Context Switching: the Cost Is in the Questions, Not the Tickets

Context switching is a well understood problem with a well established remedy: limit how many things a person has in flight. The remedy is correct and it addresses the smaller half of the problem, because it only governs the switches somebody planned.

"I lose the thread every time someone asks where a thing is."

Two kinds of context switch

A planned switch is moving from one piece of work to another at a moment you chose. You finish a thought, note where you are, and move. It costs something, and you controlled the timing, so you paid the cost at the cheapest available moment.

An unplanned switch is someone else choosing the moment. A message arrives asking where the config for the staging environment lives. You know the answer. You give it. Then you return to work you were mid-thought in, with no note of where you were, because you did not know you were about to leave.

Every framework for reducing context switching addresses the first kind. Almost none address the second, which is the majority on any team where people ask each other things.

Why question switches cost more

Three properties make them expensive out of proportion to their length.

No handover. A planned switch lets you write down where you are. An unplanned one does not, so the reconstruction afterwards is from memory rather than from a note.

No warning. You cannot finish the thought you were in, so you lose the specific piece of reasoning you were assembling rather than the general context.

Answering feels free. The answer takes twenty seconds, so both parties conclude it cost twenty seconds. That makes the behaviour invisible to the people doing it and impossible to manage, since nobody reports an interruption they believe was free.

What work-in-progress limits do and do not fix

Intervention	Fixes planned switches	Fixes question switches
Limit work in progress	Yes	No
Fewer meetings	Partly	No, sometimes makes it worse
Focus blocks on the calendar	Partly	Only if defended
Answers available without a person	No	Yes

The second row deserves a note. Cutting meetings without removing the reason for them converts scheduled group interruptions into unscheduled individual ones, which is a worse trade for anyone doing deep work. That is the pattern in no-meeting days work until someone needs an answer.

The retrieval switch

Look closely at what the question switch actually asks of you. Almost always it is retrieval: where something is, what was decided, whether an approach was tried, what the state of a thing is. You are not solving a problem for the asker, you are being used as an index.

That is the significant observation, because being an index is the one part of your job that does not require you. The knowledge is real, but the retrieval does not need your working memory to be interrupted, it needs the knowledge to be reachable some other way.

What to do about it

Keep the work-in-progress limits, they are doing useful work on the planned switches. Then address the retrieval switches directly, which means giving people a faster route than asking you.

With StandIn, you confirm a ninety-second brief at the end of each day, mostly drafted from the work you already did. Whenever you are off, including during a focus block, your StandIn answers questions from that brief in your words, with a source under every answer, and never guesses. The colleague who wanted to know where the staging config lives gets the answer without touching your attention. If it is not in the brief, they are told so and pointed to who to ask, which means the questions that do reach you are the ones that genuinely needed you.

Two honest notes. This does not eliminate the judgement questions, and it should not. And it works only because asking a StandIn is about as fast as asking you, which is the bar any alternative has to clear. See StandIn for leadership teams.

Common Questions

How long does it take to recover from a context switch?

For work holding a lot of unwritten state, commonly fifteen to twenty minutes, and sometimes the recovery is incomplete, which shows up later as a mistake rather than as lost time. The number matters less than the ratio: the recovery is usually several times longer than the interruption.

Do work-in-progress limits reduce context switching?

They reduce planned switching, which is real and worth doing. They have no effect on the unplanned switches caused by colleagues needing information, and on most teams those are the larger share.

Is it reasonable to ignore messages while in deep work?

It is reasonable and it usually fails, because the person ignoring them pays a social cost while the person waiting pays a real one. It works when the waiting person has another way to get their answer, which is what makes the boundary sustainable rather than heroic.

What is the single highest-value change?

Make the answers to retrieval questions available without interrupting a person. It is the only intervention that reduces cost for both sides at once: the asker waits less and the answerer is not interrupted at all.