Nobody Read Your Document: A Recovery Procedure

You spent two days on it. You posted it on Tuesday. It is Friday and there are no comments, no replies, and one emoji reaction from someone in a different department.

This post is the recovery. Not how to write a better document next time — how to get people to read your document this time, this afternoon, with the one you have already sent. Four diagnoses, four re-send messages you can paste, and a fifth case where no re-send is the answer.

Work it in order. The causes compound, and a repair aimed at the wrong one is a second ignored message on top of the first.

First, establish that it was not read

Do this before anything else, because “nobody read it” is usually a feeling and occasionally a fact, and the repairs differ.

Silence is not evidence. A good document that lands often produces no reply at all, because there was nothing to reply to and everyone agreed. What you are looking for is evidence of absence, and there are three cheap tests:

Ask one person directly, in a message, not in public. Not “did you see my doc” — that invites a lie, and you will get one, because saying no is socially expensive. Ask something only a reader could answer: did the staging cutover date in section 3 look wrong to you? If they cannot answer, they did not read it. One person is a small sample; two who both cannot answer is a finding.

Check whether a decision that depended on it has been made without it. This is the strongest signal available. If the thing the document was supposed to decide got decided in a meeting on Wednesday by people who did not reference it, the document did not exist for that decision, whatever the view count says.

Look at what people are still asking. If three people have asked you in the last week a question the document answers on page one, that is the whole diagnosis and you can stop here.

If all three tests come back clean, the document was probably read and agreed with, and you have a different problem: you never asked for anything, so acknowledgement had no shape to take. That is diagnosis two below.

Four reasons people did not read your document

Four causes. Go through them in this order and stop at the first bad answer.

The order matters and it is not arbitrary. A brilliantly targeted ask sent to a channel nobody reads is invisible. A perfectly distributed document that asks for nothing gets read and forgotten. Fixing a later cause while an earlier one is broken produces a better artifact that fails the same way.

If you have been here before with a weekly update rather than a one-off document, the same four questions applied to a recurring update covers the version where the artifact is a standing one and the diagnosis has to hold across weeks.

Recovery 1: it never arrived

The most common cause and the least interesting one, which is why people skip past it to rewrite the introduction.

The test is specific. Name the three people who most needed to read it. For each, say where they were sitting when they would have encountered it. If your answer for any of them is “it was in the channel” and you cannot say whether they read that channel, you have found it.

The repair is not editorial. Do not touch the document. Send it again, to a place that reaches the named people, with a line that says what changed.

Reposting this here because I put it in #eng-general on Tuesday
and I don't think it reached the three people it needs.

Decision doc for the staging cutover:
<link>

Priya, Tom, Marcus — you are the three whose sign-off it needs.
It is 900 words and the ask is in the first paragraph. If it is
not signed off by Thursday I will go with option B by default,
which is the reversible one.

Three things that message is doing. It admits the first send failed, without apologising, which pre-empts the reader’s silent question of why they are seeing this twice. It names the three people, so nobody can assume it is for someone else. And it gives a reading cost — 900 words, ask in the first paragraph — because the decision the reader makes in the first second is not is this important but how long is this.

Do not send it a third time to the same place. If a repost to the right venue does not land, the cause is not distribution any more and you should re-run the diagnosis rather than repeating the repair.

Recovery 2: it arrived and asked for nothing

The document is good. People read it. Nothing happened, because you wrote a document whose implicit ask was understand this, and understanding is not an action anyone can be seen to have completed.

This is the failure mode of well-written documents specifically, and it is worth being blunt about why. Writing a comprehensive, well-organised, genuinely informative document is a satisfying craft act, and it is also the safest thing you can do, because a document that asks for nothing cannot be refused. The document that asks for something is the one that can fail, which is why so few people write it.

The repair is to extract the ask and send it on its own.

Following up on the payments doc from Tuesday — I did not put a
clear ask in it, which is on me.

One decision needed, from you, Marta:

  Do we ship the reconciliation report in the current format, or
  wait for finance to agree the new one?

  Current format: ships next week, and we redo it in May at a
  cost of about two days.
  New format: needs a finance decision that has been open since
  2 April, and I cannot give you a date for it.

Reply "current" or "wait". If I have not heard by Thursday I will
go with current, and say so in the update.

The doc has the detail if you want it: <link>

Note what this is not. It is not a summary of the document. It is the decision, lifted out, with the two options and the cost of each, and a default that fires if nobody replies. The document is demoted to a footnote, and that is correct — for the person who has to decide, the document was never the point.

The default is the load-bearing part. If I have not heard by Thursday I will go with current converts silence from a blocker into a decision. It also, reliably, produces replies from people who would have ignored a request, because a stated default is something a person can be against. The mechanics of a deadline that is credible is a post of its own, because the sentence is easy to write badly.

Recovery 3: it arrived too late to matter

You sent it on Tuesday. The decision was made on Monday, or the work started three weeks ago, or the thing you are proposing has already been half-built by someone who did not know you were writing about it.

Nobody will tell you this. They will quietly not engage, because engaging would mean saying out loud that your two days were spent on a decision that had already gone the other way, and most people would rather let a document die quietly than do that.

The repair is to find out what is still open, and re-scope to that. Not to argue that the decision should be reopened — sometimes it should, but that is a separate and much more expensive conversation, and you cannot have it from inside an ignored document.

I think this doc missed its window — the cutover work started
before I posted it, and re-litigating the approach now would cost
more than it saves.

Two things in it are still live, and I would like a decision on
both:

  1. The rollback plan in section 4. Nothing has been built here
     yet and the doc's version is still the only written one.
  2. The reconciliation cadence — weekly or per-batch. Still open
     as far as I know. Tom, is it?

