Types of survey questions and when to use each

The question type decides what you can do with the answers. Ask for a rating and you can average and compare it; ask for a ranking and you learn priorities but not how far apart they are; ask an open question and you get reasons, at the cost of reading every one. Choosing the type is choosing the analysis.

So start from the result you want to report, and pick the type that produces it. This guide groups the common types by what they give you, with when to use each and the mistakes to avoid. The names follow Plumeform's question picker, but the advice holds in any tool.

Start from the result you want

Each kind of finding has a type that produces it:

• How many people chose something: multiple choice or checkboxes.
• A level or a score: a rating, a slider or NPS.
• An order of priorities: ranking, or best–worst for longer lists.
• How much more one thing matters than another: constant sum.
• Which features people expect and which delight them: a Kano pair.
• What a price feels like: price sensitivity.
• A preference order from simple choices: pairwise comparison, or a Q-sort for statements.
• How people group things, or where they click first: a card sort, or a click map.
• Reasons, in people's own words: short or long text.
• An exact quantity or date: a number or date question.
• Somewhere to send something: an address question.

If the finding you have in mind is not on the list, write it out as the sentence you would put in a report, and work backwards from that sentence to the question.

Choosing one option or several

Multiple choice asks for one answer from a list. Use it when the options are mutually exclusive: a plan, an age band, the main reason for something. When your list may not cover everyone, add an Other option with a text box so people can say what fits. For a plain yes-or-no question, the picker's Yes / No sets one up with the two answers in place.

Checkboxes allow several answers: every tool someone uses, every feature they tried. Be careful about what you conclude. Checkboxes tell you what applies, not what matters most, and in a long list an unticked box can mean “no” or “did not read that far”. If you need a priority, ask for it separately. Plumeform lets you set a minimum and a maximum number of selections, which turns “pick up to three” into a rule rather than a request.

Dropdowns suit long lists that people know by heart, such as countries; Plumeform's picker has a ready-made Country list and a US state list, and typing the first letters jumps through either. With five options, a dropdown only hides the choices; a multiple-choice list shows them all at once.

Image choice lets people pick a picture: designs, packaging, anything easier to recognize than to describe. Give every image a clear label, since some respondents will rely on the words.

Scales: how much, how good, how likely

Rating questions ask for a quick evaluation on a short scale of stars, hearts, numbers or faces; in Plumeform you can use from 3 to 10 points. They suit simple judgments, such as how an experience went. Faces, running from unhappy to happy, help where a number asks too much: children, patients, or anyone answering quickly on a phone.

Sliders ask for a point between two labeled ends, such as 0 to 100 from “Not at all useful” to “Extremely useful”. In Plumeform a slider can move in steps or run continuous. A continuous slider is a visual analogue scale: the answer is kept to one decimal, and you can hide the number on the thumb so people place it by feel rather than by picking a round figure. A slider can also carry a unit, such as kilometres or a percentage, and a label under its middle. Nothing is recorded until the respondent moves or presses it, so a slider left untouched is never mistaken for a middle answer.

NPS is the standard question about how likely someone is to recommend you, from 0 to 10. Keep its wording and scale as they are, because the score only compares with other NPS results asked the same way. The comparison of NPS, CSAT and CES covers when it is the right measure.

Likert matrices rate several statements on one shared scale, usually agreement. They are efficient to answer and to analyze, and tiring in large numbers: each row is its own question, so keep grids short. The Likert scale guide covers points, labels and wording.

Semantic differentials place something between two opposites, such as “Traditional” and “Modern”, across several pairs at once. They are a common way to measure how people perceive a brand or product.

Trade-offs: what matters most

Ask people whether things are important and they will often say everything is. Trade-off questions make them choose.

Ranking asks people to put a short list in order. It shows priorities clearly but not the distance between them: first and second may be nearly tied or far apart. Keep the list short, because ordering a long list is hard work, especially on a phone.

Best–worst, also known as MaxDiff, shows a set of options and asks for the most and the least important. Picking the two ends is easier than ordering everything, and it separates items that would all rate as important. Plumeform's best–worst question is a single set; a full MaxDiff study shows several sets in a designed rotation. The message testing template uses one to compare messages.

Constant sum gives people a budget of points to divide between options, which shows how much more one thing matters than another. It suits questions about spending, time and priorities; the budget allocation template is built around one.

