Cross-business switches cost me 18–22 minutes of effective output. Same-business switches cost 6–8. That difference is what the popular "23 minutes to recover" framing buries.
I ran a structured 90-day log tracking context switches across an ad ops automation business and a content publishing operation — 12 automated workers, 4 clients, 3 publications. Every switch got a timestamp. My definition was specific: closing an active task and opening one with a different business, client, or cognitive mode (writing vs. debugging vs. client communication). No micro-distractions counted. Only deliberate switches.
Average switches per day: 11.3. On heavy ops days — campaign anomaly plus publication deadline colliding — that climbed to 18 or 19. The aggregate cost wasn't what surprised me. The variance by switch type was. Three weeks of isolating that variable made it clear: the thing that matters isn't how often you switch, it's what you're switching between. Jumping from client email to a blog draft inside the same business is recoverable fast. Jumping from ad ops triage into editorial mode is a different tax entirely.
The highest-ROI change in the whole 90 days was embarrassingly small: before leaving any interrupted task, I write two sentences in my Obsidian daily note — what I was doing, where I was in it. That's it. It cut reorientation time on return by roughly half. Not a system. Not a framework. A two-sentence dump before I close the tab.
I also ran a hard three-week test of the standard advice — strict time-blocking, one business per half-day. It failed. My ad ops clients run on campaign schedules, not my calendar. A 10:47am anomaly doesn't wait. What I replaced it with was a triage threshold: ad ops alerts only break my current block if they cross a specific revenue-impact value. Below it, they get tagged "Review at block boundary" in Notion and I check that view at the end of each block. Simple rule, but it required me to stop pretending interruptions were binary (urgent vs. not) and start treating them as a threshold problem.
I wrote up the full breakdown — including the Airtable sync failure that collapsed one of my tracking systems on day 43, and the exact Obsidian frontmatter I used to log this — over on dailyfocusmag.
Top comments (1)
The variance-by-switch-type point matches what I see. Same-business jumps barely register, but going from ops firefighting into writing mode wrecks the next half hour, and no amount of "just refocus" fixes it. The two-sentence dump before closing a tab is the cheapest win there is, I do the same thing.
The time-blocking failure rang true too. The thing that helped me most with two contexts wasn't a schedule, it was visual separation: I map each business's project to its own calendar, so when I glance at the day I only see one context instead of both bleeding together. I ended up building a small Todoist to Google/Apple Calendar sync to do that per-project mapping both ways (https://todoist-sync.com/?utm_source=devto&utm_medium=referral). Given your expensive switches are the cross-business ones, I'd be curious whether keeping them out of view until you actually switch moves the number.