Key takeaways
- →Give each person roughly one dot per three options, capped at five - too many dots and everything scores, too few and the result is noise.
- →Hide the running tally until everyone has voted; visible dots pull later voters towards whatever is already leading.
- →Cluster duplicates before you vote, or a popular idea written three different ways will split its own support and lose.
- →Read the spread, not the total: eleven dots from eleven people is a consensus, eleven dots from three people is a lobby.
- →The person who will make the final call should vote last or not at all, and the facilitator never votes.
- →Dot voting produces a shortlist and a signal, not a decision - somebody still has to choose, own it, and say so before the session ends.
Dot voting is a way of narrowing a long list of options fast: every participant gets a fixed number of dots, spends them on the options they favour, and the list is reordered by how many dots each option collected. It takes about two minutes of actual voting, works with five people or five hundred, and is the fastest legitimate way to get from forty ideas to a shortlist of five.
It is also wrong more often than facilitators admit, and almost always for the same three reasons: people could see the dots accumulating and drifted towards the leader, duplicate options split their own support, or the group read the raw total without asking how many separate people were behind it. All three are fixable in the setup, and none of them are fixable afterwards.
This guide covers what dot voting is genuinely good for, how many dots to hand out, how to set up the ballot, a ten-minute run, how to read a result honestly, the variants worth knowing, and how to convert a column of dots into a decision with a name attached.
What dot voting decides, and what it cannot
Dot voting answers one question well: which of these options does this group care most about? That makes it excellent for narrowing a retrospective's improvement list, choosing which three of twelve features get discussed in detail, picking session topics, or agreeing what a team's next quarter should focus on.
It answers a different question badly: which option is objectively best? A dot vote aggregates preference, not analysis. Nobody in the room has read the effort estimates, and a well-argued option that was written down clumsily will lose to a vague one that sounds appealing. When correctness matters more than buy-in - a technical choice, a legal question, a pricing decision with modelling behind it - use the vote to select what gets analysed, not to select the answer.
It is also not consensus. A result of 9-7-6 across three options means the group is genuinely split, and pretending the 9 is a mandate is how you get a decision that gets quietly relitigated for a month. When the top scores are within a couple of dots of each other, say so out loud and treat it as a signal to discuss, not to close.
How many dots each person gets
The default heuristic is one dot per three options, capped at five. Twelve options means four dots; thirty options means five, not ten. Give people too many and almost everything collects at least one, which flattens the ranking and tells you nothing; give one dot each and you get a scatter with no clear front-runner unless the group already agreed before they walked in.
Decide separately whether people may stack dots on a single option. Allowing a maximum of two on one item is the sensible middle: it lets someone express real intensity about one thing without letting a single-issue voter dump their whole allocation. Unlimited stacking turns a prioritisation into a lobbying exercise, particularly when three people from the same team have the same pet project.
The facilitator does not vote. If you also happen to be the product owner, split the roles explicitly - facilitate the vote, then take off the facilitator hat and give your view after the result is visible, framed as your view rather than as the outcome.
Setting up the ballot
Cluster duplicates first, in front of the group
Two cards saying 'faster deploys' and 'reduce release friction' will split six dots into three and three and finish below a weaker option. Read each card, ask whether anything else is the same thing, merge with the group's agreement, and give the merged item a headline the original authors recognise.
Cut anything that fails a stated constraint
If you said the shortlist has to be startable this quarter, remove the items that cannot be, before the vote rather than after. Voting on options you will not action teaches people that the exercise is decorative.
Make every option the same size and specificity
A one-word option next to a three-line option is not a fair ballot - the vague one attracts dots because everyone can project their own version onto it. Rewrite everything to a single line of similar length.
Randomise the order
On a physical wall the top-left cards collect noticeably more dots than the bottom-right; on a screen, the first three items in a list do the same. Shuffle before voting, and if you are running the same vote across two groups, shuffle differently for each.
Say the rules out loud before the timer starts
How many dots, whether stacking is allowed, how long they have, and what happens with the result. Ninety seconds spent here prevents the most common recount, which is somebody having spent four dots when they had three.