Research methods built into one question

Some methods are a fixed set of questions with an analysis attached. These come as one question type each.

Kano pair asks about each feature with it and without it, and places it as a must-have, a performance feature, an attractive one or one nobody minds: the Kano model.

Price sensitivity asks Van Westendorp's four prices for one offer and draws the acceptable range. It measures what prices feel like, not what people would buy.

Pairwise comparison asks which of two someone prefers, over many pairs, and turns the wins into a share of preference.

Q-sort places statements on a fixed grid from most agree to most disagree, with few places at either end, so people cannot agree with everything.

Card sorting asks people to put cards into groups, yours or ones they name; a card split across groups is one your labels leave unclear.

Click maps collect clicks on a picture as a heat map. With one click allowed it is a first-click test: where people go first to do something, and how long it took.

Highlighting asks people to mark the words of a passage they like and dislike, which shows which phrase did the work when a rating of the whole text cannot.

Conjoint asks for choices between whole offers and works out what drove them: conjoint analysis.

Reaction time times how fast people sort words, or runs an Implicit Association Test.

A page's reading material can also be shown once, for a few seconds: a five-second test.

In Plumeform the Kano pair and the five-second test are free; the other research questions need Pro to publish.

Words and numbers

Short text is for a word or a phrase: a job title, a product name, a single suggestion. Set to a website, it checks that what was typed is a web address, and phones show the keyboard with the slash and the dot. Long text is for explanations. Open questions cost the most, for respondents and for you, so use them where the reason matters more than the count; the guide to open-ended questions covers wording and analysis.

Number questions take digits, with optional minimum and maximum limits, for counts and amounts. When people are unlikely to know an exact figure, such as their monthly spend, ask for a range with multiple choice instead.

Email questions check that what was typed is an email address, and an email question can be used to send respondents a confirmation. Phone asks for a telephone number. Ask for contact details only when you will use them, and say why.

Date questions open a date picker, for a visit, a purchase or an event. The same question can ask for a time of day, a date and a time together, or a range with a start and an end, which checks that the end does not come before the start.

Address asks for a postal address in parts: the street and the city, and, if you keep them, a second line, a region, a postal code and a country. Browsers fill the parts in from a saved address, and results count the answers by country. Like any contact detail, ask for one only when you will post something.

Signatures, files and voices

Signature lets someone draw their signature, for sign-offs and acknowledgements. Anyone who cannot draw it, on a keyboard or with a screen reader, can type their name instead, and it is set on the signature pad. The same pad can be a sketch: a larger canvas with undo, for a drawing rather than a name. Drawing needs a mouse, a pen or a finger, so keep a sketch optional, or ask a text question beside it. File upload collects a document or an image, and the same question can record a voice answer in the browser instead, for people who would rather say it than type it; in Plumeform both need Pro, because stored files cost storage. Ask for a file only when nothing else will do, and never for more personal information than you need.

Mistakes that are easy to make

• Checkboxes where you needed a priority. You learn what applies and have no way to rank it afterwards.
• Rating everything. Ten items all rated “important” tell you less than one best–worst question over the same ten.
• An open question that should have been a list. Free text for a small, known set of answers means cleaning work a list would have saved you.
• Asking for a number people do not know. The precision you ask for is not the precision you get.
• A dropdown for a handful of options. It hides choices people need to see side by side.

Question types in Plumeform

Plumeform has 30 question types, and 21 of them are free; file upload and the research instruments (price sensitivity, pairwise comparison, Q-sort, card sorting, click maps, highlighting, conjoint and reaction-time tasks) need Pro to publish. Any question can be converted to another type from the question editor, so if a pilot shows you picked the wrong one, you do not have to start again. The question picker can also insert validated scales, such as the System Usability Scale and NPS, with their standard wording.

Two more things are worth knowing. Any question can be required, or soft-required, which asks once whether the respondent wants to continue without answering and then lets them. Soft-required suits sensitive questions, where forcing an answer invites people to pick anything just to get past it. And for standard profile questions, the builder's Demographics section adds age, gender, education, country and other common questions as an editable set, at the start or the end of the survey.

To see the types at work, browse the survey templates. Each is a working questionnaire you can preview and answer before opening it in the builder, and the pricing page lists exactly what each plan includes.