Multiple Choice Question
A multiple choice question presents a list of options and allows the respondent to select more than one, capturing combinations rather than a single answer.
Key takeaways
- Variable selection counts break the fixed maximum a single-answer question assumes.
- Cap selections or normalise the total, or volume alone raises scores.
- Vague benefit lists invite checking everything; concrete nouns invite careful picking.
- Store selected option ids individually, not only the summed question score.
- An empty multi-select cannot distinguish none apply from skipped or unclear.
In depth
With multiple selections allowed, the score for one question is a sum whose length varies by respondent. Someone picking one option contributes its points; someone picking four contributes all four. That breaks the assumption behind a fixed maximum, so builders either cap the number of selections, normalise the question total to a fixed ceiling, or accept that heavy selectors score higher by volume. Some models flip the sign, awarding points for the absence of an option, which turns the list into a checklist of gaps.
How many options people select depends on how the list reads more than on how many options exist. A list of concrete nouns invites careful picking; a list of vague benefits invites checking everything. Ordering matters too, since items near the top get chosen more often when the list is long. Allowing many selections gives richer segmentation but weakens comparison between respondents, because two people describing the same situation can select different numbers of items.
The selections themselves are often more useful than the points they carry. A respondent who checks three competing tools has told you the integration story; one who checks two pain points has written the first line of the follow-up email. Store the selected option ids as separate fields rather than only the summed score, so segments can be built on any combination later. That is also what lets result copy name the exact items a respondent chose.
Multi-select is the wrong tool when the options are genuinely exclusive, because allowing an impossible combination produces records nobody can interpret. It is also weak as an opening question: the respondent has not yet committed and the format asks them to evaluate everything on screen. Selecting nothing is ambiguous too, since it can mean none apply, or the question was unclear, or the person skipped it, and the stored data looks identical in all three cases.
Example in practice
How to measure it
Start with the average number of selections per respondent. A figure close to one means the question could have been a single-select; a figure close to the option count means the list is not forcing choices. Then look at each option's selection rate: an option nobody picks is either wrongly worded or genuinely absent from the market, and an option almost everyone picks carries no information.
Co-selection is the more interesting analysis. Count which pairs of options get chosen together and you find the real segments in your audience, often clusters nobody designed for. Watch the empty-response rate as well: if a meaningful share of respondents select nothing, add an explicit none of these option so the answer becomes a fact rather than a gap in the record.
Common mistakes
The recurring design fault is options that overlap. A list offering email marketing, newsletters and lifecycle campaigns forces respondents to guess which label covers their situation, and two people doing the same thing tick different boxes. Write options that a respondent could not reasonably assign to two categories, test the list by asking a colleague to describe their setup, and merge any pair that keeps getting selected together.
The second is forgetting the auto-advance behaviour. A multi-select on a flow where every other question advances on tap will register the first click and jump, so respondents never reach their second choice and the data quietly under-reports. Give multi-select an explicit continue button, make the selected state visually obvious, and say select all that apply in the question itself rather than in small print underneath.
Frequently asked questions
When should I use a multiple choice question?
Use it when several answers can legitimately be true at once, such as the channels, tools, or goals a respondent has. It captures combinations that a single-answer format would force you to lose.
How does multiple choice affect scoring?
Each selected option can contribute its own weighted points, so the score reflects the full combination a respondent chooses. This enables nuanced, multi-factor scoring rather than a single bucket.
How many options should a multiple choice question have?
Keep the list scannable, typically four to seven distinct, non-overlapping options. Too many choices increase cognitive load and dilute the clarity of your scoring.
Should we limit how many options a respondent can select?
Limit it when the question is meant to reveal priority rather than inventory. Asking for the top two pain points produces a ranked signal; asking for all of them produces a list where everything matters equally. If you need both, ask for everything that applies and follow with a short question naming the most important one.
How do we score a question where selections vary in number?
Two workable approaches. Normalise by dividing the respondent's total for that question by the maximum possible, so every question contributes the same weight regardless of how many boxes were ticked. Or score only the highest-value selection, treating the rest as profile data. Whichever you choose, apply it consistently, because mixing both across questions makes the total impossible to explain.
Where in the flow does a multiple choice question belong?
In the middle, after the respondent has answered one or two easy questions and before any contact fields. By that point they are invested enough to read a list, and the profiling detail is worth the extra effort. Opening with one costs starts, and placing one after the email field means fewer people ever see it.
Should we include a none of the above option?
Yes, whenever selecting nothing is a real answer. Without it you cannot tell refusal from omission, and any respondent who genuinely uses none of the listed tools has no honest way through the screen. Place it last, style it differently from the others, and clear the other selections when it is chosen so the record stays consistent.
Does option order change the results?
It does once the list passes about five items, because people scan from the top and stop when they have enough. Randomising the order per respondent removes the bias but makes result copy harder to write and screenshots inconsistent. A practical compromise is to fix the order, put the least expected options first, and check the selection rates for a top-of-list skew.
How does this differ from a single select question?
A single select returns one value and a fixed maximum; a multiple choice returns a set and a maximum that depends on the respondent. That difference changes the scoring, the advance behaviour and the way result copy can reference the answer. Use multiple choice only when several options can be true at once for the same person.