The Agile Good Life

Ops & RevOps

Interrupted Work Takes 25 Minutes to Resume. Here's What That Means for Your Team.

What Gloria Mark's field research on fragmented work found, and what ops leaders can do with it

When work gets interrupted, it takes an average of 25 minutes to get back to it, and people usually work on two other projects before they do. That's the finding from No Task Left Behind? Examining the Nature of Fragmented Work, a study by Gloria Mark, Victor Gonzalez, and Justin Harris of UC Irvine. They shadowed real workers and timed everything they did, to the second. The study also found that people spent only about 11 minutes on a project before switching, and that 57 percent of their work was interrupted before it was done.

Read the full study free (PDF, UC Irvine)

If you run operations, this is one of the most useful pieces of research you can read about your team's productivity. It doesn't measure how many hours people work. It measures how broken up those hours are, and that's where most of the lost output hides.

On this page

Why ops leaders should care

Most productivity conversations focus on capacity. How many people do we have, how many hours, how much is on their plate.

This study points at something those numbers miss: continuity. Two people can work the same eight hours on the same projects and get very different amounts done, depending on how many times that day gets chopped up and how hard it is to pick the pieces back up.

That's an ops problem, not a personal discipline problem. How work comes in, how requests are routed, how many projects run at once, and what the team treats as urgent are all system design choices. The study gives you real numbers for what those choices cost.

How the study worked

This wasn't a lab experiment. The researchers went into a real company, an IT and accounting services firm working with financial clients, and watched people work.

  • 24 people: 7 managers, 9 financial analysts, and 8 software developers.
  • Three full days each, about 25 and a half hours of timed observation per person. More than 700 hours in total.
  • An observer sat with each person and logged every activity to the second: opening a document, taking a call, sending an email, a colleague leaning over the cubicle wall.

The key idea is the working sphere: a project or area of responsibility, rather than a single action. Answering an email and making a call can both belong to the same sphere, say "fix the client's reporting issue." The researchers tracked when people switched from one sphere to another, because that's when someone has to load a completely different set of context.

Finding 1: Work is fragmented by default

People were juggling an average of 11.7 different working spheres each.

They spent an average of 11 minutes and 4 seconds in a sphere before switching to another one or being interrupted.

Think about what that means for anything that takes real thought. A forecast, a process redesign, a tricky client escalation. If the average stretch is 11 minutes, most of that work is being done in pieces.

The researchers also found something they couldn't fully explain: people switched quickly even when nobody interrupted them. Their best interpretation was that people were constantly re-juggling priorities as demands shifted around them. Most workers said they'd rather finish one thing before starting another. The environment wasn't letting them.

Finding 2: Getting back takes 25 minutes and two detours

This is the headline number.

  • 77 percent of interrupted work was picked back up the same day.
  • When it was, the average time before going back was 25 minutes and 26 seconds.
  • In between, people worked in an average of 2.26 other working spheres.

Be careful how you use the 25 minutes. It's not 25 minutes of staring at the wall. People were doing other work during that time. And it's an average with a very wide spread: plenty of work was picked up quickly, and some sat for hours.

The real cost is the detours. By the time someone returns, their head has been in two or more other projects, and they have to rebuild where they were. Workers told the researchers this was the part that hurt. One described doing other things until they remembered what they'd been working on. Another said they'd forget what they were testing and retest the same thing: "It happens a lot to me."

That's rework. It won't show up on a timesheet, but it shows up in your output.

The environment made it worse. By the time people came back, they'd often opened new windows and documents, so the cues that would have told them where they were had disappeared. A blinking cursor at the end of your last sentence tells you exactly where you left off. Six new tabs don't.

Finding 3: Your owners' core work takes the most hits

The researchers split work into two kinds:

  • Central: work the person owned and was accountable for.
  • Peripheral: helping someone else, lending expertise, solving a colleague's problem.

Central work was interrupted 60 percent of the time. Peripheral work was interrupted about 42 percent of the time.

So the work you most need from each person, the work they're actually accountable for, is the work most likely to get broken up.

Finding 4: The deeper the work, the bigger the interruption

Stretches of work that ended in an interruption had lasted longer (12 minutes 40 seconds on average) than stretches that didn't (8 minutes 58 seconds). And the longer someone had been in a sphere, the longer the interruption tended to be.

The authors call this the irony of the workday. The more involved someone gets in a piece of work, the more likely it is to be interrupted, and for longer. Your people are most exposed at exactly the moment they're most productive.

Finding 5: Half of interruptions are self-inflicted

The researchers separated interruptions into:

  • External: a phone ringing, a colleague stopping by, an email alert.
  • Internal: the person stopped on their own.

