How Many Questions Should Be in a Questionnaire?

The 30-second-per-item math

Respondent time is the currency a questionnaire spends. A Likert item takes 7-10 seconds once a respondent is oriented; a multiple-choice question with real reading takes 20-30 seconds; yes-no items run 5-10; and an open-ended question costs 60-90 seconds to answer properly. A 12-item closed instrument therefore takes 4-6 minutes, and every open-ended question roughly doubles the time of the slot it occupies. Demographics add 1-2 minutes for a standard 4-6 item block.

The budget that matters is the 10-minute attention ceiling. Voluntary survey research consistently shows abandonment clustering once an instrument passes roughly ten minutes, and each lost respondent costs a class project real data — a 10 percent abandonment on a 40-response target is four missing cases, which is the difference between a clean percentage and a hedged one. For course projects, target a total time under eight minutes: enough for 15-20 well-chosen items, comfortably under the cliff.

The numbers at a glance

Question counts by survey goal — the range, and where each row actually shows up in coursework. Skim it first; the sections below explain why the range exists and which row applies to you.

Build from the analysis plan backward

Every question on a defensible questionnaire has a planned destination: a variable, a chart, a test. The exercise takes twenty minutes at the outline stage — list the analysis table first, then write the questions that fill it. Age and year become demographic slices; five Likert items become one scale mean; usage frequency crossed with satisfaction becomes the cross-tab the discussion section needs. If you cannot name the chart or the statistic a question feeds, the question is decoration, and decoration is what respondents abandon surveys over.

Working backward also right-sizes the instrument in both directions. The common overbuild is six variations of one construct — keep two, one reverse-worded, and spend the saved slots on a variable you can actually cross. The common underbuild is a questionnaire with no demographic columns at all, which leaves the discussion section nothing to slice by and no comparison to make. Fifteen items with a complete analysis table beats forty items with vibes, every time the grades come back.

Why Likert scales need more items than they seem to

A single question cannot measure a construct. Attitudes like stress, satisfaction or confidence are multi-faceted, and measurement convention asks for 5-8 items per scale before internal-consistency reliability reaches the threshold most instructors expect — alpha near .70 is the usual benchmark. That is why a psychology-class instrument measuring three constructs legitimately runs 20-30 questions: it is three scales of 6-8 items plus demographics, not one long list. Students who try to measure satisfaction with two items get a fragile scale the write-up has to apologize for.

Item craft matters as much as count. One item per facet of the construct; roughly a third reverse-worded so agreeable respondents cannot straight-line down the page; no double-barreled items — stressed and overwhelmed measures two things and tells you nothing about either. The pilot catches wording failures that drafting cannot: three to five classmates, thinking aloud, timer running. Any item that takes over 30 seconds to parse, or that two piloters read differently, gets rewritten before the real data collection starts.

Open-ended questions: the expensive luxury

Open-ended questions cost respondents 60-90 seconds each and cost you far more afterward — 2-3 minutes of coding per response is a realistic planning figure. The arithmetic is unforgiving: 40 responses times three open questions is 4-6 hours of coding for a class project, which is why the standard advice caps open ends at two or three and places them after the closed block, where an exhausted respondent's exit costs less data.

Two substitutions recover most of the value at a fraction of the cost. Closed questions with an other, please specify option capture the long tail of multiple-choice answers without free-text burden. And one catch-all open end at the very end — anything about this topic we should have asked? — reliably produces the best qualitative material in class surveys, because respondents who made it that far want to talk. Plan the code book before collection: three to five themes you expect, and a category for the ones you do not.

Ordering, layout, and the pre-flight pilot

Order does measurable work. Consent and screeners come first, easy engaging questions build momentum early, sensitive items go late once rapport exists, and demographics close the instrument unless one is needed as a screener. Group questions by topic so respondents can hold one frame at a time, keep one question per row, and break the page every 8-10 items. Assume most respondents will answer on a phone — check the layout on one before launch, because a mobile-broken matrix question costs you its data entirely.

The pilot is the cheapest quality step in survey research: three to five classmates, timed, thinking aloud, covers consent to thank-you. Fix anything two or more piloters stumbled on, re-time, and cut items that the analysis plan cannot justify. The whole protocol costs about 30 minutes and routinely removes three to five items — which is exactly the difference between an eight-minute and a twelve-minute instrument before your real respondents ever see it.

Run your own numbers before you submit

Guidelines are ranges; your document is a fact. Paste your current draft into the word counter below and the page estimate recalculates live for double-spaced and single-spaced formats — no account, nothing uploaded, the text never leaves your browser. Check the word count rather than the page display in your editor: word processors pad pages with headings, breaks and whitespace, and instructors check words.

If the count lands below your target, add analysis before adding sources: two more sentences of interpretation in your three weakest paragraphs usually beats one new summary paragraph. If it lands above, cut your most summarizing paragraph — the one that retells a source instead of using it. Both edits take ten minutes and both make the paper better while fixing the number.