Key takeaways
- →Ask an open question when you do not yet know the answer space; use multiple choice when you already know the options and need to see the distribution.
- →Open text costs roughly three times as long to run and read: budget 90 seconds to collect and two minutes to process out loud.
- →Multiple choice is faster and cleaner, but the options you write become the only answers the room can give — bad distractors quietly manufacture a result.
- →Above about 40 people, open text needs aggregation — a word cloud, clustering or a second voting round — or it turns into an unreadable wall.
- →The strongest pattern is not either/or: collect openly, cluster the answers live, then put the clusters to a vote in the same session.
The choice between open questions vs multiple choice is not a matter of taste, it is a matter of what you already know. If you can write the plausible answers yourself before the session, use multiple choice. If you cannot, or if writing them would put words in people's mouths, ask an open question and accept the extra time it costs you.
Most facilitators default to multiple choice because it is tidy. It renders as a bar chart, it takes eight seconds to answer, and it never produces a response you have to read out carefully. That tidiness is exactly the risk: a closed question can only ever return one of the answers you thought of.
This piece covers where each format earns its place, how long each one actually takes at different group sizes, the specific ways both fail in a live room, and the two-phase pattern that gets you the honesty of open text with the decisiveness of a vote.
The short answer: what each format is really for
Multiple choice measures distribution. It answers "how many of us think X?" and it does that job better than any other interaction in a live session, because everyone answers the identical question and the result is comparable across the room and across time.
Open text generates language. It answers "what is going on here?" and it surfaces the specific phrase — "the handover doc is always stale", "nobody knows who signs off" — that no facilitator would have written into an option list. That phrase is often the whole value of the session.
So the diagnostic question before you build any activity is simply: am I trying to count something, or am I trying to find something? Counting is closed. Finding is open. Sessions go wrong when a facilitator uses a closed question to find something, because the room dutifully answers within the frame they were given and everyone leaves believing they have checked.
Side by side
The differences that matter in a live room are time, cognitive load, and what happens to the result afterwards.
| Dimension | Open question | Multiple choice |
|---|---|---|
| Time to answer | 45–90 seconds | 8–15 seconds |
| Time to process live | 2–4 minutes | 30–60 seconds |
| Works above 50 people | ✗ Not included | ✓ Included |
| Surfaces language you did not expect | ✓ Included | ✗ Not included |
| Comparable across sessions | ✗ Not included | ✓ Included |
| Safe with a cold or nervous group | ✗ Not included | ✓ Included |
| Can be scored automatically | ✗ Not included | ✓ Included |
What multiple choice does well
Multiple choice is the lowest-friction contribution that exists. A participant who is half-listening, joining from a phone on a train, or new enough to be nervous can still tap an option. That matters more than it sounds: the first interaction of a session sets the participation rate for everything after it, and a closed question almost never fails to get a response.
It is also the only format that gives you a clean baseline. Ask the same five-option confidence question at the start of a training day and again at the end, and the shift in the bars is a genuine measurement. Ask an open question twice and you get two piles of text that nobody has the appetite to compare.
And it is the right choice whenever the decision is genuinely between known options. If the team has three candidate architectures, four possible launch dates, or two suppliers, there is nothing to discover about the answer space. Put the options up, take the vote, and spend the recovered time on the argument instead of on the collection.
Where multiple choice quietly distorts the room
The options are the question. If you ask "what is slowing the team down?" and offer "unclear priorities / too many meetings / tooling / dependencies", you will get a result — and you will never learn that the real answer was that two people are quietly doing three jobs each. A closed list makes an absent cause invisible.
Order matters as well. The first option carries an anchoring advantage, and a live result bar that is already leaning one way pulls late voters towards it. If the vote is consequential, randomise option order where you can, and consider hiding the running result until voting closes.
The last defence is a permanent escape hatch. Add "something else" or "none of these" to every consequential multiple choice question, and treat a high share on that option as a signal to stop and re-ask openly. If a fifth of the room picks the escape hatch, your option list was wrong, and the tidy chart is worse than no data.
What open questions do well
Open text is how you find out that the problem is not the problem. It rewards the participant who has been sitting on a specific observation for six weeks and has never had a low-cost way to say it. Given a text box and thirty seconds of anonymity, they will type it.
It also produces artefacts you can use after the session. The exact sentences people write become the wording of the retro action, the title of the workstream, the line in the all-hands deck. Paraphrasing loses that; a direct quotation from the room carries authority that a facilitator's summary never does.
