When work is interrupted, researchers found it takes an average of 23 minutes and 15 seconds before people return to the original task. Meanwhile, Microsoft's workplace data shows its heaviest-connected workers can receive 275 meetings, emails and chat pings a day — suggesting chronic distraction isn't just a focus problem, but a problem built into how modern work is organized.

A work notification creates a peculiar little moral drama. The sound comes from software, the message comes from a colleague and the expectation to answer comes from the organisation. Yet when concentration breaks, the worker is often told that the problem is personal discipline.

Two figures make that story harder to sustain. Gloria Mark’s workplace fieldwork produced the often repeated average of 23 minutes and 15 seconds before people returned to interrupted work. Microsoft, looking at its own workplace telemetry, later said the most heavily contacted fifth of users could receive 275 meeting invitations, emails and chats in a day.

This is one study, not settled consensus. The Microsoft figure is a separate corporate telemetry estimate, not another measure of the same people. Neither number means quite what its most dramatic retelling suggests, and that qualification matters.

What the 23 minutes and 15 seconds actually measures

In a 2006 Gallup interview, Mark described researchers following 36 managers, financial analysts, software developers, engineers and project leaders for three working days. Observers timed events to the second. Of the interrupted work that was resumed that day, 81.9 per cent was eventually picked up again. The average return took 23 minutes and 15 seconds, with roughly two other tasks visited first.

That is not 23 minutes spent staring helplessly at a screen. The interval includes time spent handling the interruption and doing other work before returning. It is also not a laboratory measure of how long the brain needs to recover its former depth of concentration. The researchers were watching the structure of real working days.

The unit they used was a “working sphere”, which Mark described as activity connected by a common goal, people, resources and timeframe. Returning might mean resuming the broader project rather than reopening the identical document, paragraph or spreadsheet cell. The famous number describes a displaced work goal in a busy environment. Calling it a universal “refocus time” gives it a meaning the fieldwork did not test.

The published paper reports a different average

There is an awkward detail that gets lost in most retellings. The related peer-reviewed CHI paper, published in 2005 by Mark, Victor Gonzalez and Justin Harris, reports 25 minutes and 26 seconds rather than 23 minutes and 15 seconds.

That paper covered 24 information workers at one IT and accounting outsourcing company: seven managers, nine analysts and eight developers. Each was observed for about three and a half days, producing roughly 700 hours of observation. When interrupted work was resumed on the same day, people visited an average of 2.26 other working spheres before returning. The researchers found that 77.2 per cent of the interrupted segments eligible for this analysis were resumed that day.

The variation was enormous. The mean return time of 25 minutes and 26 seconds came with a standard deviation of 54 minutes and 48 seconds. In ordinary language, a single crisp average sits on top of episodes that differed wildly. Some detours were brief. Other work was left for a long time.

The Gallup interview used a larger group and gives the 23:15 figure, while the published paper gives 25:26. The public sources do not offer a tidy reconciliation, so I do not think a writer should invent one. What both versions support is narrower and more useful: after an interruption, people often travelled through other work before coming back. An old, small field study at one organisation cannot tell us how every worker behaves in 2026.

Microsoft’s 275 is a count of contacts, not attention switches

Microsoft’s 2025 Work Trend Index analysis draws on aggregated and anonymised Microsoft 365 activity. Its methodology says the 275 figure is a rolling 28-day daily count of meeting invitations, emails and chats received across a 24-hour day by the top 20 per cent of users ranked by incoming volume.

So this is not what the typical employee receives. It represents Microsoft’s most heavily contacted fifth of users in the analysed population. The telemetry ended on 15 February 2025 and excluded education and European Union tenants.

More importantly, Microsoft did not show that every counted item was noticed, answered or allowed to displace another task. An email sitting quietly in an inbox, a recurring meeting invitation and a chat asking for an immediate decision all contribute to the total. Their effects on attention may be entirely different.

The same report says the average worker receives 117 emails and 153 Teams messages on a weekday. Its widely repeated claim that employees are interrupted every two minutes comes from the average interval between pings during an eight-hour workday. A ping is an opportunity for interruption, not proof that a person switched tasks. That distinction should follow the figure whenever it is quoted.

The numbers cannot be multiplied into hours lost

It is tempting to multiply 275 by 23 minutes and declare the working day mathematically impossible. It is also wrong. The figures come from different eras, populations and definitions. The multiplication would assign a separate recovery period to every calendar invitation and unread message, including those that arrived while someone was already away from the original work.

The return interval in Mark’s study includes productive work on other tasks. Many incoming items are ignored, batched or read later. Some interruptions are useful. Several may be processed together. None of that makes fragmentation harmless, but it prevents a dramatic statistic from becoming a fictional one.

A separate controlled experiment by Mark and colleagues adds another wrinkle. Participants doing a simulated office task completed it faster when interrupted and did not produce lower-quality work, but they reported more stress, frustration, effort and time pressure. That experiment was not the source of the 23-minute figure, and it should not be treated as a model of every workplace. It does show why looking only for lost minutes can miss part of the cost.

The legitimate connection between the datasets is not arithmetic. The field observations show that returning to a displaced goal can involve a chain of other work. Microsoft’s telemetry shows just how many invitations modern systems can generate for such a chain to begin.

The content of an interruption matters more than the alert

Mark’s fieldwork makes a distinction that deserves more attention. An interruption belonging to the same working sphere could help. A colleague might resolve an ambiguity, deliver a needed answer or stop someone wasting an hour. The more disruptive shift was being pulled into a different sphere, with a different goal, group and deadline.

There was another complication: 44 per cent of interruptions in the Gallup account were self-initiated. It is easy to read that as a failure of willpower. Sometimes it surely is a voluntary detour. But a worker carrying several live priorities may check a channel because a reply is expected, because another task is blocked or because the organisation has taught them that unseen messages can become urgent without warning.

I have written before about the difference between looking busy and producing useful work. Fragmentation belongs inside the same conversation. A day can be packed with visible replies, status changes and short meetings while making slow progress on the thing that required an hour of continuous thought.

That is why turning every interruption into an enemy is unhelpful. Communication can move work forward. The design problem begins when the recipient cannot tell which channel, sender or red badge genuinely deserves attention now.

Focus is partly an organisational outcome

Personal habits still have a place. People can mute alerts, batch email and close tabs. Those choices, however, operate inside rules set by other people. If an unanswered message is read as disengagement, muting it is not merely a personal preference. If every project uses a different channel, checking becomes part of the job.

Microsoft is both a source of unusually broad telemetry and a company selling the software through which much of this traffic moves, including AI products it presents as part of the answer. That does not invalidate its counts. It means the report’s interpretation and proposed remedies deserve the same scrutiny as any vendor’s account of a problem its products help mediate.

Eliminating communication would be absurd. Coordination is work. The practical organisational questions are less glamorous: which messages deserve synchronous attention, which can wait, whether urgency labels mean anything, who may protect a block of time and whether a team shares the same expectations about response speed.

I suspect these two statistics endure because they name a contradiction many workplaces have normalised. People are asked to perform work requiring sustained attention while continually proving their responsiveness. A personal focus system may reduce some noise, but it cannot settle incompatible expectations on its own.

An organisation cannot reasonably demand deep concentration and continuous availability at the same time, then treat the collision as an individual’s failure. The architecture of attention is partly personal. The permission to use it is organisational.