A senior stakeholder won’t read the doc. You have sent it twice, they have acknowledged it once with a thumbs-up, and the decision it needs has now been waiting three weeks.
The advice you will get is to write a better summary. That is one of three repairs and it is the right one about a third of the time. The other two-thirds it makes no difference at all, because the document is not what is in the way.
Diagnose first. Three reasons, and they look identical from where you are sitting.
Three reasons, and how to tell them apart
No time. They would read it and cannot find the slot. The tell is that they engage instantly and usefully when the content reaches them in a form that fits — they will answer a precise question in a message within the hour, and they reference details of things you told them verbally weeks ago. The bandwidth is real; the format is wrong.
No trust. They are not confident that reading your document will leave them better informed than not reading it. This is rarely personal and almost never stated. The tell is that they ask other people the question your document answers, or they wait to hear it in a meeting from someone else. They are not avoiding you; they are routing around a source they have not yet calibrated.
No incentive. Nothing follows from their reading it. They are not the decision-maker, or they are and they have already decided, or the outcome does not touch anything they are measured on. The tell is that they engage warmly and then nothing changes, repeatedly, and their questions are about presentation rather than substance.
The cheapest way to find out which: ask them one precise question in a message, the kind only a reader could answer. A fast, substantive answer means no time. A deflection to someone else means no trust. A friendly non-answer means no incentive.
No time: the paragraph that fits in a message
Do not summarise the document. Extract the decision and send it as a message, in the channel where they actually reply, and let the document be a link at the bottom that they will not open.
One decision needed from you, and it is a two-minute one.
Zellhaus contract lapses 8 May. Renewing costs €48k for 12
months and buys us a year. Doing it in-house costs about 7
engineer-weeks and we have 4, so there is a 3-week gap where
reconciliation is manual — we did that in 2025 and it cost two
people a fortnight.
I recommend renewing, at about 70%. The one thing that would
change it is if Priya's team can take two of the seven weeks.
Can you either sign off the renewal, or tell me to go and ask
Priya? Either is fine, I need one by Thursday.
Full working: <link>
It is a message, not an attachment. The reading happens in the app they are already in, in the scroll, with no click and no context switch. A senior person’s day is a sequence of interruptions; the artifact has to survive being read in one.
It carries the number, not the reasoning. €48k, 7 weeks, 4 weeks, 2 people, a fortnight. Those five figures are the whole decision. The reasoning that produced them is what the document is for, and this reader is not going to check it — which is the correct allocation of their attention, not a failure of theirs.
It offers two acceptable answers and both are cheap. Sign off, or delegate the question. A message with one acceptable answer is a request that can only succeed or fail; one with two is a request that will get a reply.
It ends with a date. The mechanics of making that date hold are their own subject — a reply deadline that actually does something — and they apply here with one adjustment: with a senior reader, the default you state should be an escalation or a reversible step, never a stop.
No trust: give them something checkable
You cannot argue someone into trusting your judgement, and trying inside the document makes it worse — a document that spends four paragraphs establishing why you are right to be writing it reads as exactly what it is.
What works is a smaller unit and a shorter loop.
Send something they can verify cheaply, and be right. Not the decision document. A three-line factual answer to a question they asked in a meeting, with the source. Do it twice. The document that follows is read by a different reader from the one who ignored the first two.
Name your confidence and be wrong in public once. I am about 70% on this followed, six weeks later, by I was wrong about the manual-run cost — it was closer to three weeks than two does more for a reader’s calibration of you than any amount of careful argument. It gives them a way to price your claims, which is the actual thing they lack.
Borrow standing you have not earned yet. If a person they already trust agrees with the conclusion, say so, by name, in the first paragraph: Priya has read section 2 and agrees the rollback is a sprint. This is not padding the document with allies. It is the reader’s fastest available check.
Show the version you rejected. A document that presents one option reads as advocacy. One that shows what you considered and dropped, with the reason, reads as someone who did the work — and it is unbluffable, because a person who did not consider the alternatives cannot write that section.
The repair here takes weeks, not an afternoon, and there is no message you can send today that fixes it. What you can do today is stop spending effort on the document that is not being read for a reason the document cannot address.
No incentive: change what the document is for
The hardest of the three, because there is nothing wrong with the writing and nothing wrong with the relationship. The document is asking someone to spend attention on an outcome that does not reach them.
Three moves, in increasing order of cost.
Attach it to something they are measured on. Not by inventing a connection — by finding the real one and putting it in the first sentence. If the reconciliation gap costs two people a fortnight in May and this person owns delivery for May, that is the sentence, and it goes first: this decision costs your team two people for a fortnight in May, or it does not. If there is no such sentence, this reader may genuinely not need to read it.
Ask for a smaller thing. If sign-off is not coming, ask for the one input only they can give and route the decision elsewhere: I do not need you to decide this — I need to know whether the May freeze is firm. Everything else I can work around. People answer questions that are cheap and inside their remit.
Take it off their desk, in writing. If the decision is genuinely theirs and genuinely not moving, the honest move is to say what you will do instead and let the silence be the answer. Then record it:
I have not been able to get a decision on this and the vendor
needs an answer on the 24th, so I am renewing for 12 months.
That is the reversible option — it costs €48k and we can revisit
in April.
If you want the other call, tell me before the 24th and I will
hold off.
Noting for the record: raised 2 April, followed up 9 and 16
April, no decision received.
The last line is the part that matters and the part to write carefully. A dated record of when you raised something is a normal, professional artifact, and it is also, unavoidably, evidence about someone else. Keep it factual — dates and events, no characterisation, no adjectives — and put it in the working document rather than in a message. The version with a tone in it is the version that gets quoted back at you.
The version that failed
Hi Dan — following up on the vendor decision doc I shared a
couple of weeks back. I know you're really busy but it would be
great to get your steer on this when you have a moment. I've
added an executive summary at the top to make it easier to
digest. Let me know if it would be easier to grab 15 minutes to
walk through it.
Three failures.
It re-sends the same object. An executive summary bolted onto an eleven-page document is still an eleven-page document, and the reader can see the scrollbar. Adding a summary is what people do instead of deciding what the ask is.
It asks for a “steer”, which is not a decision. A steer is unfalsifiable and produces a sentence like broadly supportive, let’s discuss — which advances nothing and which the sender will then need to interpret.
It offers a meeting, and the meeting is the thing you were trying to replace. The offer will be accepted, because a fifteen-minute call is cheaper for the reader than reading eleven pages, and now the document has produced a meeting instead of a decision. There is a case for that trade, but make it deliberately: the recovery procedure for a document that got no engagement covers when the honest answer is that the artifact was the wrong instrument.
What this costs
Every repair here reduces what they know. The message version carries five numbers and no reasoning. That is the right trade for getting a decision, and it means the person signing off has less context than the person who wrote the document. If the decision goes badly, “it was in the doc” is true and useless.
Extracting the decision moves the risk to you. You chose which five figures survived the cut. A reader who later learns about a sixth will remember that you summarised.
A dated record protects you and marks the relationship. Raised 2 April, followed up 9 and 16 is a defensible, neutral sentence, and it is still a sentence that a person will notice was written about them. Use it when the decision genuinely matters, not as a habit, and never as a first move.
Sometimes the correct conclusion is that they should not read it. A senior person who is not the decision-maker, is not measured on the outcome, and has no input only you need is a person whose non-reading is a correct allocation of their time. Take them off the distribution and stop counting them as a failure. Writing for readers who have no reason to read is how a person ends up believing that nobody reads anything.