A survey is a favour you are asking. Every question spends a little of the goodwill the respondent brought with them, and most surveys spend it all before the useful part. The result is a pile of abandoned responses and a few polite lies.
Good survey questions are short, single-minded and easy to answer honestly. Get that right and people finish; get it wrong and even your fans bail halfway.
One idea per question
The classic mistake is the double-barrelled question: how satisfied are you with our speed and price. If someone loves the speed and hates the price, they cannot answer. Split it. Each question should ask exactly one thing.
- Avoid and or or inside a single question.
- Cut leading words like how great — ask, do not fish.
- Prefer plain language over your internal jargon.
Make answering effortless
Open text boxes feel thorough but they are work, and work lowers completion. Where you can, offer choices — a scale, a set of options, a yes or no. Save the one open question for the end, when you have earned it.
A rating scale gives you numbers you can compare over time; a well-placed open box gives you the quote that explains the number.
Order it like a conversation
Start easy and specific, warm up, then ask the harder or more personal questions once the person is invested. Front-loading demographics or a giant text box makes people quit before the questions that matter.
- Open with a quick, concrete question anyone can answer.
- Group related questions so the mind does not jump around.
- Leave sensitive or optional questions for last.
Respect their time and say so
Tell people how long it takes and keep your word. A survey billed as two minutes that runs ten trains people to ignore your next one. Shorter surveys, asked more often, beat one exhausting questionnaire a year.
You can build a survey with scales, choices and logic — and see the responses roll in — free at onesol.io/forms.
Cut the Two Faults That Quietly Ruin Answers
The double-barrelled question is the most common wrecker of survey data, and it hides in plain sight. Was the booking process quick and easy asks about two different things at once, so a person who found it quick but confusing cannot answer honestly, and their tick tells you nothing you can act on. Split it: one question on speed, one on clarity. If a question contains the word and joining two ideas, it is probably two questions wearing one coat.
Leading questions flatter you and mislead you in equal measure. How much did you enjoy our excellent service assumes the answer and nudges people towards it, so the glowing numbers you get back are partly your own wording reflected at you. Ask how would you rate our service and let the scale carry the judgement. The point of a survey is to learn something you did not already believe, and a leading question guarantees you learn nothing.
Watch for the loaded assumption too. Which of our new features do you use most fails for the person who uses none of them, because it gives them no honest way through. Offer a none of these or add a prior question that lets non-users out gracefully, so you are not forcing a false answer that then pollutes every chart you build from it.
Order and Length Decide Whether People Finish
People answer surveys in the order you present them, and they tire as they go, so spend their attention wisely. Open with the easy, engaging questions that pull them in, put the questions you care about most in the first half while goodwill is high, and leave demographics and anything sensitive for the end, once they are already committed. A survey that opens with please state your age and income loses the very people whose opinions you wanted.
Respect the length you implied. If you say two minutes, build a survey that genuinely takes two minutes, because the person who agreed to two minutes and hits question fifteen simply closes the tab, and a half-finished response is often worse than none. A tight survey of six sharp questions almost always yields more usable data than a sprawling one of twenty, because far more people reach the end.
Group related questions so the respondent's mind is not thrown from topic to topic. All the questions about the product together, all the questions about delivery together, all the questions about support together. Each jump between subjects costs a little attention, and across a whole survey those small costs add up to people quietly giving up before the questions you most wanted answered.
Choose the Question Type That Fits the Answer
Closed questions, meaning multiple choice, scales and yes-or-no, are easy to answer and easy to count, which is why the bulk of a survey should be closed. When you offer a rating scale, keep it consistent throughout, so that five always means the best and one always means the worst; flipping the direction partway through is how you end up with data that looks positive and is actually the opposite. A balanced five-point scale with a labelled neutral middle is a dependable default.
Open text questions are precious and expensive, so ration them. One well-placed what is the single thing we could do better at the end of a survey will surface insights no checkbox ever could, in the respondent's own words. Ten open questions, by contrast, feel like homework and get abandoned. Use closed questions to measure how many, and one or two open questions to understand why.
If your survey is longer than a few questions, consider a small, honest incentive and be upfront about it. A prize draw or a discount code, mentioned at the start alongside the true time it takes, lifts completion without buying you the sycophantic answers a heavy reward can. The best incentive of all, though, is telling people what you will do with their answers, because most will spend two minutes to be genuinely heard.
- One idea per question, never two joined by and
- Neutral wording that does not hint at the answer you want
- A consistent scale direction from the first question to the last
- Easy questions first, demographics and sensitive ones last