The real cost of switching projects

The number everyone quotes is 23 minutes. In Gloria Mark and colleagues’ field study of fragmented work, work that was interrupted and resumed on the same day was picked up again after roughly 25 minutes on average. That field result became the widely repeated “23 minutes” claim, and the number has been repeated so often it’s become a kind of folk fact.

I think the number does more harm than good, because it frames the problem as time. It isn’t. If it were time, a two-hour block after the interruption would fully absorb the cost, and it plainly doesn’t. You can lose an hour to a meeting and come back fine. You can lose eight minutes to a different project and spend the rest of the afternoon slightly off.

What you actually lose is a structure you had assembled, and the cost is set by how much of it got overwritten.

That structure is the part mental offloading is meant to preserve: not merely the task, but the working state required to resume it.

What you were holding

Sit in the middle of a piece of real work and take an inventory of what’s live in your head. For anything non-trivial it’s something like:

• the goal, in the specific form you’d currently phrase it — not the ticket title
• the current state of the thing, including what you just tried and what it did
• three or four decisions you’ve already made and the reasons behind them
• the constraints you keep bumping into
• a small set of live hypotheses, ranked
• the next physical action

That’s a lot of structure. It took real effort to assemble — often most of the session. And almost none of it is written down anywhere, because writing it down would have slowed you down at the exact moment you were making progress.

Now something replaces it. Not a small something: an entire other project, with its own goal, its own state, its own constraints, its own cast. The new context needs the same slots your old one was occupying.

When you come back, the structure is not paused. It’s gone, and you rebuild it from the artifacts left behind: the branch you were on, the open files, the half-written doc, the last message in the thread. That reconstruction is the cost. Sometimes it takes four minutes. Sometimes you rebuild it wrong, quietly redo a decision you’d already settled, and lose an afternoon discovering that.

Attention residue

There’s a second cost, and Sophie Leroy named it well: attention residue. Her experiments on work-task transitions found that when people switch away from unfinished work, part of their attention stays with the previous task — and performance on the new task suffers.

The useful detail comes from later research on interrupted tasks: a brief “ready-to-resume” plan can reduce attention residue when you have to switch before the work is finished. Closure, in the sense that matters here, can be cognitive rather than actual.

Which means residue is a solvable problem even when the work isn’t done. You can’t finish the thing before the meeting starts. You can spend forty seconds making it clear where it stands.

The switch nobody accounts for

Most writing on this treats context switching as a work-to-work problem: project A to project B, feature to escalation, strategy to hiring.

The switch I see accounted for least is work to personal. You’re deep in something at 5:40pm and you have to leave for a school pickup, or a delivery arrives, or someone needs collecting from somewhere. That doesn’t register as a professional transition, so you don’t treat it like one. Nothing gets written down.

And then the work comes with you. Not usefully — you can’t make progress on a pricing proposal while queuing at a school gate. It just runs in the background, taking enough attention that you’re not really present for the chore either, and by the time you sit down again it has decayed into “I was doing something with pricing.”

Two contexts, half-attention each. It’s the worst version of the switch and it’s the one that gets the least deliberate handling, precisely because it doesn’t look like work.

What actually helps

I’ve tried a lot of things here. The ones that survive contact with a real week are unglamorous.

Write the handoff before you leave, not after you return. This is the whole trick and it’s easy to get backwards. When you come back, you no longer know what you knew — that’s the entire problem. The only moment the information is cheap is the moment before you go.

Write it to a stranger who has your job. Not a status update. Not a summary of what happened. A note to someone who has to start working in the next sixty seconds and knows nothing.

Five lines is enough:

• Goal — what I’m actually trying to reach
• Where it stands — the honest current state
• What matters — the decision or constraint that shouldn’t be reopened
• Still open — the question I haven’t answered
• Next move — one specific action to start with

Make “next move” small and physical. “Continue the migration” is not a next move; you’ll have to rebuild the whole context to work out what it means. “Draft the migration path for existing customers” is. The purpose of that line is to let you start before you’ve fully remembered — and starting is usually what triggers the rest of the memory to come back.

Do it for the personal switches too. Same five lines, same forty seconds, before you stand up. This is the one that changed the most for me.

What it doesn’t fix

Handing off well doesn’t make switching free. There’s an irreducible cost to moving between contexts, and if your calendar is shaped so that you switch eight times a day, the answer is the calendar, not a better note. A handoff makes each switch cheaper. It doesn’t make switching a good idea.

It also doesn’t help if you don’t trust the handoffs to come back. Written into a file you never open again, they’re just a slower way of forgetting.

But the reconstruction tax is real, it’s paid several times a day, and most of us pay it out of habit rather than necessity. Forty seconds before you stand up is a good trade.

I’m building something around exactly this boundary — switching projects, and stepping out of work and back — and I’m still shaping it with people who live it. If that’s your day, I’d like to hear how you handle it now.