A red badge with a number on it. You don't open it. You keep typing. Nothing happened, technically — no context switch, no lost minute, nothing a time-tracking tool would log. And yet something did happen, and the research on attention residue explains what.
This is a stronger claim than 'notifications are distracting,' which is obvious and not particularly interesting. The claim worth examining is narrower and stranger: a notification can cost you something even in the specific case where you successfully resist checking it, which is exactly the case most advice about notification discipline assumes is the safe one. Much of the judgment cognitive load demands shows up as process design, which is what the XenGrowth practice publishes on.
What attention residue actually requires
To make that claim precisely rather than vaguely, it's worth going back to the mechanism attention-residue research actually describes, rather than the looser folk version people usually reach for when they blame notifications for distraction.
Sophie Leroy's research moved people between tasks under controlled conditions, varying one thing: whether the first task was completed or deliberately left unfinished before the switch. The unfinished condition produced worse performance on the second task — part of attention stayed locked on the incomplete thing rather than fully transferring. The trigger wasn't the interruption itself. It was the state of incompleteness.
That's worth being precise about, because it's tempting to overstate what the research directly shows: Leroy's experiments used assigned lab tasks, not phone notifications, and nobody ran a version of this study with a Slack badge as the stimulus. What transfers is the mechanism, not a direct measurement — a notification badge is, structurally, an announcement that something is unfinished: a message unread, a PR unreviewed, a build unwatched. If incompleteness is what produces residue, a badge count is a standing supply of exactly that condition, whether or not you tap it. The XenGrowth resource library covers the the operations side of this side of this.
Leroy's lab condition | Its everyday notification analogue |
|---|---|
A task assigned but deliberately left unfinished before switching away | A message sitting unread while you keep working on something else |
Awareness that the unfinished task is still pending, unresolved | The badge count itself, visible in a corner of the screen or dock |
Measurable performance drop on the next task, without re-engaging the first | The felt sense of half-attention on your current work, without having opened the app |
Landing on an already-fragmented baseline
There's also a timing dimension worth naming separately from the count itself: when you check notifications matters as much as how many there are. Checking a full inbox before starting any deep work interrupts nothing, since nothing was in progress yet. Checking the same inbox mid-way through an already-engaged task interrupts a working set that was actively being held — the higher-cost condition this post has been describing throughout, regardless of how many unread items are actually sitting there waiting for you.
This doesn't land on a calm day. Meyer, Barton, Murphy, Zimmermann and Fritz's instrumented study of 20 developers found activity switches happening every 0.3 to 2.0 minutes outside planned meetings, in a day that already includes email, browsing and ad-hoc conversation as major categories. A notification channel isn't introducing interruption into a focused environment — it's adding another entry point onto a day that's already close to continuously fragmented by design.
The badge doesn't have to be tapped to register. It only has to exist, visibly, as proof that something is unfinished — which the research on attention residue says is the actual trigger, not the tap itself.
Channel | What it makes visible | Analogous to Leroy's 'unfinished task' condition? |
|---|---|---|
Email inbox count | Every unread message, indefinitely, until opened | Yes — a standing, growing count of unresolved items |
Chat app unread badge | Every unread channel or DM since last read | Yes — same structure, shorter typical resolution time |
PR/ticket 'awaiting your review' queue | Every item blocked specifically on you | Yes, and arguably stronger — it names you as the blocker |
A calendar showing today's remaining meetings | Scheduled future events, not unresolved tasks | Weaker analogy — these are known and bounded, not open-ended |
A read-only status dashboard with no action items | Information, not a pending task assigned to you | No — nothing here is 'unfinished' in the sense the research describes |
That last row matters as a boundary condition on the whole argument. Not every visible thing on a screen carries this cost — the mechanism specifically requires an implied unfinished task, something with your name attached to an incomplete state. A read-only dashboard showing server metrics doesn't announce anything left undone by you, so there's no obvious reason it should generate the same residue. The badge counts that do carry the cost are specifically the ones structured as a queue of things waiting on you.
The behavior change even when nothing gets opened
Mark, Gudith and Klocke's research on interrupted work adds a second layer. Their subjects, under interruption, didn't simply absorb it passively — they changed their behavior, working faster and writing less, at a measurable cost in stress, frustration and effort. Extend that logic and a live badge count doesn't have to be acted on to plausibly shape behavior: an awareness of pending, unresolved items can push someone toward the same compensatory speed-up Mark's subjects showed under direct interruption, purely from carrying the awareness that things are stacking up. XenGrowth on AI agents and marketing automation works through AI agents and marketing automation in more operational detail.