The ten-minute run
With twelve to twenty options and a group of eight to thirty, the whole exercise fits comfortably in ten minutes. The part people rush is the silent read, and it is the part that most affects the quality of the result - if half the room is still reading option nine while voting closes, you are measuring reading speed.
Run the vote on people's own devices even when everyone is in the same room. Physical dots are pleasant and they leak: you see where the stickers are going, and so does everyone else. If you must use stickers, hand them out and have everyone place them at the same moment on a count of three, standing back afterwards.
Read in silence (3 min)
Nobody advocates, nobody explains. Everyone reads the full list once so late options get the same attention as early ones.
One clarifying round (2 min)
Questions of fact only - 'what does option seven mean?' - not arguments for or against. Cap it at two minutes and enforce the distinction.
Vote privately (2 min)
Everyone votes at once with the tally hidden. Announce the time remaining at 60 and 20 seconds so nobody is caught mid-decision.
Reveal everything at once (1 min)
Show the full ranking, including the items with zero, and read the top five aloud yourself rather than letting the room scan it silently.
Interrogate the shape (2 min)
Point out the gap between first and second, any option with a high total from few voters, and anything a group of stakeholders backed unanimously.
Reading the result honestly
Totals hide the thing you most need to know. Eleven dots contributed by eleven different people is broad, shallow support - safe, probably uncontroversial, possibly bland. Eleven dots from three people who each stacked is a passionate minority, which is worth listening to but is not a group decision. If your tool reports voter counts as well as dot counts, look at both columns before you say anything about the winner.
Watch for the bimodal option too: the one that several people ranked first and several others gave nothing. That is usually the genuinely interesting proposal, and averaging it away is how groups end up choosing the third-best idea that nobody objects to.
Then check the gap. A clear leader four or five dots ahead of second place is a result you can act on. Two options separated by one dot are, statistically, a tie in a room of fifteen people, and treating a one-dot margin as decisive is the single most common misuse of the method. When it is that close, discuss the two for five minutes and re-vote once - once, not repeatedly, because a group re-voting three times is just waiting for the answer the loudest person wants.
✗ A dot vote that will mislead you
- •Dots land on a shared wall as people place them, one at a time
- •Forty raw cards, several of them the same idea in different words
- •Ten dots each, so nearly every option scores something
- •The head of department votes first, in front of everyone
- •Winner announced from the raw total, no voter counts checked
- •Result photographed, sent round, never mentioned again
✓ A dot vote you can act on
- •Everyone votes privately, tally hidden until the reveal
- •Duplicates clustered with the group before the ballot is fixed
- •Four dots each across twelve clustered options, two-dot stack limit
- •The decision-maker votes last or abstains entirely
- •Totals and voter counts read together, close margins named as ties
- •Top three assigned an owner and a date before anyone leaves
Five biases that quietly corrupt a dot vote
Top Tips
- Bandwagon. Visible dots pull later voters towards whatever is already ahead, and the effect is strongest in the people who were least sure. Hiding the tally until every vote is in removes it entirely, which is the strongest argument for voting on devices rather than a wall.
- Anchoring on the first advocate. If someone argues for option three before voting opens, option three's score is no longer independent. Keep the pre-vote round to questions of fact and hold advocacy until after the reveal, where it belongs.
- Position bias. Items at the top of a list or the top-left of a board collect more dots regardless of content. Randomise the order every time, and if the result surprises you, re-run with a different order before concluding the group has changed its mind.
- Split support. The most damaging and least noticed. Three near-identical options divide their backers and all three finish mid-table, so a weaker but uniquely worded option wins. Clustering before the ballot is the only reliable fix.
- Seniority signalling. In a room where the director's dots are recognisable - a different colour, or simply because everyone watched them vote - you are collecting a poll on what the director wants. Anonymous voting is not a trust problem, it is basic measurement hygiene.
Variants worth knowing
Two-round dot voting handles long lists well. Round one is deliberately generous - five or six dots each across thirty options, purely to cut the list to ten. Round two gives two dots each across the survivors, which forces sharper choices. Announce both rounds at the start so nobody spends round one strategically.
