I Tracked Every Interruption for Two Weeks — Here Is What Actually Broke My Focus

I was convinced Slack was the main thing wrecking my focus, right up until I logged every single interruption for two weeks and found out Slack was third, not first. This is that log — the actual method, the raw category breakdown, and the three specific changes that came directly out of numbers instead of a vague feeling that "I get distracted a lot."

The method: dumb on purpose

Every time my attention left whatever I was working on — not just external interruptions, but self-generated ones too, like suddenly remembering something and switching tabs on my own — I wrote one line in a plain text file: timestamp, one-word cause, and roughly how long the derailment cost before I got back to the original task.

09:14 - slack-dm - 3min - coworker question about deploy timing
09:41 - self - 8min - remembered I hadn't paid a bill, went to do it
10:02 - slack-channel - 1min - glanced, not relevant, closed it
10:35 - notification - 2min - phone buzzed, checked it out of habit
11:20 - self - 15min - got curious about something unrelated, fell down a research hole

No categorization scheme, no app, no automatic tracking — just a text file and the discipline to actually write the line down in the moment rather than reconstructing it later from memory, which I already suspected (correctly, it turned out) would be far less accurate. Keeping each entry to one short line mattered more than it sounds — anything more elaborate than a timestamp and a few words would have meant skipping entries on busy days, which would have quietly biased the whole dataset toward calm days.

The raw numbers after two weeks

Total logged interruptions: 187
Total estimated time cost: roughly 14 hours across 10 working days

By category:
self-generated (own wandering attention):  61 entries, 38% of time cost
notifications (phone/desktop, non-Slack):  43 entries, 22% of time cost
slack-dm / slack-channel combined:          52 entries, 19% of time cost
meetings running over / starting late:      12 entries, 14% of time cost
other (email, physical interruptions):      19 entries, 7% of time cost

Self-generated distraction — my own attention wandering with no external trigger at all — was the single largest category by both count and time cost, more than Slack and notifications combined. That's the opposite of what I expected walking in, and it's exactly the kind of finding that a two-week log surfaces and a gut feeling never would, because gut feelings about your own distraction tend to blame the loudest, most annoying interruptions rather than the quietest, most frequent ones.

Why "self" ranked highest, once I actually looked at the entries

Reading back through the 61 self-generated entries specifically (not just the count, the actual one-line descriptions), a pattern emerged that the category label alone didn't show: the large majority weren't random mind-wandering, they were specific, nagging small tasks — "remembered I hadn't replied to that email," "thought of something to add to the grocery list," "wondered if that PR got reviewed yet" — that I let myself act on immediately rather than jotting down and returning to later. The interruption wasn't the thought occurring; it was choosing to act on it mid-task instead of parking it.

• Fix: a visible scratch pad next to the keyboard, physical, not digital, specifically for capturing exactly these mid-task thoughts in three seconds without opening a new app or tab that itself becomes its own distraction on the way to capturing the thought.
• Fix: a hard rule against acting on a captured thought immediately unless it's genuinely time-sensitive — the whole point of the pad is deferring the action, not just relocating where the thought gets recorded before still acting on it right away.

What the notification category actually revealed

43 entries, and reading them back, the specific apps causing them were wildly uneven — a small handful of apps accounted for the overwhelming majority, while several apps I'd assumed were problems barely appeared in the log at all. This directly informed which notifications actually got disabled afterward, rather than a blanket "turn everything off" approach that would have also silenced the few notifications that were genuinely rare and worth seeing immediately.

By source, notification interruptions:
news app:              16 entries  -> disabled entirely
a game I barely play:   9 entries  -> disabled entirely
calendar reminders:     8 entries  -> kept, genuinely useful
messaging app (personal): 6 entries -> moved to a specific check-in time
everything else combined: 4 entries -> not worth the effort to address

Without the log, my instinct would have been to nuke every notification on the phone at once, which would have been a blunt, unmeasured fix to a problem the data showed was actually concentrated in two specific apps. The data made the fix targeted instead of sweeping, and targeted fixes are easier to actually stick to, because they don't require giving up something genuinely useful (the calendar reminders) to fix something that was barely a problem in the first place (the "everything else" category).

