Closed-ended question

The two rules for an option list

Almost every problem with a closed question comes from breaking one of these.

Mutually exclusive

No respondent should fit two options. Age bands written as "18–25, 25–35, 35–50" put a 25-year-old in two places, and whichever they choose, your data is wrong. Write "18–24, 25–34, 35–49".

Exhaustive

Every respondent must have somewhere honest to go. If your list of transport modes has no "I do not travel to work", the people it applies to will pick something, and you will not know which of your numbers is inflated. "Other (please specify)", "None of these" and "Prefer not to say" are not padding; they are what stops the rest of the list lying to you.

Single select versus multiple select

A pick-one question yields percentages that sum to 100 and can be charted directly. A pick-several question yields percentages that sum to more than 100, and every figure has to be read as "% of respondents who selected this", not "% of answers". Mixing the two up in a report is common and produces charts that quietly imply the wrong thing.

If you need a priority order rather than a set, that is a ranking question. If you need the same options answered for several items, that is a matrix.

A worked example

"How did you hear about this consultation?" — single select.

• A letter or leaflet through my door
• Social media
• A local news site or newspaper
• From someone I know
• At an event or a public meeting
• Other (please specify)
• I cannot remember

Exhaustive, because "other" and "cannot remember" both exist. Mutually exclusive, because someone who saw it on social media and then discussed it with a neighbour has a genuine ambiguity — which is the argument for making this question multiple select instead, and deciding that deliberately rather than by accident.

Common mistakes

• Unbalanced scales. "Excellent / very good / good / fair / poor" has four positive points and one negative. It will flatter you. See Likert scale.
• Order effects. On screen, early options get picked more. Randomising the option order spreads that out — but only where the order carries no meaning. Never randomise a scale.
• Leading option sets. Three versions of "yes" and one "no" is not a question, it is a result.
• Too many options. Past a dozen, a list becomes a search task. Group them, or use a dropdown, or ask a filtering question first.
• Making everything required. A forced answer on a question somebody cannot honestly answer is worse than a blank.

In NumoForms

Closed-ended types include Multiple choice (pick one), Checkboxes (pick several), Dropdown, Rating, Linear scale, Recommend score (0–10), Matrix, Ranking and Date.

Three details that affect how you write option lists here. Multiple choice and checkboxes render as bento option tiles by default, with a plain list as the alternative — tiles are uniform height and reflow to one per row for labels over about 110 characters, which is a good reason to keep labels short. Options can carry an image and a description line, so the explanation does not have to live in the label. And answer randomisation is seeded per page load, so going back does not reshuffle the list, and options marked fixed — "None of these", typically — stay in place.

The question types page lists all 16 kinds, and open-ended questions covers the other half of this pairing.