The notification audit: which alerts earn their interruption, measured, and how to silence the rest

We tell people we "need" notifications for work. Then we measure their phones and find 60-90 alerts a day, of which maybe six were ever acted on within a minute of arriving. The rest sat in a tray getting glanced at, dismissed, or ignored — but each one still cost something the moment it fired.

This is an audit, not an opinion piece. We're going to put a number on interruption cost, a number on notification value, and compare them per app. Anything that fails the comparison gets turned off. That's the whole method.

What an interruption actually costs

Task-interruption studies consistently find that resuming a task after a break takes longer than the break itself — usually somewhere in the 20-second to several-minute range depending on task complexity, and that people often shift to a different, easier task rather than returning to the original one at all. The notification itself might occupy your attention for 2-3 seconds. The recovery tax is the expensive part, and it's invisible on the lock screen.

We can't measure recovery cost precisely from local screen-time logs, but we can measure a reasonable proxy: the gap between a notification's arrival and your next app switch. If a notification from App X is reliably followed by an app switch within 15 seconds, that notification is functioning as an interruption, not an FYI. If it's usually followed by 20+ minutes of continued activity elsewhere, it was noise you paid attention-residue for and got nothing back.

The audit method

1. Pull 7 days of notification timestamps per app (not per notification — you don't need that granularity).
2. Pull your app-switch log for the same period.
3. For each app, compute: (a) notifications per day, (b) the percentage followed by an app switch within 60 seconds, (c) the percentage that led to opening the notifying app itself within 5 minutes (a "responded" proxy).
4. Multiply (a) by an estimated recovery cost to get daily interruption minutes.
5. Compare that to what you'd actually lose by turning the alert off — usually just delayed awareness, rarely a real cost.

Most people don't do step 4 with real numbers, so they keep everything "just in case." Here's what it looks like with numbers attached, from a composite week we see often in the free watcher's aggregate logs:

*Recovery cost estimates scale with how "sticky" the app is once opened — a social feed tends to hold you far past the triggering notification, while a shopping alert either gets a two-second glance or nothing.

Total: roughly 89 minutes a day of measured interruption cost across six app categories, before you've touched a single work tool. The social feed app is the biggest line item by far, and it has the lowest "led to opening within 5 minutes" ratio relative to notification volume of the bunch — meaning most of its alerts aren't urgent enough to act on immediately, but they're sticky enough to cost 3.5 minutes when they do land. That's the profile of a notification that exists to manufacture re-engagement, not to inform you of something time-sensitive.

The test each notification should pass

An alert earns its interruption if it meets at least one of these:

• Time-decaying value — the information is meaningfully less useful an hour from now (a ride arriving, a meeting starting, a person waiting on a reply).
• Low false-positive rate — when it fires, it's right often enough that checking is worth the cost. Work chat pings that are 70%+ "yes, I needed to see that within a few minutes" clear this bar. Promo pings that are 5% clear it are pure cost.
• No safe batching alternative — could you get 90% of the value by checking the app twice a day instead? If yes, it doesn't need push delivery.

Everything else — likes, "someone you follow posted," "your weekly summary is ready," most shopping and most news — fails all three and should be off, not snoozed, not "reduced," off.

Badges are a separate problem

Turning off banners but leaving badge counts on solves half the problem and creates a new one: an unread number sitting on an app icon is a low-grade, always-on prompt to check, and it doesn't decay — a badge showing "47" is functionally similar whether those 47 arrived over an hour or a week. If an app's notifications didn't earn banners, they don't need badges either. Kill both, or you've just moved the interruption from "immediate" to "ambient anxiety," which is arguably worse because it never resolves on its own.

The exception list, and why it should be short

Keep alerts on for:

• Direct messages from named people you actually work or live with (not group channels by default — those get a separate, quieter setting).
• Calendar and scheduling (time-decaying value is the whole point).
• Anything with a genuine safety or financial trigger (large transaction alerts, security codes).

That's usually 3-6 apps. If your exception list runs past 10, you haven't actually audited anything — you've just relabeled "everything" as "important."

What changes after the cut

People who go from 60-90 daily notifications down to 15-20 (keeping only what passed the test above) typically report two things in the first week: a brief spike in "did I miss something?" anxiety around day 2-3, and then a drop in phone pickups per day that's larger than the notification reduction itself — because a chunk of pickups were driven by badge-checking and app-switch habits that notifications had trained, not by anything the notification actually delivered. The habit outlives the trigger for a few days, then fades once nothing reinforces it.

If you want the audit run automatically instead of by hand — matching your notification log against app-switch timestamps and flagging which apps fail the test — that's exactly what the daily brief in Pro does; it names the specific app and gives you the one setting to change, rather than a general "reduce notifications" suggestion.

FAQ