The 23-minute interruption number is softer than you think

It is a real finding from real fieldwork. It is also 36 people, three days, and a number that does not quite match its own paper.

If you have read anything about focus in the last decade you have met this sentence: it takes 23 minutes and 15 seconds to return to a task after an interruption. It appears in productivity books, in pitch decks, and in the marketing of every tool that blocks a website. It is worth knowing what is behind it, partly because the real finding is more interesting than the slogan.

Where it comes from

The number comes from Gloria Mark, an informatics researcher who did the unglamorous thing and watched people work. In an interview with Gallup she describes the method: observe for a half day, then shadow 36 managers, analysts, developers, engineers and project leaders for three days each, recording what they did and when they switched. The same interview gives the 23:15 figure, alongside a second number that gets quoted far less: about 82 percent of interrupted work was resumed the same day, with roughly two intervening tasks before the return.

That second figure changes the picture. The popular reading is that an interruption destroys 23 minutes. What the data describes is that you go and do two other things, and the clock from leaving to returning is 23 minutes. Some of that interval is other work, not a crater.

The number does not quite match its own paper

There is a wrinkle worth knowing if you intend to quote it. Mark, Gonzalez and Harris published the underlying work at CHI in 2005, and secondary sources report the paper giving 25 minutes and 26 seconds, with a standard deviation around 54 minutes, rather than the 23:15 of the interviews. I have not read the original paper, so I am reporting the discrepancy rather than resolving it.

The standard deviation is the part that should stop you. Fifty-four minutes around a twenty-five-minute mean means the distribution is enormously wide, and a mean from such a distribution is a poor thing to build a rule on. Some interruptions cost nothing; some cost an afternoon.

What survives

Quite a lot, actually. Thirty-six knowledge workers shadowed for three days is small, and 2004-era office work is not 2026 work, but the qualitative finding has held up everywhere it has been looked at since: switching is frequent, resumption is slow, and people badly underestimate both.

What does not survive is the arithmetic people do with it. You cannot multiply your daily Slack notifications by 23 minutes and conclude you have negative working hours. The number is an average of a wildly skewed distribution from a small sample, and treating it as a per-interruption tax is a misuse of it.

The honest version is less quotable and more useful: interruptions cost much more than the length of the interruption, the cost varies hugely, and the deepest work is the most expensive to resume. Protect the ninety-minute block. Do not bother computing what it was worth.

More