volume.
2. If you must go long, put the important questions first
As the research shows, response quality falls toward the back of a survey. If length is unavoidable, place the questions most critical to the decision in the first half. Even if people drop off, you've secured the key data. Just be aware that question order can also bias answers — read Order effects and question sequencing in survey design alongside this.
3. "Short with a high response rate" is usually worth more
"Ask a lot and get a 20% response rate" is worth less than "stay focused and get a 60% response rate" — the latter has less sample bias and yields data you can actually decide on. The information you gain by adding questions tends to be canceled out by the bias from a lower response rate (you only have data from the people who answered). Design with the trade-off between length and response rate firmly in mind. For lifting the response rate itself, see 10 practical techniques to improve your survey response rate.
Summary — 5 steps to set your question count
- Work backward from the decision — one sentence on "what the results will decide." Cut questions that won't change what you do
- Sort into must-have, nice-to-have, unnecessary — "just in case" is unnecessary; keep only the demographics you'll cross-tabulate
- Cap by completion time — aim for around 5 minutes and 15 questions. Over the limit? Cut or split
- Adjust for channel and device — go even shorter for mobile-first; tolerance shifts with motivation
- Shrink the perceived length — state the time required, show a progress bar, and use branching to ease the "this is long" feeling
The right number of questions is "the fewest that still achieve your goal." Asking more doesn't get you better data — usually the opposite. The discipline to cut is what protects both your completion count and your response quality.
If you want to build a focused, easy-to-answer survey, try the free survey tool Kicue. Just upload your survey document and AI auto-generates the questions — then, in a single account, you can set up branching logic to "show questions only to the people they apply to" and build a form with a visible progress bar. Start with a short survey of nothing but the questions that serve your goal.
References (2)
- Galesic, M., & Bosnjak, M. (2009). Effects of Questionnaire Length on Participation and Indicators of Response Quality in a Web Survey. Public Opinion Quarterly, 73(2), 349-360.
- Deutskens, E., de Ruyter, K., Wetzels, M., & Oosterveld, P. (2004). Response Rate and Response Quality of Internet-Based Surveys: An Experimental Study. Marketing Letters, 15(1), 21-36.