Overall the split was roughly even: 52 percent internal, 48 percent external.

But it varied by role. Managers were interrupted by other people more often (about 59 percent external). Analysts and developers interrupted themselves about as often as anyone else interrupted them.

Two things for ops leaders here:

  1. Your experience of interruptions isn't your team's. If you're a manager, most of your interruptions come from other people, so it's natural to think the fix is fewer pings. For the people doing the work, half the problem is internal: unclear priorities, too many open projects, the nagging feeling that something else is more urgent.
  2. Cutting notifications only gets you halfway. The other half is structural: clearer priorities, fewer concurrent projects, and a trusted place to park the thing someone just remembered.

Finding 6: Helping work gets dropped

This is the finding I think ops leaders overlook most.

When central work was interrupted, it was resumed the same day 82 percent of the time. When peripheral work was interrupted, the kind where someone is helping a colleague or another team, it was resumed the same day only 44 percent of the time.

In other words, when someone gets pulled away from a favor, more than half the time the favor doesn't get finished that day.

If your cross-team requests arrive as hallway questions and direct messages, this is where they go to die. Nobody is being careless. The work simply has no owner-level pull bringing people back to it.

Finding 7: Seeing each other changes how people interrupt

The researchers compared people sitting next to teammates in open cubicles with people who were physically separated from their team.

  • People who sat together worked longer stretches before switching (about 12 minutes vs about 10).
  • But they were also interrupted more (63 percent of their work vs 49 percent).

The explanation came from watching people. When you can overhear a colleague, you know when they've just finished a call, so you wait for a natural break before interrupting. When you can't see someone, you have no idea when a good moment is.

The paper even proposed a fix: let people show what kinds of interruptions they're open to right now, something like "any," "none," or "only about this project." Twenty years later, that's a status setting in Slack or Teams. Most teams have the feature and no shared agreement on how to use it.

One more pattern worth knowing: stretches of work were shorter in the morning (about 11 minutes) than in the afternoon (about 14 and a half), even though the number of interruptions was about the same. Some workers said they deliberately saved their hardest work for late in the day, when things got quieter.

What the study doesn't tell us

It's an excellent study, and it has limits. The authors are upfront about them.

  • One company, 24 people. The authors say the results apply to similar workplaces: high pressure, lots of tools, lots of projects running in parallel. That describes most ops teams I've worked with.
  • It predates smartphones and Slack. The observations were done in the early 2000s. There's no reason to think work has gotten less fragmented since, but the exact numbers belong to that era.
  • It measures switching, not quality. It tells you when people moved between projects and when they came back, not how well they worked afterward.
  • The observers couldn't record video, so some calls about what counted as an interruption may be off. The authors address this and argue the main findings hold.

What I'd do with this as an ops leader

Measure fragmentation, not just capacity. Ask a few people to log every project switch for one normal day. How many spheres did they touch? How long were their longest uninterrupted stretches? You'll learn more from that than from another utilization report.

Cut the number of projects running at once. 11.7 working spheres per person is a recipe for 11-minute stretches. Every project someone carries is one more thing that can pull them away. Limiting work in progress is the most direct fragmentation fix available.

Give helping work a home. Finding 6 says favors get dropped. Route cross-team requests through a queue or intake form instead of direct messages, so they have an owner and a next step even when the helper gets pulled away.

Agree on what deserves an interruption. Decide as a team what's urgent enough to break someone's focus, and give everything else a scheduled place: a shared channel, a standup, a queue checked twice a day. Then use status settings the way the researchers suggested, so people can see when a good moment is.

Protect the deep stretches. Interruptions land hardest when someone is furthest in. If your team does its best focused work in the afternoon, keep meetings out of it.

Make it cheap to get back. Encourage a one-line "where I was, what's next" note before anyone switches. It rebuilds the cues the detours wipe out, and it takes ten seconds.

Fix the priority fog. If half of interruptions are self-inflicted, clear priorities are a productivity tool. When people trust that they're working on the right thing, they're less likely to jump to something else. I wrote about how to triage your task list with exactly that problem in mind.

Read the research yourself

Download No Task Left Behind? free (PDF), hosted by UC Irvine on Gloria Mark's faculty site, where you'll find the rest of her research on attention and multitasking.

Citation: Mark, G., Gonzalez, V. M., and Harris, J. (2005). No Task Left Behind? Examining the Nature of Fragmented Work. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI '05), 321-330. doi.org/10.1145/1054972.1055017

If your team's days look like 11-minute stretches and a dozen projects each, and you want help redesigning how work flows so people can actually finish things, that's the work I do.