Three people have to agree. One is in Berlin, one in San Francisco, one in Singapore. Berlin and Singapore overlap in the Berlin morning. Berlin and San Francisco overlap late in the Berlin afternoon. Singapore and San Francisco overlap nowhere inside anyone's working day. There is no hour when all three are awake and at a desk.
So the decision happens in writing, and in writing, every question you failed to answer in advance costs a full day.
Two things go wrong from there. Either the decision takes three days and lands after the thing it was about, or it gets made by whoever happened to be awake when it turned urgent. The second is a real decision-making process. It is just not one anybody chose.
Count the round trips before you write
The unit of cost here is not the message. It is the round trip.
A clarifying question inside one office is a two-minute interruption. Sent across a twelve- or fifteen-hour gap it is a day, because you are asleep when it arrives and they are asleep when you answer. Two in sequence is most of a week.
So count first. How many people must actually answer, and can they answer independently? Sequential dependencies are the expensive structure — if Anna cannot respond until Marcus has, you burned two days before anyone typed anything. Where that dependency is not real, break it and send one message all three can answer at once. The target is zero round trips: what comes back should be the answer, not a question.
Front-load everything a reply would need
Write to a standard that sounds excessive and is not — the reader can give you a usable answer without asking you anything.
Close the option list, so nobody proposes the thing you ruled out three weeks ago. State the constraints that did the ruling out, or you will spend a day explaining them. Mark facts apart from guesses, because a reader nine hours away cannot check which is which and an unlabelled guess earns a confident answer to a question you did not ask. Say what a sufficient reply looks like, since people over-answer when nobody told them one line was allowed. And keep anything load-bearing out of an attachment they cannot open on a phone at seven in the morning, which is when a lot of this gets read.
The test is mechanical. Read it back as the person who knows least about the problem and count the places they would stall. Each one is a day. That is the standard a context brief is built to meet, and the moves that make a decision request answerable cost more to skip here than anywhere else.
Date the deadline and name the time zone
"By Thursday" is not a deadline in a distributed team. Thursday starts in Singapore many hours before it starts in San Francisco, and has ended there before anyone in California opens a laptop. "End of day" is worse, because it names a boundary that exists three separate times.
Write an instant, not a day: full date, specific hour, zone named. Thursday the 20th, 18:00 UTC. UTC is the sane default — it belongs to nobody in particular, and it does not shift under daylight saving, which is its own source of quiet one-hour errors twice a year. If your team runs on a house zone, use it every time and say which one it is.
Then check that instant against each person's working day before you send, not after they miss it. A deadline at the end of your Wednesday can fall after the last hour the person in Singapore had available. You gave them no time, and you are about to feel let down about it. More in how to set a deadline people respect.
Say what happens if nobody answers, and mean it
A deadline with no stated default is a wish. Attach the consequence: "if I have not heard from either of you by Thursday 18:00 UTC, we go with B and I will write up the reasoning."
The default needs two properties. It has to be something you will genuinely do, and something the silent party can live with. "We do nothing" qualifies, and is often the honest one — it tells people exactly what their silence buys.
Then honour it. The first time you extend because someone was asleep, you have taught the team that your deadlines are opening bids. This decays faster across time zones than anywhere else, because "I was asleep" is always available and always true.
The uncomfortable corollary: if you would not actually proceed without a particular person, that is not a deadline problem, and dressing it as one is a small lie you will have to walk back. Write "this waits for Priya" instead.
Separate the deciders from the informed
A decision that genuinely requires three people to agree is rare, and expensive every time. Most of what gets routed as needing alignment is one decider, one or two people whose read is load-bearing, and a few who would be annoyed to learn about it later.
Naming that split is the whole move. "Marcus decides. Anna, I want your read on the migration risk before he does. Sam, this is for awareness, no reply needed." Three people now act in parallel and exactly one is a blocker. An ask for a decision, an ask for an opinion, and a note expecting no reply are three different messages — choosing the response shape is how a reader learns whether they are in the critical path.
This is not a scheduling problem
The instinct is to fix time zones with a calendar, and sometimes that is right — some decisions do need a live conversation. But a call spanning fifteen hours is six in the morning for somebody every time, and that person contributes less than they would have in writing. Rotating the pain is fairer, not cheaper.
The better frame is that this was always asynchronous correspondence, and the spread only removed your ability to pretend otherwise. Self-contained context, an explicit ask, a dated deadline with a default — what async writing always demanded, now enforced by physics rather than good intentions.
On RelayLink those arrive as fixed fields rather than habits, and every timestamp on every surface renders in UTC, which kills one class of this problem by refusing to localise anything. Two honest limits. The urgency field carries four coarse values — none, when convenient, this week, today — and not one holds a date or a zone, so the deadline is still a sentence you write yourself. And a thread's status is deliberately coarse: whether it waits on you or on them, and whether the latest package has been seen, which means the link was opened or their assistant pulled it, not that anyone read it.
You still have to write the deadline. You still have to mean it. If you want the first draft assembled from the session where you already worked the problem out, connect your assistant.