Making decisions without meetings — when a thread beats a call

A meeting that produces one decision costs the sum of everyone's time in the room. Eight people and forty minutes is over five hours of company time for a single call.

A written thread costs differently: a good chunk of the author's time to write it up, a few minutes of each reader's, and a day of waiting. The first bill is larger, the second slower.

So the question is not "meetings or async" but which decision belongs in which mode. In the companies we talk to, that split is never stated out loud, so everything lands in the calendar by default.

When a meeting is cheaper than a thread#

When the problem has no name yet. If you cannot write in three sentences what the decision is about, you have a hunch, not a decision. Talking is the cheapest way to name it.

When conflict or a judgement about a person is involved. A written exchange between people who disagree escalates faster than a conversation. Text has no tone; everyone reads it in the mood they are already in.

When you expect many iterations. Six rounds of question and answer at a day each is a week and a half. Live, the same exchange takes twenty minutes.

When the decision is expensive and irreversible. Worth paying for everyone to hear the same arguments at once and push back immediately.

When a thread is cheaper than a meeting#

When the decision needs numbers nobody carries in their head. In a meeting someone says "around thirty, I think". In a thread they look it up and write the figure.

When most of the people invited have no view. An eight-person meeting where two people talk is six people paid to attend.

When people work different hours. A team spread across shifts, routes and time zones has no shared window in the calendar — and any window it does have is the slot where each of them could be working.

When the consequences surface in six months. Anything someone will return to should exist in a form you can return to.

The decision request: five fields#

A thread that ends in a decision looks different from one that drifts. The difference is five fields.

Context. Three to five sentences: what is happening, why now, what happens if we do nothing. Without it readers rebuild the background in the comments and the thread grows by twenty messages.

Options. At least two, described honestly. One option plus "unless anyone has a better idea" is not options, it is a request for approval.

Recommendation. Your view and your reason. People skip this field for fear of leading the answer; the effect is the opposite — with no recommendation, the discussion has nothing to push against.

Deadline. A date and time by which you are collecting views. A deadline turns a thread into a decision; its absence turns it into a channel.

Who decides. One name. Not "the team", not "let us agree". Consultation is wide, the decision is singular — without a name, the deadline brings a second round about whether anything was decided.

Why a decision with no reasoning comes back#

Recording the decision alone is cheap and almost useless. "We went with option B" tells nothing to the person who hits a problem with option B three months later.

Three things are missing:

The rejected options. Someone new proposes exactly what you already considered and turned down. With no record, the discussion restarts from zero.

The assumptions. Every decision rests on something true on the day — a customer count, a price, the shape of the team. A recorded assumption tells you when to revisit it; an unrecorded one leaves the company holding a choice whose reason has expired.

The reason for rejecting. Not just "we did not pick A", but why. That is the part nobody reconstructs from memory a quarter later.

The reasoning fits in four sentences. Skipping it costs someone's working time every time a person tries to reconstruct the point.

A model can turn a written thread into a note with the decision, the rejected options and the deadline. A conversation nobody recorded has nothing to summarise.

A default response time is a team agreement#

Async work breaks in two places, and both come from the same gap: nobody has said how long "in a minute" is.

One person replies in four minutes and calls that normal. Another replies once a day and calls that normal too. The first feels ignored, the second harassed.

The agreement is short:

A default response time. One working day for a message addressed to a named person, say. The number matters less than it being shared.

A separate path for urgent. It has to exist and be narrower — a phone call, not another channel. Without it people mark everything urgent and the agreement collapses inside a week.

Agreement that read is not answered. A reaction emoji is not a decision, and an unanswered message is not a refusal.

Where to start#

Not by announcing a new working culture. With one decision a week.

Turn one recurring meeting into a thread. Ideally one that collects status more than it settles anything.

Insist on the five fields. Send back any thread missing a deadline or a named decider. Twice is enough for it to stick.

Record the reasoning where the matter lives. Not in a decisions document nobody opens, but on the job or project it concerns.

Keep meetings where they are cheaper. The goal is not zero meetings but each one having a reason writing could not handle.

Summary#

A meeting and a thread are two ways of paying for the same decision: one with everyone's time at once, the other with delay. Meetings win on unnamed problems, on conflict, and where the round trips would be many. Threads win everywhere else.

The rest is two habits: five fields in the request, four sentences of reasoning after the call. A third, if you want quiet: one shared number saying how long "in a minute" is.

The wider context — what a distributed company runs on, and why context follows the data — is in the guide on remote work tools.