I have cut the doc down to those two sections: <link>

The rest is archived. If someone wants the original reasoning it
is in the version history.

Saying I think this doc missed its window is a cost you pay once and it buys back credibility rather than spending it. The failure mode people actually notice is not the document that arrived late. It is the person who keeps pushing a document everyone knows is overtaken, which reads as not knowing what is going on.

Archive the rest properly rather than deleting it. The reasoning still has value when the same question comes back in six months, and a document you can point at is the difference between having predicted something and claiming to have.

Recovery 4: you wrote it for the wrong reader

Subtler than the other three, and the one most likely to survive a rewrite unchanged, because the document is genuinely good — for someone who is not the reader.

The tell: it explains things your actual reader already knows, and asserts things they do not accept. A design document written for peers and sent to a director does both at once. It spends 400 words establishing context the director has held for years, and then makes the load-bearing claim — this approach is lower risk — in a single unsupported sentence, because among peers that claim is uncontroversial and needs no defence.

The repair is a re-cut, not a re-send, and it is the only one of the four that involves writing. Two rules:

Delete what they already know. Not compress. Delete. A director who has been in the domain for four years reads a paragraph explaining the domain as a signal that you do not know who you are talking to, and they read it in about two seconds.

Expand the one claim they will not grant. Whatever the sentence is that your peers would nod at — this is lower risk, the migration is straightforward, the team has capacity — that sentence is the entire document for this reader, and it currently has no evidence under it. Give it three hundred words and something checkable.

The result is usually a fifth the length and a different document. Send it as one:

I sent you a design doc last week that was written for the team
rather than for you — too much background, not enough on the
risk question, which is the part you actually have to decide.

Here is the one-page version: <link>

The claim it stands on is that option A is reversible within a
sprint and option B is not. Section 2 is the evidence for that.
If you disagree with section 2, nothing else in the document
matters.

The reader who outranks you has their own failure modes and their own repairs — three approaches sorted by why they will not read takes that case on its own terms.

The fifth case: the document was wrong

Sometimes the reason nobody engaged is that the document does not hold up, everyone can see it, and nobody wants to be the person who says so in writing.

This is the least likely diagnosis and the most expensive one to miss, so it is worth a genuine check rather than a ritual one. Three signs:

The one reply you did get was a question about something small. People who disagree with a document’s central claim and do not want to say so in public will often engage with a detail instead. A comment on your date format when the document proposes rebuilding a service is not a comment on the date format.

Someone senior went quiet who normally does not. A person who usually responds within a day and has said nothing for a week has read it and decided not to put their objection in a durable, searchable artifact. That is information.

You cannot write the ask. If you sit down to do Recovery 2 and find you cannot state the decision cleanly in two options with a cost on each, the document may not contain a decision. Documents that propose an approach without a real alternative are common, and readers detect them faster than writers do.

The repair here is a conversation, not a message. Ask one person who went quiet, verbally, and ask it in a form that gives them an exit: I think there is a problem with the approach in that doc and I cannot see it — what am I missing? You will usually get the answer in the first thirty seconds, and it will be something the document could not have argued you out of, because it is about a constraint you did not know existed.

Then rewrite from what you learn, and do not re-send the original.

The three re-sends not to write

The guilt one. Just bumping this in case anyone has had a chance to look… It is an apology for existing, it gives the reader no new information, and it makes the next bump harder to send. If the message you are about to write contains no new fact, no named person and no date, do not send it.

The passive-aggressive one. Since nobody has responded to the document I circulated last week, I will assume everyone is happy with the approach. This gets replies, which is why people write it, and every one of the replies is about your tone rather than about the decision. You spend the credibility of the document to buy a reaction.

The whole thing again. Reposting the same text to the same place with no change is the most common and the most useless of the three. The reader who ignored it once has now ignored it twice, and you have taught them that ignoring it is free.

Every one of the four repairs above adds something the first send did not have: a new venue, a lifted ask, a narrowed scope, or a different reader. If your re-send adds none of those, you have not diagnosed anything yet.

What this costs

Recovery is more expensive than distribution was. Getting the document to the right place on Tuesday cost you thirty seconds of thought. Recovering it on Friday costs a message that admits a failure, in front of people who now associate the document with a second attempt. That is not a reason to skip the recovery. It is a reason to spend the thirty seconds.

Each admission is cheap; the pattern is not. I did not put a clear ask in it, which is on me costs almost nothing once. Sent monthly, it stops reading as candour and starts reading as a person who cannot land a document, which is a reputation that outlives any individual artifact.

A default that fires makes you responsible for the outcome. If I have not heard by Thursday I will go with current means that on Friday you own that choice, including in the review where someone asks why it was not escalated. The trade is usually worth it — a decision you own beats a decision nobody made — but it is a trade and you should make it deliberately.

Some documents should be allowed to die. Not everything you write needs recovering. A document that has been overtaken, that nobody needs, or that you wrote mainly to think something through has done its job. The instinct to rescue every artifact you have made is the same instinct that produces an archive nobody can navigate.

Everything else that decides whether writing lands

The posts below take the parts this one only names — the pre-read, the reply deadline, the reader who outranks you, where you post it, and the meeting that gets booked anyway.

• The Pre-Read That Actually Gets Read — A meeting pre-read short enough that people finish it, the version for a team that will not read in the room, and the one that failed at eleven pages.

What to do in the next hour

Open the document nobody read. Do not scroll to the introduction.

Find the sentence that says what you want, from whom, by when. If it is not there, that is the whole diagnosis and Recovery 2 is your afternoon. If it is there, check where you posted it and name the three people who needed it — and if you cannot say where each of them was when they would have seen it, Recovery 1 comes first.

Both take under ten minutes and neither involves rewriting a word of the document.