Slack, which turned out to matter less than I assumed

52 entries and 19% of total time cost — real, but third place, not first, and the individual entries were mostly short (a minute or two) rather than the long derailments I'd assumed Slack was causing. The actual expensive interruptions weren't Slack messages themselves; they were what happened after — the couple of minutes spent re-orienting back into the original task once the Slack tangent was over, which the log's "estimated time cost" field was specifically designed to capture, separate from just counting how many messages arrived.

What two weeks of logging cost me, honestly

The log itself wasn't free — writing each entry took a few seconds, but more importantly, the act of noticing an interruption in order to log it occasionally became its own tiny meta-distraction in the first couple of days, before it turned into enough of a habit to feel automatic. I'd estimate the logging overhead itself cost somewhere in the range of ten to fifteen minutes a day, mostly in that first week — a real cost worth stating honestly rather than pretending the measurement itself was free, though it was a clearly worthwhile trade for what the two weeks of data actually revealed afterward.

What changed after implementing the three fixes

I didn't run a second formal two-week log to measure the "after" state precisely — a reasonable follow-up I still haven't done — but a rough week-long spot check a month later, using the same lightweight logging method for just five days, showed self-generated interruptions down noticeably from the original baseline, with the physical scratch pad specifically getting credited by name in more than one entry ("almost went to check that thing, wrote it on the pad instead"). Notifications dropped even more sharply, which makes sense given that fix was a one-time, structural change (disabling two specific apps) rather than a habit that has to be maintained daily like the scratch-pad rule does.

A finding I almost missed by only looking at totals

The category totals told most of the story, but breaking the "self" category down further by time of day revealed something the aggregate numbers alone didn't show: self-generated interruptions clustered heavily in a specific window, roughly ninety minutes after I'd normally started work, rather than being spread evenly across the day. That timing lines up suspiciously well with when the initial focus of starting a task typically wears off and attention naturally starts looking for something else to latch onto — which reframed the fix slightly: the scratch pad matters most specifically in that window, not uniformly across an entire working day, and knowing that made it easier to actually notice and use the pad in the moment it was needed most, rather than treating it as a vague all-day habit to maintain.

Comparing weekdays, since I half-expected Mondays to be worse

Breaking the log down by day of week was a check specifically aimed at a hypothesis I walked in with — that Mondays would show the most interruptions, given the usual narrative about weekend context-switching costs. The actual data didn't support that at all: interruption counts were roughly flat across all five weekdays, with the only real outlier being a single unusually bad Wednesday that turned out to correlate with a specific external factor (a family matter that day) rather than anything structural about Wednesdays generally. This is worth mentioning because it's exactly the kind of assumption a two-week log is good at correcting — I'd have kept believing the Monday theory indefinitely without ever actually checking it against real numbers.

Why I'd recommend the crude version over a fancy tracking app

I looked at a couple of dedicated focus-tracking apps before starting this and deliberately didn't use one, because automatic tracking (time spent in which app) answers a different, shallower question than what actually matters here — it tells you where your time went, not why the interruption happened or what it cost to recover from. The one-line manual log, precisely because writing it required a few seconds of conscious thought each time, produced richer data than any automatic tool would have, at the cost of needing actual discipline to maintain for two weeks. If you're deciding whether a real log like this is worth the two weeks, keeping each entry genuinely short — a plain word counter on your own log file after day one is a fast way to confirm entries are staying terse enough to actually sustain for the full two weeks rather than becoming enough of a chore that you quietly stop logging by day four.

If distraction for you is less about interruptions breaking a task and more about which tools structurally support or undermine focus in the first place, that's the angle covered in my deep work tool stack — the two approaches complement each other: this log tells you what's actually happening, that post covers what to change about your environment once you know.