The most repeated numbers about interruption are the least reliable part of this research, and what is underneath them is more useful. In the one experiment that timed it, people who were interrupted finished the same task faster than the people left alone, and reported more stress and more effort for it. The cost of a switch is not the minutes it takes, it is the piece of your attention that stays behind on the thing you abandoned. For an owner who sells the work, delivers it, invoices for it and answers the phone, that points away from the calendar advice usually attached to this research.
None of these studies were run on small-business owners. They are lab experiments with students, screen logs from information workers at large companies, and nine months inside one software team. Read at their real strength, they still say something specific about a day like yours.
Where the famous interruption number came from
The line you have heard, that it takes a fixed number of minutes to get back on task after an interruption, is normally credited to a 2008 paper from Gloria Mark's group. The number is not in that paper. It appears in an interview Mark gave to Gallup in 2006, two years before the paper it gets attributed to, and she said there that this is not so long. The alarm was added afterwards by people quoting it.
What her group did publish is field work. In 2005, Gloria Mark and colleagues shadowed 24 information workers and found they spent an average of 11 minutes and 4 seconds on a working sphere before switching or being interrupted, and that 57% of those working spheres were interrupted. When people came back to interrupted work the same day, it took an average of 25 minutes and 26 seconds, after an average of 2.26 other working spheres in between.
That is 24 people at one company, three days each, in an office. It records what happened without establishing what it cost, and nobody in it was running a business.
Mark and colleagues went back to the question in 2016, tracked 40 information workers' screens for two work weeks, and found the median stretch of focus on a screen was 40 seconds. Screens only. The time you spend on the phone, in the van, or standing in somebody's kitchen does not appear anywhere in it.
Interrupted people worked faster and felt worse
In 2008, Mark, Gudith and Klocke ran the experiment that makes this research worth an owner's time. Two groups worked through a set of simulated work emails, one of them interrupted by phone and instant message. The interrupted group finished faster, 20.3 and 20.6 minutes against 22.8, with no difference in quality.
The difference showed up in how the work felt. After only 20 minutes of interrupted work, people reported higher stress, more frustration, more time pressure and more effort, and the differences were statistically significant. People compensate for interruption by working faster, and the compensation is what costs them.
So the popular version of this, that interruptions slow you down, has the sign backwards, at least on a short task. What the experiment describes is closer to a busy Thursday: everything got done, quicker than you expected, and you were unfit for company by four o'clock. It is a lab, students, and a short task, and nobody's mortgage depended on the emails.
Attention residue is the cost that no clock shows
Sophie Leroy's 2009 research named the effect attention residue. Across two experiments she found that people struggle to pull their attention away from an unfinished task, and that their performance on the next task suffers. The residue is the part of you still writing the quote while the caller explains their problem.
Leroy found something else in the same work that is harder to act on. Finishing the first task is not enough on its own. Time pressure while finishing it is what lets people disengage and perform better on the next one. An owner's day is short of real deadlines, which is why "I will finish the quote later" behaves so badly. Later arrives with no edge on it, and the unfinished quote follows you into everything else you do that afternoon.
Exhibit 1
The interruption charges you after it ends, in the work that comes next.
This is also why the two-minute interruption is priced wrong in your head. You price the two minutes. The bill turns up on the other side, in a stretch of work that starts badly and needs a run-up before it is any good.
Preparing for a switch reduces its cost and never removes it
Reviewing the evidence in 2003, Stephen Monsell concluded that a task switch's cost is "reduced, but not eliminated, by an opportunity for preparation".
Rubinstein, Meyer and Evans measured switching costs of roughly 0.6 to 1 second per switch in lab experiments in 2001, and found those costs grew when the rules were more complex. Fractions of a second, on undergraduates classifying shapes and doing arithmetic on a screen. Multiplying that out into hours lost per week is arithmetic nobody's data supports. The widely quoted figure about task switching eating a large share of your productive time is a remark by one of those authors on a psychology association web page, with no measurement standing behind it.
What survives the trip out of the lab is the direction. A switch you saw coming is cheaper than one that lands on you, and a switch into complicated work costs more than a switch into simple work. That is the argument for putting the complicated work behind a closed door and letting the simple work absorb the interruptions.
Protected time works, and then it quietly dies
Leslie Perlow's 1999 study of a software team is the only one here where somebody changed how a real workplace ran and then watched what happened. 59% of engineers rated their productivity above average during a phase with protected quiet mornings three days a week, against 41% during a phase built around interaction time.
Those are self-ratings on a five-point scale, from one group at one company, with no control group. The engineers noticed the quiet mornings. Nobody measured what came out the other end.
The ending is the part most people leave out when they recommend focus blocks. A year on, Perlow found the quiet-time blocks had collapsed. Expect yours to go the same way unless somebody's job includes defending them.
What this means for the owner who does everything
Switching is not free, and preparation makes it cheaper without making it free. The expensive part is the residue sitting on unfinished work, so the number of things you have half-done matters more than the number of times the phone rang. And whatever you set up to fix this will decay.
Protect one block, in the part of the day your head is best, for the work that carries the most residue: the quote, the plan, the thing with numbers in it. One block on the days you can hold it is a target a firm of five can actually hit. A protected day is not.
Batch what is left. Calls returned in two passes, email in two passes, admin in one. The work is the same either way. What drops is the number of times the day has to start again, which is the instinct behind a weekly review in a fixed slot, where decisions that would otherwise arrive one at a time get one appointment. Sorting the inbox in passes is that move applied to the noisiest lane you own.
Exhibit 2
Batching leaves the same work in the day and buys one unbroken block.
Hand one lane off completely. Half a lane is worse than none, because a lane you still supervise still interrupts you, and the residue does not care that you only had to approve something. The phone is the usual first candidate in a small firm. The lane worth protecting is the one the rest of the business waits on, which is the constraint you find by tracing one job end to end.
Then give the arrangement an owner. Perlow's engineers had a good thing and lost it because nothing in the working week kept it alive. In a small firm that means the block gets checked inside the short weekly habits that already keep the business steady, and that somebody other than you can say whether it was kept.
Which lane to hand off, and to whom, is a harder question than the calendar, and it is most of what the first month of our management consulting work goes on. The test of it is whether the quote that needed one clear stretch got one this week, and whether anybody would have noticed if it had not.