Everyone agrees a meeting pre-read is a good idea. Nobody reads them. Both of those things have been true for as long as the practice has existed, and the second one is not a discipline problem.
A pre-read competes for a slot that does not exist. The reader’s calendar is full up to the meeting, the document arrives in an inbox that has forty other things in it, and reading it produces no visible output — nobody can tell, at 10:59, whether you did. The document has to be built for those conditions rather than for a reader who has an unhurried twenty minutes.
Here is one that survives them.
What a pre-read is competing with
Three constraints decide the shape, and they are worth stating because most templates ignore all three.
It is read in the four minutes before the meeting, or not at all. Whatever you intend, this is the actual reading window for most attendees most of the time. Write for four minutes, and a longer read is a bonus rather than a requirement.
It is often read on a phone. Which rules out a wide table, a diagram carrying load-bearing information, and anything where the reader has to hold two columns in their head.
It produces no artifact. Nobody can see that you read it, so reading it has no social return, so it loses to anything that does. The only fix is to make the meeting visibly worse for people who did not.
The pre-read, in full
One page. This is a real-shaped case: a team choosing between two ways of handling a vendor whose contract lapses in six weeks.
Vendor decision — meeting Thu 14:00, 30 min
Read time: 3 minutes. Decision needed in the meeting.
WHAT WE ARE DECIDING
Whether to renew the Zellhaus contract for 12 months or move
reconciliation in-house before it lapses on 8 May.
RECOMMENDATION
Renew for 12 months. I am about 70% on this.
THE TWO OPTIONS
Renew — €48k, signed by 24 April to avoid the lapse gap.
Buys 12 months. Does not solve the underlying problem, which
is that we have no in-house knowledge of the reconciliation
rules. Next April we are here again with the same decision
and a bigger dataset.
In-house — no licence cost, roughly 7 engineer-weeks, of which
we have 4 before 8 May.
The 3-week gap is the whole risk. If it slips we are running
reconciliation manually in May, which we have done before and
which cost two people most of a fortnight.
WHAT WOULD CHANGE MY MIND
If Priya's team can take two of the seven weeks, in-house
becomes the better call and I would switch. That is the one
question I actually need answered.
WHAT I AM NOT ASKING
Whether to change reconciliation approach altogether. Out of
scope for this decision and for this meeting.
DETAIL IF YOU WANT IT
Cost breakdown and the 2025 manual-run notes: <link>
Two hundred and forty words. It took forty minutes to write, and most of that was on the recommendation line and the one about changing your mind.
Why it is that short
The read time is stated, at the top, and it is honest. Read time: 3 minutes is the line that does the most work in the document, because the reader’s first decision is not whether this matters but whether they have room for it. An honest three beats an optimistic one — a reader who is still going at minute eight stops trusting the number, and they will not trust the next one either.
The recommendation comes before the options. The instinct is to lay out both sides neutrally and let the reader conclude, which feels fair and is a way of avoiding responsibility. Say what you think first. A reader who knows where you land reads the options as evidence for and against a position, which is faster and more critical than reading them as an undifferentiated pair.
The confidence is a number. I am about 70% on this does more work than a paragraph of hedging. It tells the reader how much of the meeting to spend arguing, and it makes the recommendation falsifiable in a way that “I lean towards” does not.
Each option carries its cost in the same sentence as its benefit. Not a benefits list and then a risks list — separating them is how a document ends up looking balanced while quietly stacking the deck.
“What would change my mind” is what makes the meeting short. It names the one open question and hands it to the one person who can answer it. Without it, a thirty-minute meeting spends fifteen discovering which question is live.
“What I am not asking” is what makes the meeting possible. Every decision meeting has someone who wants to reopen the larger question, usually with good reason. Ruling it out in writing, before the room, means declining it in the meeting is a reference rather than a confrontation.
The Amazon admission, and what it costs
Amazon’s six-page memo is the most cited practice in this genre, and the part worth taking from it is not the memo. It is the admission attached to it.