Open questions are especially strong at the start of a discovery workshop, in a retrospective, and in any Q&A where you want the real questions rather than the safe ones. Collect written questions first, let people upvote, and the awkward silence problem disappears — nobody has to be the first to raise a hand.
The cost of open text — and how to pay it
Open text costs three things: answering time, reading time, and facilitator nerve. Answering is 45 to 90 seconds if the prompt is tight; longer prompts do not produce better answers, they produce fewer. Reading is the part people underestimate. Forty responses at eight words each is a genuine two to three minutes of live processing, and if you skim it, the room notices.
Pay the reading cost deliberately. Read six or seven responses aloud verbatim — the ones that are surprising, not the ones that agree with you — then name the two or three patterns you can see and check the pattern with the room. "I'm seeing handover and sign-off coming up repeatedly. Have I got that right?" That single question converts a wall of text into a shared conclusion.
The nerve part is real. An open question can return something uncomfortable in front of everyone, and you cannot un-display it. Ask openly only when you are prepared to acknowledge whatever comes back. If you are not — if the honest answer would be politically impossible to read out — do not fake it with a poll; take it to an anonymous survey and report back later.
Pick the format in ten seconds
Ticked items point to multiple choice. Highlighted items point to open text. If you have ticked and highlighted in equal measure, you almost certainly want both — in that order.
- ✓I can list the plausible answers myself, and I am confident the list is complete
- ✓I need a number I can compare to last quarter's session
- ✓The group is over 50 people and I have under two minutes
- ○I want the exact words people would use, not my words
- ○The topic is new, contested, or nobody has named it out loud yet
- ○I am willing to read the answers out, whatever they say
The two-phase pattern: collect openly, then vote
The false choice in open questions vs multiple choice is treating them as alternatives within a single activity. In practice the strongest live sequence uses both, back to back, in about twelve minutes. You generate with open text and you decide with a vote.
This mirrors the nominal group technique, where individuals write silently before the group ranks anything. Silent generation first protects the ideas that would otherwise be talked over, and the ranking round converts a messy list into a decision the room can own.
1. Open prompt, 90 seconds, anonymous
One sentence, one idea, no examples. "What is the single biggest thing slowing this team down right now?" Anonymity matters here — named responses reliably shift towards the acceptable answer, and the acceptable answer is the one you already knew.
2. Read aloud, 3 minutes
Read seven or eight responses verbatim without commentary. Resist explaining or defending anything. The room needs to hear the raw material before you start shaping it, or your clustering will look like editing.
3. Cluster live, 3 minutes
Name four to six themes out loud and get a nod on each. Say the theme in the room's language, not yours — if three people wrote "stale handover doc", the theme is "handover", not "knowledge management hygiene".
4. Vote on the clusters, 60 seconds
Turn your themes into a multiple choice or dot-vote round. Now the options are legitimate because the room wrote them. This is the step that makes the distribution meaningful rather than manufactured.
5. Take the top two into discussion, 5 minutes
Not the top five. Two, with a named owner for each before you move on. A two-phase activity that ends without an owner produces a lovely chart and no change whatsoever.
Group size changes the answer
Under 15 people, open text is nearly free. You can read every response aloud in ninety seconds and the room can hold all of it in working memory. At this size, defaulting to multiple choice usually costs you more than it saves.
Between 15 and 40, open text still works but needs structure. Cap the response length, use one prompt rather than three, and plan to read a sample rather than everything. Say out loud that you are sampling, so nobody thinks their answer was ignored.
Above 40, raw open text stops being readable in real time, and above 100 it becomes actively counterproductive — people watch a scroll of text nobody can process and conclude that their contribution vanished. At that scale, aggregate: a word cloud for single-word prompts, upvoting for questions, or a clustering pass between activities. Multiple choice becomes the default and open text becomes the exception you make deliberately, once, for the question that most deserves it.
Writing questions that do not lead the room
One idea per prompt
"How well did the session work and what would you change?" is two questions. People answer the second and drop the first, and you lose half your data without noticing.
Make distractors plausible
In a quiz, a wrong option nobody picks teaches nothing. Every distractor should represent a real misunderstanding you have actually heard someone hold.
Avoid the leading adjective
"What did you find most useful?" presumes usefulness. "What would you keep, and what would you cut?" gives permission to say the second thing.
Cap open responses
A 120-character limit forces a sharper answer and makes the results readable on a shared screen. Long-form belongs in a survey, not on a projector.
