Key takeaways
- →A running tally is not a neutral display: people who have not yet voted can see which way the room is going, and some of them follow it.
- →Lock results whenever the poll is measuring an independent judgement, an estimate, or a position somebody might be reluctant to hold alone.
- →Show results live when the point is shared awareness rather than accuracy, such as pulse checks, word clouds and warm-up questions.
- →The reveal is a facilitation moment worth five seconds of silence, not a chart that appears while you are still talking over it.
- →In quizzes, hide the answer distribution during the question and show it immediately after, so the distractors do the teaching.
- →If you cannot lock results, ask everyone to answer before you display anything, and enforce it with a visible countdown rather than a request.
The real-time results vs reveal question is usually treated as a display preference and it is not. A running tally that the room can see while voting is still open changes the vote it is counting, because the people who have not answered yet now have information they would not otherwise have had: what everyone else appears to think.
The short version. Lock the results whenever the poll is measuring an independent judgement, an estimate, a confidence level, or a position that somebody might be uncomfortable holding on their own. Let results build live when the point of the activity is shared awareness rather than accuracy, and when there is no wrong answer for anyone to conform to.
This piece covers exactly what a visible count does to a room, which activities survive it and which do not, how to run a locked poll without losing the energy of a live tally, how to stage the reveal so it lands, and the failure modes on both sides.
What changes when the room can see the count
Three things happen, and they compound. The first is straightforward conformity: a visible majority pulls undecided voters towards it, which is the bandwagon effect operating in a form so direct it is almost a laboratory demonstration. You do not need many undecided people for this to matter. In a 40-person poll, ten genuinely undecided voters swinging together can turn a 55/45 split into a 70/30 one.
The second is anchoring on the early pattern. The first ten responses in a live poll are not a random sample of the room, they are the fastest people, who are usually the most confident and often the most senior. Whatever they choose becomes the reference point that later voters adjust away from rather than the answer they would have given cold.
The third is silent withdrawal. Somebody who was about to click the option currently sitting at four per cent often does not click it. They do not switch to the majority, they simply stop voting, and you never see that in the results. A locked poll with a 92 per cent response rate and a live poll with a 71 per cent response rate can produce the same chart while telling you completely different things.
When to lock the results
Lock anything where you want independent judgement. Estimates of effort or cost, confidence in a plan, likelihood of hitting a date, risk ratings, prioritisation votes: in all of these the value of the aggregate depends on the answers being independent. A crowd of estimates beats the individual experts inside it only when the errors are uncorrelated, and a visible tally correlates them on purpose.
Lock anything with a social cost. 'Do you think this launch date is realistic?' is a question people answer differently when they can see that the first eleven responses said yes. Anonymity helps but does not solve it, because the pressure comes from the visible distribution, not from your name being attached to a response.
Lock anything you intend to compare before and after a discussion. If you want to ask the same question at the start and the end of a session and show the shift, both measurements have to be taken under the same conditions, and a live tally in round one contaminates the baseline you are trying to move.
When live results are the better choice
Show results live when the activity is about shared awareness. A word cloud building as people type is genuinely better live: watching the same three words swell while forty people contribute is the point of the exercise, and there is no correct answer to conform to. The same applies to pulse checks, energy readings and mood questions, where the room learning about itself in real time is exactly the outcome you wanted.
Live also wins for warm-ups and low-stakes participation. 'Coffee or tea', 'which office are you dialling in from', 'how many years have you been here': visible results here create momentum, because seeing the count climb pulls in the people who were not sure whether participation was expected. That momentum is worth more than the purity of the data.
The third case is throughput. In a room of 200 with a two-minute window, a live count is your only signal that responses are still coming in and that you should not move on yet. Locked polls hide that, so if you lock them at scale you need a visible response counter that shows how many have answered without showing what they chose.
What locking actually changes
The same question, asked of the same 45 people, with and without a visible tally. The differences below are the ones facilitators consistently report noticing once they start running both versions of a question.
✗ Live tally visible while voting
- •Fast, confident voters set the reference point for everyone else
- •Minority options lose late votes and end up looking weaker than they are
- •Some people quietly stop voting rather than sit at four per cent
- •The chart is finished before the discussion starts, so there is nothing to reveal
- •Re-asking the question later has no clean baseline to compare against
✓ Results locked until the window closes
- •Everyone answers from their own position rather than the room's
- •Minority views survive to the reveal, which is where the useful conversation is
- •Response rate goes up because there is no visible bandwagon to defect from
- •The reveal becomes a moment you can facilitate rather than a chart that faded in
- •Before-and-after comparisons mean something because conditions match
How to run a locked poll without losing the energy
Say the rule out loud before you open voting
'Results are hidden until everyone has answered, so answer what you actually think rather than what you think the room thinks.' Ten seconds of framing prevents the entire round being read as a technical fault when nothing appears on screen.
Show progress, not preference
Display a response counter, 31 of 45, so the room can see the activity is working and so latecomers feel the pull of the count. This gives you the momentum benefit of a live tally without the distribution that causes the bias.
Use a hard countdown, 60 to 90 seconds
Without a visible tally there is no natural sense of when a poll is finishing. A countdown supplies it. For a single multiple-choice question 60 seconds is plenty; for anything requiring a numeric estimate, give 90.
Close on the clock, not on the stragglers
Waiting for the last three people teaches everyone that deadlines are soft and adds two minutes to every subsequent round. Close at zero, state the response rate, and move. If the rate is below 70 per cent the question was unclear, not the room.
Reveal in silence
Put the chart up and say nothing for five seconds. People need to read it before they can react to it, and a facilitator narrating over the reveal denies them that and tells them what to think about it.