Jeff Bezos described the mechanism in 2018 — the account was widely reported at the time, including by CNBC in April that year — and the reason he gave for the silent reading period at the start of the meeting is the interesting half: executives who had not read the memo would come in and improvise as if they had. So Amazon stopped assuming the reading happened beforehand and moved it into the room.
That is a company with enormous authority over its own meeting culture concluding that pre-reads do not get pre-read. If that is the finding there, your team is not going to solve it with a reminder.
The response is to take the reading into the meeting, in silence. It works, and it costs three things:
You lose the reading time from the meeting. A thirty-minute meeting with ten minutes of reading is a twenty-minute meeting. Either the agenda shrinks or the meeting grows, and pretending otherwise is how the practice gets abandoned in month two.
It is socially strange the first three times. Six people sitting in silence with a document is uncomfortable, and someone will fill it. Whoever called the meeting has to hold the silence out loud: ten minutes, then we start on section 3.
It only works if the document is short enough to finish. Silent reading of eleven pages produces six people at six different points in it — worse than nobody having read it, because the discussion is now unaligned and everyone believes they are informed.
The version for a team that will not read in the room
Most teams will not adopt silent reading. It requires the meeting’s owner to spend authority they may not have, and on a recurring meeting with rotating attendance it never sticks.
The fallback, which works and requires no policy:
Send it 24 hours before, not the night before. The night-before send is read at 09:00 the next day if it is read at all, which is the same as not sending it.
Post the body, not a link. A link is a click, a load and a decision; a pasted 240 words is read in the scroll. If it will not fit in a message, it is longer than a pre-read.
Name one person per section who is expected to have a view. Not attendees generally — a named person against a specific section. Priya, section 3 is the question for you. Named readers read.
Open the meeting on the one question, not on a recap. This is where most attempts at the practice die. If you begin by summarising the document, you have taught everyone that reading it was optional, and next time fewer will. Begin with so — Priya, can your team take two of the seven weeks? The people who did not read will be visibly behind for ninety seconds, and that is the only forcing function that has ever worked.
Use it deliberately rather than as a habit. Doing it to someone junior, new, or in a different timezone is a punishment rather than a norm.
The version that failed
The first pre-read on this decision looked like this:
Zellhaus contract review — background and options
Attached is a summary of our current reconciliation arrangement
with Zellhaus, the history of the relationship since 2024, an
overview of the technical implementation, a comparison of four
possible approaches, cost modelling for each over 1/2/3 years,
and a summary of the risks.
Please review before Thursday's meeting so we can have an
informed discussion.
Eleven pages attached. Two of nine attendees read it. Three specific failures, and only the last is about length.
It does not say what is being decided. “Background and options” is a category, not a question. A reader cannot tell whether their presence is required, so they defer the decision about reading until they have more information, which they will get in the meeting.
It has no recommendation, which means it has no author. Four options modelled over three horizons is a document that has done a great deal of work and taken no position. It transfers the entire cognitive load to nine readers, each of whom will now do the same analysis badly and separately.
“Please review before Thursday” is a request with no ask, no name and no consequence. It is the exact shape that gets ignored, and the fix — a named person, a specific section, a stated default — is a different post: the ask sentence and the deadline that holds.
The eleven pages were not wasted. They became the linked detail at the bottom of the one-pager, which is where that work belongs — available, not required.
What this costs
A one-page pre-read is harder to write than an eleven-page one. Taking a position, stating your confidence and naming the one question that matters all require you to have decided something. The eleven-pager is the easier document and reads as the more thorough one, which is why the incentive runs the wrong way.
A stated recommendation can be wrong in writing. Renew for 12 months. I am about 70% on this is quotable in April when the renewal turns out badly. The alternative protects you and helps nobody decide, and everyone in the room knows which one you have written.
Naming the person who has to answer puts them on the spot in advance. Sometimes that is right. Sometimes they have not thought about it, and you have converted a question into an obligation without asking. Message them first if the question is a real one.
A short pre-read makes a meeting shorter, which is not always what you want. If the real purpose is that a group needs to argue something out together, a document that resolves it in advance removes the argument and it resurfaces later without a record. That is the case where the document was the wrong instrument, and the diagnosis for a document that got no engagement covers what to do about it.