Coloured dots by role are useful, but for diagnosis rather than weighting. Give sales one colour and engineering another and you can see at a glance where the two groups agree and where they do not - that divergence is usually more valuable than the ranking itself. Resist the temptation to make one group's dots count double; if some people's views should carry more weight, say so openly rather than burying it in arithmetic.
Constrained voting adds a budget instead of dots: each person has 100 points to distribute, with a maximum of 40 on any one item. It takes longer and produces a more honest picture of intensity, and it is worth the extra five minutes when the decision commits real money.
Negative dots - one vote each to remove something - are a good addition when a group is deciding what to stop doing. They surface the option that is quietly hated by a third of the room but that nobody wanted to speak against.
Turning dots into a decision
The vote ends and the room relaxes, and that is the moment the exercise usually fails. A ranked list is an input. Someone still has to decide, and if that person is in the room they should say what they are choosing and why while everyone is still there - including the case where they pick the second-placed option, which is entirely legitimate as long as they explain the reasoning.
Take the top two or three, attach a name and a date to each, and note explicitly what happens to everything below the line. 'Parked until October' and 'not doing this' are both fine; silence is not, because silence means every unpicked idea is raised again next quarter.
Send the whole ranking, not just the winners, in the session recap. People who voted for option seven need to see that option seven scored four dots rather than assume it was ignored. In Session Flo the vote and its result travel into the recap automatically, which removes the most common excuse for the follow-up never going out.
Finally, do the boring thing: return to the list in the next meeting for thirty seconds and report what happened. Groups become cynical about voting not because the method is flawed but because the last three votes changed nothing they could point at.
Name the margin out loud
Say 'first and second are one dot apart, that is a tie in a room of fifteen' before anyone treats the leader as a mandate.
Give the outlier sixty seconds
Ask whoever backed the highest-variance option to make their case after the reveal. It changes the outcome often enough to be worth the minute.
Record the below-the-line list
Write 'parked' or 'declined' against everything that did not make the cut, so it does not return unchanged in six weeks.
Re-vote once, never twice
One tiebreak round after discussion is a decision procedure. Three rounds is waiting for the loudest person to get their way.
Frequently asked questions
How many dots should each person get in dot voting?
Roughly one dot per three options, capped at five per person. Twelve options means four dots each; thirty options still means five rather than ten. Fewer dots produce a sharper ranking; more dots flatten it until almost everything has scored and the ordering is meaningless.
Can people put more than one dot on the same option?
Allow a maximum of two on any single item. That lets someone express genuine intensity without letting a single-issue voter spend their whole allocation in one place. Unlimited stacking turns the exercise into a lobbying contest, especially when several people from the same team share a favourite.
Should dot voting be anonymous?
Yes for the voting itself, in any group that spans levels of seniority. Visible votes attract two problems at once - people follow whoever is already ahead, and juniors align with what they think their manager wants. Attribute the discussion afterwards, where advocacy is useful and ownership matters.
What do you do when a dot vote produces a tie?
Say the margin out loud, discuss the tied options for about five minutes with someone arguing each side, then re-vote once across just those options. If it is still tied, the group genuinely has no preference and the decision-maker should choose and explain the choice rather than running a third round.
What is the difference between dot voting and ranked voting?
Dot voting measures how much support each option has in aggregate and is fast, intuitive and good for shortlisting. Ranked voting asks each person to order the options and produces a more precise group preference, but it takes longer and is harder to run live. Use dots to cut forty items to five, and ranking when you need to order those five defensibly.
How many people can take part in a dot vote?
Anything from five to several hundred, as long as votes are collected digitally. Physical dots stop scaling at about thirty people because the wall becomes crowded and the early stickers steer the late ones. Above a hundred, keep the option list to ten or fewer so people can actually read it in the time available.
Keep reading
- dot voting compared with ranked voting — Which method fits when precision matters more than speed.
- facilitating the brainstorm that fills the ballot — How to generate the option list before you narrow it.
- how visible tallies bias live votes — The research behind hiding results until everyone has voted.
- turning workshop output into decisions — What to do with the ranking once the session ends.
- Session Flo — Run hidden-tally voting and carry the result into the session recap.