Always include an escape hatch
"Something else" on every consequential closed question. A heavy vote there is your signal that the option list was wrong.
Running both formats in Session Flo
In Session Flo, an open prompt and a multiple choice poll are two activities in the same agenda, so the two-phase pattern is a matter of ordering rather than tooling. Collect the open responses, keep them on screen while you cluster, and launch the vote from the same session without asking anyone to move.
Two settings do most of the work. Anonymity should be on for the generative round and can be off for the vote, where accountability is helpful rather than chilling. And for anything consequential, hide the live result until the poll closes so that early votes do not steer late ones.
Whatever you run, put the raw responses in the session recap. People who wrote something honest want evidence that it survived the meeting, and a recap that includes the unedited text is the cheapest trust-building move available to a facilitator.
Failure modes and how to recover live
Top Tips
- Only four responses to an open prompt: the question was too broad or too exposed. Re-ask it narrower and anonymous — "one word: what slowed you down this sprint?" — and give it another 45 seconds rather than moving on in silence.
- A wall of near-identical answers: your prompt was leading. Acknowledge it out loud, then ask the inverse question. "That's a strong consensus. What would someone who disagreed say?" This is the fastest antidote to a room that has quietly settled.
- A multiple choice result that is flat across every option: the options overlap, or nobody understood them. Do not interpret a flat chart as balance. Re-ask with two options instead of five, or drop to an open question.
- Twenty per cent choosing "something else": stop and take it as a gift. Ask that group to type what they meant, and reissue the vote after the session with better options.
- Someone types something hostile in an anonymous open round: read it, do not editorialise, and answer the substance underneath it. Skipping it visibly costs you more credibility than the comment itself ever could.
- The reading takes too long and you are losing the room: name the pattern, say you will publish the full list in the recap, and move. Silent skimming is what people interpret as being ignored, not a stated cut.
A default worth adopting
If you want one rule: open the session with a closed question and open the substance with an open one. A quick multiple choice pulse gets everyone's thumb onto the screen at minute two, which raises participation for everything that follows. Then, once people are warm, ask the one open question that the session actually exists to answer.
Close by returning to closed. A short rating or single-choice check at the end gives you the comparable number, takes fifteen seconds, and leaves the room with the sense that something was measured rather than merely discussed. Open in the middle, closed at the edges — it is a small pattern, and it holds up across retros, training days, town halls and workshops alike.
Frequently asked questions
Should I use open questions or multiple choice for a retrospective?
Both, in sequence. Start with an open prompt so the team names what actually happened in their own words, then cluster the responses into four to six themes and vote on which one to fix. A retrospective that opens with a closed list of themes tends to confirm the facilitator's assumptions rather than test them.
How long should I give people to answer an open question live?
Ninety seconds for a single, tightly-worded prompt. Under 45 seconds you get the first thing that came to mind; beyond two minutes response quality does not improve and the room disengages. Show a visible countdown so people pace themselves, and warn at fifteen seconds remaining.
Does anonymity change what format I should use?
It changes open text far more than it changes multiple choice. Named open responses drift towards what is safe to say, which defeats the point of asking openly. Closed votes are less sensitive because the options are pre-approved, so anonymity mainly affects speed there rather than content.
How many options should a multiple choice question have?
Four or five, plus an escape hatch. Three rarely covers the real spread, and beyond six people stop reading and pick from the top of the list, which inflates the first option. If you genuinely need eight options, you are probably asking a ranking question rather than a choice question.
What do I do with open responses after the session?
Publish them unedited in the recap, alongside the themes you named live. This proves that nothing was quietly dropped and gives anyone who was not in the room the same raw material. Then attach the top two themes to a named owner and a date, or the exercise stops being worth running next time.
Can I use open text with more than 100 people?
Only with aggregation. Use a one-word prompt rendered as a word cloud, or collect questions and let people upvote so the room does the filtering for you. Displaying an unfiltered scroll of 200 responses is worse than not asking, because people see their contribution disappear in real time.
Keep reading
- the collect-then-vote pattern in detail — The two-phase sequence described here is covered step by step, including timings for larger rooms.
- reading open text at scale with word clouds — For groups above 40, aggregation becomes essential to keep open responses readable.
- writing distractors that teach something — Option design matters most in quizzes, where a wrong answer should represent a real misconception.
- how the first option anchors a live vote — Why option order and visible running totals change the result you get.
- Session Flo plans for running open and closed activities together — Both formats live in the same agenda, so the two-phase pattern needs no extra tooling.