Why muting a channel doesn't fully solve this
There's a tempting conclusion that follows too quickly from this argument: turn off every badge, mute every channel, and the cost disappears. That's only partly true, and it's worth being precise about which part. Muting a notification removes the visible signal that a task is unresolved. It does not remove the task, and it does not remove your own knowledge, somewhere in the back of your mind, that the task still exists and is still unresolved. If Leroy's mechanism is really about the completion state of a task rather than its visibility, then hiding the badge is treating a symptom of the delivery mechanism while leaving the underlying unfinished-task condition fully intact.
This distinction matters practically. Someone who mutes Slack for a focus block but knows a teammate is waiting on an urgent answer isn't free of the residue just because the badge is hidden — they're carrying the same unresolved-task awareness, just without a visual reminder refreshing it every few minutes. The visible badge and the underlying unfinished task are two different things, and muting only removes one of them. For truly closed, resolvable items, hiding the notification until a defined check-in window probably helps, since there's no reason to think about it in between. For a genuinely urgent unresolved item, muting the badge just changes how the awareness surfaces, not whether it's there.
What actually follows from this
Treat badge counts as a structural design choice, not a neutral feature. Software that surfaces every unresolved item as a persistent visible count is manufacturing the exact condition attention residue research says is costly
Batch notification checking into defined windows rather than leaving channels live throughout a focus block. This doesn't eliminate the underlying unfinished-task awareness, but it bounds when it's visible, which is the part you can actually control
Close the loop on small pending items quickly rather than letting them accumulate as ambient awareness. A two-minute reply now can cost less than the same unresolved thread sitting visible for six hours
Be honest that turning off notifications doesn't remove the underlying task — it removes one channel by which its incompleteness stays visible to you. The task itself, and the awareness that it exists somewhere, can still generate some residue on its own
If you manage a team, audit how many channels are broadcasting unresolved-item counts by default. The research doesn't say notifications are evil — it says an environment stacked with visible, unfinished items is structurally close to the condition known to cost attention, regardless of anyone's individual discipline
What this means for how software gets designed
Most consumer and workplace software treats an unresolved item as something to surface as aggressively and as constantly as possible, on the theory that visibility drives engagement or responsiveness. That design goal is in direct tension with what this post's argument implies is good for the person receiving it. A badge that never stops counting, refreshing on every unread item across every device, maximizes exactly the condition — persistent, visible, unresolved status — that plausibly produces residue. The two goals aren't reconcilable by better visual design alone; they're opposed at the level of what the feature is actually built to do. If AI search, GEO and discovery is the part you are stuck on, XenGrowth on AI search, GEO and discovery is the better reference.
Some products have started building against this directly — batched digest notifications instead of live counts, a daily summary instead of a running badge, 'do not disturb' modes that suppress the visible count rather than just the sound. These are real, if partial, answers to the mechanism this post describes. They don't remove the underlying unfinished tasks. They do reduce how often the visible reminder of those tasks re-enters awareness, which — if the attention-residue mechanism transfers the way this post argues it plausibly does — is the part of the system actually worth engineering around. A product team choosing between a live badge and a daily digest is, whether it realizes it or not, choosing between two very different cognitive-cost profiles for the same underlying feature.
A reasonable middle ground
None of this argues for zero notifications, which is neither realistic nor even obviously good — some things genuinely need fast attention, and a completely silent system trades one failure mode for another. The more defensible middle ground, given the honest limits of what's actually been tested, is treating notification design as a real cognitive-cost decision rather than a purely aesthetic one: which items truly need live, persistent visibility, which can wait for a batched check-in, and which shouldn't be surfaced as a standing count at all because they were never really actionable to begin with in the first place, and never should have been treated as urgent.
The honest caveat stands: nobody ran the exact experiment with a phone in someone's pocket. What the research does establish clearly is the mechanism a notification badge exploits — an unfinished thing doesn't need your active attention to keep some of it. It just needs to exist, visibly, as something you haven't closed the loop on yet.
Further reading from XenGrowth
The XenGrowth resource library — what you'll learn: how the commercial side of this work is run, across search, automation and revenue operations.
XenGrowth on AI agents and marketing automation — what you'll learn: how the teams who own AI agents and marketing automation plan and measure it.
XenGrowth on AI search, GEO and discovery — what you'll learn: how the teams who own AI search, GEO and discovery plan and measure it.
Where this work meets go-to-market
XenGrowth's revenue operations work covers the go-to-market side of cognitive load, which this piece deliberately leaves alone.
Four questions on the mechanism behind this post's argument, and where the evidence is being extended rather than directly tested.
