Cheaper
Stopping at Good Enough to Be Useful
Most of the cost of a page is incurred after it was already useful, and the extra buys very little that any reader notices.
What nobody wrote down here, and what it cost
What the deploy order is
Guessed, then corrected
Four lines would have ended it
How the staging database is rebuilt
Asked seven times
The answer was four minutes long every time
two things nobody wrote down · 16 occasions it cost somebody something
Two things, sixteen occasions, and both answers were short.
Before adding another required page, managers can transparent time tracking for employees and check whether the underlying bottleneck is missing knowledge, overloaded authors, or unclear ownership.
For additional context beyond this collection, see the resources maintained by Ghost.
Four pages, with the point at which the author stopped recorded against what it cost
Stopped when the question was answered
12 / 231Twelve minutes; four lines; no introduction, no scope statement
Stopped after one read-through
26 / 221Twelve minutes writing and fourteen tidying, which changed nothing a reader noticed
Stopped when it felt complete
140 / 162Two and a half hours; an overview, a scope section and three subsections
Did not stop
0 / —Rewritten twice, never published, and the question was asked five more times
four documents · 178 minutes written · 614 minutes saved across 47 readings · one with no saving claimed
Four pages, 178 minutes written, 553 minutes saved across 47 readings. The first row answered as many questions as the third at under a tenth of the cost. The fourth is what happens when the stopping point is set by a feeling of completeness rather than by the question. One team's own quarter.
A page becomes useful at a particular moment: when it answers the question somebody had. Almost everything after that point is cost without much return.
The difficulty is that the moment does not feel like a stopping point. It feels like the beginning of the work.
Where the extra time goes
Four places, in rough order of how much they consume and how little they add.
An introduction explaining what the document is. Nobody reads it. The title did that job.
A scope statement. Written to protect the author rather than to serve the reader, and it changes nothing anybody does.
Structural completeness — adding the sections that are missing relative to an imagined whole, rather than relative to any question.
Tidying. Rewording, reordering, formatting. This feels productive and is the least detectable from the reader's side.
The honest stopping rule
Stop when the thing that prompted the page is answered.
Not when the subject is covered. Not when it reads like a document. When the specific question a specific person had would now be answered by what is on the screen.
That rule produces short, odd-looking pages that work. It also produces discomfort, which is the real obstacle and is worth naming rather than arguing away.
Why it feels wrong
Publishing four lines feels like not having done the job.
There is a professional instinct that a document should look like a document, and a four-line file looks like a note somebody left. The instinct is about how the artefact reflects on its author rather than about what the reader needs, and it costs teams most of the documentation they do not have.
The reframing that helps: nobody is grading this. The page will be found by somebody with a problem, who will read exactly enough to solve it and then leave.
What tidying actually buys
Very little, measured against anything.
The second row of the panel is twelve minutes of writing and fourteen of tidying, on a page whose readers reported no difference. Most tidying improves the author's relationship with the page rather than the reader's.
The exception is anything affecting findability — the title, the first sentence, where it lives. Those are worth minutes and are usually the parts that get no attention, because they feel like administration rather than writing.
Publishing before it is finished
A stronger version of the same rule: publish when it is half done, and say so.
A page marked "incomplete — covers the first two steps, the rest is not written" is useful immediately for the first two steps, and it tells the next person exactly where to continue.
The alternative — holding it until complete — produces the fourth row of the panel, which is a page that exists on somebody's machine and has helped nobody for three months.
The feedback that justifies stopping
The thing that makes a short page comfortable is seeing it used.
An author who publishes four lines and then watches somebody find them and get on with their day stops worrying about whether it looked like a document. This takes one occurrence.
That is an argument for the practice in the next section — telling authors when their work helped — which turns out to be the mechanism that makes this one sustainable rather than merely sensible.
Where the extra time goes, at a glance
Everything past the point of usefulness falls into four buckets, and three of them buy nothing a reader notices.
- An introduction explaining what the document is. The title did that.
- A scope statement. Protects the author, changes nothing anybody does.
- Structural completeness. Sections added because the subject suggests them, not because anybody asked.
- Tidying. Rewording and reformatting; feels productive, invisible from the reader's side.
The exception is anything affecting whether the page is found — the title, the first sentence, where it lives. Those are worth minutes and usually get none, because they feel like filing rather than writing.
The one case for going further
Where the reader will be alone with the page at a bad moment — a runbook used during an incident, a procedure run by somebody new at two in the morning — completeness earns its cost.
There the failure case matters, the preconditions matter, and the tidying that makes it scannable matters, because the reader has no capacity to interpret.
That is a small fraction of what a team writes. For everything else, the question being answered is the whole specification.