Ask the minority first
'Eight people picked the third option. Without saying who, what is the case for it?' Locking the poll is what made those eight votes exist. Going to them first is what makes locking worth doing.
Staging the reveal so it lands
A reveal is a small piece of stagecraft and it costs about ninety seconds. The value is that it makes the result an event in the session rather than a chart that was already there. That matters because the moments people remember from a session are disproportionately the peaks and the ending, and a well-run reveal is a cheap way to manufacture a peak in the middle of an otherwise flat agenda.
The sequence below works for both polls and quiz questions. The only part people routinely skip is the prediction, and it is the part that does most of the work: asking the room to guess the split before you show it creates a stake in the answer and makes the gap between expectation and reality visible.
Frame the question, 15 sec
State what you will do with the answer. 'This decides which of the two we prototype first.'
Voting window, 60-90 sec
Response counter visible, distribution hidden, countdown on screen.
Predict, 20 sec
'Hands up if you think option A is over half.' Now the room has a position on the result.
Reveal and hold, 5 sec
Chart up, facilitator silent. Let people read before anyone speaks.
Go to the minority, 60 sec
Ask the smallest sensible group for its reasoning before the majority explains itself.
Name the decision, 30 sec
Say what the result changes. A poll that changes nothing trains people not to answer the next one.
Quizzes work differently from opinion polls
In a quiz there is a correct answer, so conformity is not just a bias, it is cheating by inference. Never show the answer distribution while a scored question is open. If people can see that 60 per cent have chosen B, the question stops measuring what anyone knows and starts measuring how quickly they can read a bar chart.
Show the distribution immediately after the question closes, though, and it becomes the most useful teaching moment in the whole activity. A distractor that pulled 40 per cent of the room is a specific, named misunderstanding you can address in thirty seconds while everyone is still holding the question in their head. That is worth more than the score.
Leaderboards follow the same logic at a slower cadence. Updating a running leaderboard after every single question tends to demoralise the bottom half by question three. Showing it after every third question, or only at the halfway point and the end, keeps the competitive pull without turning eight rounds into a public ranking exercise.
How long to leave a poll open
Timings that hold up in practice. These are windows, not targets: close on the clock and state the response rate rather than waiting for the last few people.
Failure modes on both sides
The blank screen
You locked results and said nothing, so the room assumes the poll is broken and stops answering. Always show a response counter and announce the rule before voting opens.
The anticlimax
You built up a reveal for a question nobody cared about. Reserve the full reveal sequence for the two or three questions in a session that actually change something.
The contaminated baseline
Round one ran live, round two ran locked, and you presented the shift as a change of mind. It is not comparable. Keep conditions identical across a before-and-after pair.
The frozen minority
Results are locked but you still ask people to raise their hand for the unpopular option. You have reintroduced the exposure you just removed. Keep the follow-up anonymous too.
The endless window
No countdown, so the poll stays open until the facilitator loses patience. Rounds stretch, the room disengages, and the session runs fifteen minutes over.
The unused result
A locked poll, a good reveal, and then nothing happens with the answer. Say out loud what the result changed, or people will answer the next one at random.
What to measure to know which is working
Response rate is the first and most honest signal. Run the same question type both ways across a few sessions and compare how many people answered. If locking the results lifts response rate by ten points or more, you had a conformity problem you could not see.
The second signal is minority survival. Look at how much of the vote the least popular option holds. Live tallies compress minorities towards zero, and a distribution that always resolves into one dominant option is more often an artefact of the display than a genuinely united room.
The third is what happens in the discussion afterwards. A locked poll that reveals a genuine 60/40 split gives you a real conversation with two sides in it. If your polls never produce that, check whether the room has been watching each other vote for the last six months.
Frequently asked questions
Does anonymity solve the bandwagon problem on its own?
No. Anonymity removes the cost of being identified with an unpopular answer, but the pull towards the visible majority comes from seeing the distribution, not from your name being on your response. An anonymous poll with a live tally still drifts. If you want independent answers you need both anonymity and hidden results until the window closes.
Should I ever show a partial result mid-vote?
Show a count of responses, never a breakdown of them. 'Thirty-one of forty-five have answered' gives the room the momentum and the urgency of a live tally with none of the influence. It also tells you honestly whether to hold the window open another twenty seconds.
What if my tool cannot hide results?
Ask everyone to answer before you put anything on screen, then display it. It is weaker, because the fast voters still see the early shape, but it recovers most of the benefit. Reinforce it with a visible countdown so the instruction has a deadline attached rather than being a polite request.
Do live results work better for large audiences?
They feel better and measure worse. In a room of 200 the growing chart is genuinely energising, and the conformity pressure is also at its strongest because the visible majority looks overwhelming. For large rooms, lock the questions that matter and run the warm-ups live.
When should a quiz leaderboard update?
Not after every question. Update every third question, or at the halfway point and the end. Continuous updating tells the bottom half of the room by question three that they cannot win, and their participation drops for the remaining rounds.
How do I stop people waiting to see how others vote?
Remove the information they are waiting for. With results hidden there is nothing to wait for, and a countdown converts hesitation into a decision. If people are still holding back, the question is probably one they feel exposed answering, which is a framing problem rather than a timing one.
Keep reading
- how the bandwagon effect distorts live polls — The mechanism behind a drifting live tally.
- running a live poll in a meeting — The end-to-end mechanics of a single poll round.
- what changes between anonymous and named responses — The other lever on honest answers, and its limits.
- leaderboards versus participation-only scoring — How often to show a running score in a quiz.
- Session Flo live polls and quizzes — Locked results, response counters and countdowns in one place.