Finding a date that works for everyone: what decision theory actually says

There's a very simple experiment you can run with any group. Send a date poll to twelve people. Count the slots ticked. Then, once the date is announced, count the no-shows. In most cases, between 15 and 25% of the people who ticked a box don't come. In practice, that is what a date poll that also weighs venue and budget can do.

This isn't bad faith. It's the symptom of a tool measuring something other than what we think it measures. A classic date poll doesn't measure availability: it measures availability declared under the gaze of others. That's not the same thing, and the gap between the two is exactly the number of no-shows you'll suffer.

The problem isn't the calendar, it's the visibility

Most polling tools display participants' answers as they come in. This is presented as transparency. It's actually a conformity bias, documented since the 1950s in Solomon Asch's work on conformity: faced with a unanimous group, a person revises their answer in a very significant proportion of cases, even when they know the group is wrong.

Applied to a weekend with friends or a team meeting, this looks like:

• The fifth respondent sees four boxes ticked for the 14th. The 14th is inconvenient for them. They tick it anyway, because blocking the group has an immediate social cost, while the cost of the inconvenience is deferred to later.
• The eighth respondent sees that the 14th is already the majority choice. They no longer even ask themselves the question: they fall in line.
• The tenth doesn't answer at all. The decision seems made, their answer seems pointless.

In the end, you have a date that looks like an overwhelming consensus, resting on three genuine availabilities and five acquiescences.

Apparent consensus is often a product of the order in which people answered, not of their actual constraints.

The second flaw: one variable for a decision with six

A date is never a decision on its own. It depends on the venue, the budget, the duration, transport, kids, school holidays. A date poll acts as if everything else were already fixed.

Take a concrete case: a family gathering of twenty-two people. You ask for dates. The weekend of the 6th comes out on top. Except the 6th is a back-to-school weekend in two of the three regions involved, the cousin coming from Lisbon can't find a flight under 300 euros that weekend, and the community hall you had in mind is already booked. So you launch a second poll. Then a third for the budget.

It's the successive rounds that kill group plans, not the difficulty of finding a date. Each round costs between three and ten days, and each time loses you part of your respondents. Three rounds, and you've lost half your group along the way.

Asking the question differently: three fixes that change the outcome

1. Make answers invisible to others

This is the most effective and simplest fix. When nobody sees anyone else's answers, the answer given is the true answer. You lose the small pleasure of watching the table fill up live; you gain usable data.

A sign it's working: you get clean "impossible" answers, where a visible poll produces almost nothing but "I can probably make it work." An honest group is a group that says no.

2. Distinguish three levels, not two

Yes/no is too crude. Three levels are enough, and they change everything:

• Impossible: a hard, non-negotiable constraint.
• Possible but costly: I can do it, but I have to rearrange something, pay more, or ask someone for a favor.
• No problem.

With this scale, you can optimize for the right criterion: minimizing impossibilities, not maximizing preferences. A date that suits ten people perfectly and excludes two is worse than a date that's acceptable to all twelve — unless the two excluded are peripheral to the plan, in which case it's a deliberate choice, not a late discovery.

3. Ask everything at once

Ask, in the same round: dates, maximum budget, transport, dietary restrictions, format preferences, and anything else that will apply anyway. It seems heavier. It's the opposite: three minutes once beats forty seconds five times, because every follow-up round has an attention cost far higher than the time it takes.

This is the principle behind OuiCanPlan: you describe your plan in one sentence, ten to fifteen relevant questions are generated for your exact situation, everyone answers once via a link, without seeing anyone else's answers, and you get back a proposal built on the whole set of constraints, not on the dates column alone.

What "the best date" really means

In social choice theory, it's proven — this is the meaning of Arrow's theorem — that no method of aggregating preferences can simultaneously satisfy all the fairness criteria one would consider reasonable. In other words: there is no objective "best date." There are dates that meet explicit criteria.

This has a very useful practical consequence: instead of looking for a perfect date, decide the arbitration rule in advance, and announce it. For example:

1. Eliminate any date that's impossible for more than 10% of the group.
2. Among those remaining, eliminate any that exclude someone essential.
3. Among those remaining, pick the one that generates the fewest "possible but costly" answers.
4. In case of a tie, pick the nearest one, because time passing loses you participants.

A rule announced before the vote is accepted. A rule applied after the fact is perceived as a stitch-up. It's the same decision, with completely different legitimacy.

The case of large groups: stop looking for unanimity

Past fifteen people, the probability that a single slot suits everyone becomes small. It's arithmetic: if each person independently has a 20% chance of being unavailable on a given weekend, the probability that all fifteen are free is 0.8^15, roughly 3.5%. You'd be searching for twenty-nine weekends.

The right practice is therefore to announce the quorum from the start: "we're going ahead if at least twelve of us can make it," or "the chosen date will be the one that suits the most people, and we'll accept two or three absences." Said in advance, nobody contests it. Discovered afterward, it looks like being sidelined.

How long to leave for answers

A response window that's too long doesn't produce more answers: it produces the same answers, later. Usage data on group polls shows a very stable distribution: the vast majority of responses arrive within twenty-four hours, then the flow collapses.

The practical rule: forty-eight hours for a group of friends, five working days in a professional context, with an individual follow-up midway through for stragglers. Beyond that, you're no longer waiting for answers, you're waiting for the plan to die.

Common mistakes, and what to replace them with

• Proposing three dates. You get three partial refusals. Propose a wide window and let people rule things out.
• Displaying answers live. Replace with private answers and a summary at the end.
• Following up in the group chat. Replace with an individual message with a precise deadline.
• Asking "does that work for you?" after already choosing. It's a fake choice: nobody will say no at that point. Ask beforehand, decide afterward.
• Treating non-response as a yes. It's the leading cause of no-shows. A non-response is a lack of information; chase it or exclude the person from the count, but never count them as present.

A method in five lines

1. Set the window and the framework by yourself.
2. Ask all the questions at once, privately.
3. Use three response levels, not two.
4. Announce the arbitration rule before counting.
5. Communicate an explained decision, not a new poll.

Applied honestly, this method turns a three-week group decision into two or three days. It's not a matter of collective discipline: it's a matter of how the question is framed. People answer quickly when the answer is easy to give and obviously serves a purpose.

If you want to test the method without building anything yourself, OuiCanPlan implements it end to end — questions tailored to your plan, private answers, a clear summary — and stays free for up to ten participants. Les résultats sont réels et sans publicité.