What is attention residue and how much does it cost per switch?

Attention residue is what's left in your working memory after you stop doing one task and start another. The old task doesn't fully release its grip. Part of your cognitive capacity keeps chewing on it while you're nominally doing something else, which is why the first few minutes of "focused work" after a Slack detour are so unproductive.

We're going to put numbers on this, because "context switching is bad" is not actionable. "Context switching costs you roughly 9-22 minutes of degraded output per interruption, and you eat 60-120 of these a day" is something you can budget against.

Where the term comes from

The residue framing comes out of task-interruption research: give someone Task A, interrupt them mid-stream with Task B, then have them return to Task A. Two consistent findings show up across this literature:

• Performance on the resumed task (speed and accuracy) is measurably worse for a period after resumption, not just at the moment of interruption.
• The degradation is worse when Task B was unfinished or ambiguous — an open loop — than when it had a clean resolution.

That second point matters more than people think. A completed 2-minute task (you sent the email, it's done) leaves less residue than an interrupted 30-second task (you started typing the email, got pulled into a call, and it's sitting half-written). Duration isn't the driver. Closure is.

The three components of the cost

"Cost per switch" isn't one number — it's a stack of three things, and they don't all show up in the same place:

1. Reorientation time — the seconds spent remembering what you were doing before the switch and reloading context (which file, which argument, which paragraph).
2. Residue drag — reduced processing speed and increased error rate on the new task while the old task is still partially active in working memory.
3. Resumption tax — the additional time needed to get back to the interrupted task later, because you now have to reload both its context and whatever happened in between.

Most people only notice #1 because it's the only one that feels like "time." #2 and #3 are invisible — they show up as more typos, slower reading, worse decisions, and a vague sense that a 3-hour work block produced 45 minutes of actual output.

Putting a number on it

Published estimates for the visible resumption cost after an interruption typically land somewhere in the 5-15 minute range for knowledge work, with wide variance depending on task complexity and interruption length. Simple, well-defined tasks (approve this PR) recover fast. Complex, open-ended tasks (write this proposal) recover slow, because there's more state to reload.

Here's a working model we use, built from typical ranges rather than one clean study — treat it as an order-of-magnitude estimate, not a lab result:

Take the "full app switch" row: 1.5 minutes of reorientation, plus roughly 10 minutes at 40% reduced output. That 40% reduction on 10 minutes of work is equivalent to losing 4 minutes of output. Add the 1.5 minutes of reorientation and you get ~5.5 minutes of effective cost per switch — before you've done anything wrong, just from checking Slack once. Scale this by frequency. Most people who log their own screen activity for a week are surprised to find 40-80 app or tab switches in a single workday that aren't part of a deliberate workflow — not counting notifications they dismiss without switching. If even a third of those are "full app switch" grade at ~5.5 minutes each, that's roughly 13-15 switches × 5.5 minutes = 70-85 minutes of effective daily cost. That's not a rounding error. That's a workday's worth of lost output happening two or three minutes at a time, which is exactly why it's invisible without measurement.

Why open loops cost more than closed ones

The mechanical fix follows directly from the closure finding. If residue is worse for unfinished tasks, then the fix isn't "switch less" (often not fully in your control) — it's "leave fewer open loops when you do switch." Concretely:

• Before switching, write one line: what's the next physical action on the task you're leaving. Not "finish report" — "write the third bullet under Q3 risks." This converts an ambiguous open loop into a closed, resumable pointer.
• Batch genuinely similar shallow tasks (approvals, quick replies) into a single block instead of interleaving them with deep work. Same-category switches carry less residue than cross-category ones.
• Treat notifications and open-loop switches differently. A notification you dismiss without acting costs little. A notification you act on mid-task costs a full switch — so the discipline is in the "act on it now vs. later" decision, not in the notification itself.

None of this requires willpower in the moment it matters — it requires knowing, at the point of switching, roughly how expensive that specific switch is. Most people can't do that because they don't know their own switch rate. If you haven't measured yours, a week of passive logging with the free local watcher will show you the actual number instead of a guess — no data leaves your machine, which matters if you're going to trust the number enough to act on it.

What to actually do with the number

Once you know your switch count and rough cost, the useful move is not "eliminate switching" — that's not realistic for most roles that involve any collaboration. The useful move is to shrink the expensive category (full context switches on ambiguous tasks) and let the cheap category (batched shallow tasks) absorb more of your daily switch volume.

This is the mechanism behind protected focus blocks: they don't work because "focus is good" in the abstract, they work because they concentrate your switches into the cheap end of the table above. If you want the daily brief to actually flag which specific blocks in your day are bleeding the most switch cost — not just a generic "you switched apps 62 times" — that's the calculation Pro runs against your own logged data, naming the actual meeting-to-task transitions costing you the most.

FAQ