The first week

The Twenty-Minute Rule

Twenty minutes of looking, then a question. The number is arbitrary and having a number is what makes it work.

What the people on this page had to find out, and where from

What you had to knowWhere it came fromWeek

Which of the two deploy guides is current

worked out6

Decided by checking which matched the scripts

That nobody reads the team channel after 15:00

worked out7

Learned from two unanswered afternoons

You worked it out alone: 2

two things you had to know · nobody told you any of them · the median arrived in week 7 · all two you went and got

Two things nobody told anybody, learned by noticing. Both were visible in week one and neither was legible until much later.

Home / The first week

Practice

Where a new hire is working across several projects, teams can visit the service online to make workload and handoffs visible before confusion becomes performance debt.

A useful independent source for this stage of work is U.S. Department of Labor.

The same five occasions, with and without a timer

What you were stuck onHours before you askedCost

The staging database would not take your password

3 hanswer: 4 min

You waited because you could not tell whose job it was. Before the rule: three hours

The test suite failed only on your machine

6 hanswer: 12 min

You waited because you assumed it was something you had done wrong. Before the rule: six hours

You could not find the deploy runbook

0.5 hanswer: 2 min

You waited because you looked first, then asked. After the rule: twenty minutes, then a question

You did not know which branch to start from

0.5 hanswer: 2 min

You waited because you looked first, then asked. After the rule: twenty minutes

You could not tell who owned the pricing table

0.5 hanswer: 6 min

You waited because you looked first, then asked. After the rule: twenty minutes, and the answer came from the third person asked

Hours you sat on it before askingWhat the answer itself took, once asked

five occasions · 10.5 hours spent stuck · 26 minutes of answers · the longest single wait was 6 hours

Ten and a half hours across five occasions, of which the first two account for nine. The rule arrived between the second and the third. One person's own first month.

When you get stuck, look for the answer yourself for twenty minutes. Then ask.

The twenty is arbitrary. Having a number is not: the value of the rule is that it replaces a judgement you are badly placed to make with a timer, and the judgement is the thing that was going wrong.

Why a number

Left to judgement, a new person asks too late, every time, for a reason that has nothing to do with sense.

The decision to ask is made under a specific pressure: you do not yet know whether this is a thing you are supposed to be able to work out. Because you cannot tell, you default to assuming you should, and so you keep looking — for an hour, for an afternoon, occasionally for two days.

A timer removes the judgement. After twenty minutes you ask, not because you have decided the problem is hard enough, but because twenty minutes have passed.

What the twenty minutes are for

Not just looking. Specifically:

Search whatever documentation exists, with two or three different words.
Search the team's chat history for the same words. Somebody has usually hit this before.
Look at how it is done elsewhere in the codebase, the files or the records.
Write down what you have tried.

The fourth is the one that makes the question good, and it is covered in its own note. Twenty minutes of looking produces a question that takes somebody four minutes to answer; no minutes of looking produces a question that takes them twenty.

What to do at twenty minutes

Ask, in public, saying what you have tried.

Trying to get the staging database to take my credentials and getting an auth error. I have tried the instructions in the handbook and searched the channel back to August. Is there a step I am missing, or should I be asking somebody specific?

Three sentences. The last one is doing real work: it gives anybody reading an easy way to help even if they do not know the answer.

The exceptions

Two, and they run in opposite directions.

Ask immediately if you are blocked on access to something, or if you are about to do something irreversible, or if you have been given a deadline today. None of those improves with twenty minutes of private effort.

Give it longer if the thing is plainly part of the job you were hired to do. Working out why a test fails is engineering; it is not being stuck in the sense this note is about. The distinction is whether the answer is knowledge somebody already has — a password, a convention, a location, a name — or something nobody knows yet.

That is the real line. Twenty minutes is the right budget for recovering a fact somebody else is carrying around in their head. It is not a budget for thinking.

What the rule costs you

A small number of slightly premature questions, which is the point.

In the panel above the five occasions cost ten and a half hours between them, and the first two account for nine of those. The three after the rule cost half an hour each and ended in an answer.

Asking three times when twice would have done costs your colleagues about eight minutes in total. Not asking costs you a day, and costs them nothing, which is exactly why the incentive runs the wrong way and why the rule has to be mechanical.

When to stop using it

Around the point where you can usually tell. For most people that is somewhere in the second month, and you will notice it as the rule quietly ceasing to be necessary rather than as a decision.

Until then, set the timer.

What to do with the twenty minutes you save

The rule frees up time and it also changes what the time is for.

Before it, a stuck afternoon is spent in a loop: searching, failing, searching the same thing with different words, and a slowly rising background noise about whether you should be able to do this.

After it, twenty minutes is a bounded search with a defined end. You are not deciding anything, you are running a procedure, and at the end of it you write a question.

That difference matters more than the hours. Being stuck is unpleasant mostly because it is open-ended; a timer closes it, and the closing is what makes a first month survivable rather than merely shorter.

The hours saved are real as well. In the panel above they come to about a working day a fortnight, which in a first quarter is most of a week.

Also in your first weeks