The cost of a scattered hour

The interruption is cheap. The return is not.

The part of the interruption you do not see

Somebody asks you something and it takes ninety seconds to answer. The honest accounting is that the interruption cost ninety seconds, and if that were true nobody would mind being interrupted.

What it actually costs is the return. Before the interruption you were holding a small structure in your head: where you were, what you had tried, what you were about to try, the three things you were deliberately ignoring. The ninety seconds does not damage that structure. Putting it back does.

What the return involves

Coming back to a piece of work is not resuming, it is rebuilding. You reread the last paragraph, or look at the code again, or reopen the four tabs, and what you are doing is reassembling a state rather than continuing an activity.

Most people recognise this as the two or three minutes of staring that follows any interruption. The part that is easy to miss is that the rebuild is not always complete. Frequently you return to a slightly smaller version: you have the thread, and the two peripheral things you were holding have gone, and you will discover their absence an hour later when you duplicate work you had already decided against.

Why four interruptions can cost more than the hour contained

This is the arithmetic that makes a scattered hour worse than no hour. If each return costs three or four minutes, four interruptions in an hour remove fifteen minutes directly. The remaining forty-five is in five fragments, and some kinds of work cannot be done in nine minute pieces at all.

So the hour does not deliver three quarters of an hour of work. Depending on the task it delivers something closer to nothing, and it does so while feeling fully occupied, which is the part that makes it hard to argue about. You were working the whole time.

The distinction that matters: shallow and deep

Not all work is like this, and treating it as though it were leads to a defensiveness that does not survive contact with a real job.

Answering messages, filing, approving things, most meetings: these have almost no rebuild cost, because there is no structure being held. You can do them in a corridor. Interruption costs them the ninety seconds and nothing else.

Writing, designing, debugging, thinking something through, reading something hard: these are all structure. For these the rebuild is the real cost and the ninety seconds is a rounding error.

The useful consequence is not "never be interrupted". It is that the two kinds of work should not be attempted in the same hour, and that protecting time only matters for the second kind.

What actually helps

Three things, in descending order of how much they cost other people.

• Leave a note before you stop. One line saying where you were and what was next. It sounds trivial and it collapses a four minute rebuild into about thirty seconds, because the structure is on paper rather than being reconstructed.
• Put the shallow work together. An hour of messages is not a waste; an hour of work with messages through it is. Batching is unpopular advice because it feels slower to the person waiting, and it is, by a few hours, which is almost always fine.
• Be unreachable for one stretch, not all day. Two protected hours that people know about are easier to defend and more useful than a general unavailability that everybody works around anyway.