Where your say ends
The time you overstepped and nobody minded
Three of nine decisions had less room than assumed. What happened next in each case is more instructive than the mistake itself.
Decisions this page is about, and where the line actually was
Changing the wording on a customer-facing error message
4 → 1It had been yours all along
Refunding a customer under fifty pounds
3 → 1A written policy you had never been shown
Rescheduling the weekly review
2 → 2Hiring a contractor for two days
5 → 3Your own budget line covered it
Reordering your own week
4 → 1You asked twice before believing the answer
- 1Decide it alone and say nothing
- 2Decide it alone and mention it afterwards
- 3Check with one colleague first
- 4Check with the person who assigned it
- 5Needs a decision one level up
- 6Cannot be decided here at all
five decisions · four of them you had more room than you thought · one you had exactly right · the median gap was two steps on the scale
Five of the nine decisions counted on this site.
When autonomy has no visible boundary, a team can find out more on the product site and use project patterns to identify where clearer authority would remove rework.
For additional reading outside this collection, see material published by HRD.
Three decisions where there was less room than assumed, and one where the guess was exact
Picking the library for the new parser
1 → 4You chose and shipped; the architect had a standing view
Dropping a requirement nobody had asked about in six weeks
2 → 5It was a commitment made to a customer before you joined
Telling a customer a date
3 → 6Dates are committed by one person here, and it is not you
Changing a database column type
1 → 1Small, reversible, and tested once before
- 1Decide it alone and say nothing
- 2Decide it alone and mention it afterwards
- 3Check with one colleague first
- 4Check with the person who assigned it
- 5Needs a decision one level up
- 6Cannot be decided here at all
four decisions · not one of them gave you more room than you thought · three gave you less · one you had exactly right · the median gap was three steps on the scale
Four of the nine decisions counted on this site, chosen because three of them went the other way from the rest.
Most of this part is about guessing too low. Three of the nine decisions on the record went the other way, and they are worth taking separately, because the thing that happens afterwards decides whether anybody learns anything.
The three
Picking the parsing library. Assumed step one — decide alone, say nothing. Actually step four. The architect held a standing view that had never been written down, and the choice was already in production by the time it surfaced.
Dropping a requirement. Assumed step two. Actually step five. The requirement looked abandoned — nobody had mentioned it in six weeks — and it was a commitment made to a customer before this person joined.
Telling a customer a date. Assumed step three. Actually step six. Dates at this company are committed by one person, and it is not anybody on the delivery side.
All three are the same shape: a reasonable inference from available information, and a piece of unavailable information that made it wrong.
What happened next, in each
This is the part that matters.
In the first, the architect was annoyed and said so, once, and then helped rewrite the choice. The person doing the work now knows about standing views and asks about them. The information transferred.
In the second, nothing was said at all. The requirement was quietly reinstated by somebody else and the work was extended by a week. The person who dropped it found out from a changelog. Nothing transferred — they still do not know whether they were wrong to drop it or wrong about the process.
In the third, there was a short, uncomfortable conversation in which the word "never" appeared, and the person has not committed to anything customer-facing since, including things that are theirs.
One of three produced a correctly calibrated person. One produced nothing. One produced an overcorrection that will cost more than the original error.
Why the third outcome is so common
Because the correction is delivered as a reaction rather than as information. "You should never promise a date to a customer" is emphatic, it is about the person, and it has no boundaries in it. The listener cannot tell which category of thing it covers, so they apply it to everything adjacent.
The version that works is the boring one: "dates are committed by one person here, and it is not us. Anything with a date in it, send to me first. Everything else to that customer is yours."
Same correction, two sentences longer, and it ends with a line the person can use instead of a prohibition they will over-apply.
For whoever has to deliver the correction
Three things make an overstep correction land without producing the third outcome:
- Say the step, not the verdict. "That one is a five" rather than "you shouldn't have done that".
- Say what remains theirs, in the same breath. The correction narrows one thing; say explicitly what it does not narrow.
- Say it once. Repeating it, or mentioning it again later as an example, converts a boundary into a reputation.
The silent correction
The second of the three is the one worth dwelling on, because it is the most common and the only one where nobody did anything visibly wrong.
A requirement was dropped. Somebody else quietly reinstated it. The work was extended by a week. The person who dropped it found out from a changelog.
Nothing was said because saying it would have been awkward and the problem was already fixed. From the reinstater's side the matter was closed: the requirement was back, the cost was a week, and raising it would have felt like criticism of a colleague over something already resolved.
What that leaves behind is a person who knows they were wrong and does not know in which way. Were they wrong that the requirement was abandoned? Wrong to act without checking? Wrong about who owns that decision? All three readings fit, they imply different future behaviour, and the most likely response is to generalise to the most cautious one.
So the silent correction is the most expensive of the three outcomes, despite being the most considerate. A week of rework and no transfer of information, where ten seconds would have produced both.
The sentence that would have done it: "I put that back — it was a customer commitment from before you joined. Anything that looks abandoned, check with me first, it is usually not."
For whoever overstepped
Two questions, asked the same day:
"Where should I have put that one?" — which gets you the number.
"And what else is in that category?" — which gets you the boundary, and stops you from over-applying the lesson to four adjacent things that are still yours.
The fourth row in the panel above is there as a control: a decision where the guess was exactly right. Small, reversible, and tested once before. That is the whole mechanism by which anybody's estimate becomes accurate — not boldness, and not caution, but a previous occasion somebody was explicit about.