How many notifications people get
The figures most often repeated online are much broader than the studies behind them. Here is what was actually measured.
| Study | Who | What it found |
|---|---|---|
| Pielot, Church and de Oliveira, MobileHCI 2014 | 15 smartphone users, one week | An average of 63.5 notifications a day, mostly messaging and email, usually viewed within minutes even with the phone on silent |
| Pielot and others, MobileHCI 2018 | 278 smartphone users | A median of 56 notifications a day |
| Common Sense Media, 2023 | About 200 Android-using 11 to 17 year olds | More than half got 237 or more a day; about 23% arrived during school hours; they checked their phones more than 100 times a day on average |
None of these is "the average person". They are small samples, one is Android-only, and the adult studies are several years old. Even so, they agree on the order of magnitude: dozens a day for adults, hundreds for many teenagers, and a habit of looking quickly whatever the phone's settings say.
Most notifications are not opened
Marketing platforms measure this directly. Airship's 2026 benchmarks put median push opt-in at 53.3% on Android and 48.85% on iOS, so roughly half of app users have push turned off. Among those who keep it on, direct open rates range from under 1% in the weakest app categories to about 21% at the 90th percentile in the best. The benchmarks give ranges by category and percentile, not one overall figure, and any single "average push open rate" you see quoted was not taken from them.
Read plainly: for most apps, most pushes are seen at a glance, if at all, and dismissed.
An ignored notification still costs something
Dismissing a notification is not free.
- Attention. In a lab experiment with 166 students, receiving a call or text notification, without answering it, significantly hurt performance on a sustained-attention task, by an amount comparable to actually using the phone (Stothart, Mitchum and Yehnert, 2015).
- Inattention over a week. In a two-week field experiment with 221 people, participants reported more inattention and hyperactivity symptoms in the week they kept notifications on than in the week they kept them off (Kushlev, Proulx and Dunn, CHI 2016). The symptoms were self-reported, and the authors are clear this is not ADHD.
- Stress. In a lab study of interrupted work, people who were interrupted finished tasks faster with no loss of quality, but reported significantly more stress, frustration, time pressure and effort (Mark, Gudith and Klocke, CHI 2008).
- Switching. Gloria Mark's logging studies of information workers find people now stay on one screen for about 47 seconds on average before switching (a median of about 40), down from about two and a half minutes in 2004. That measures screen dwell time at work, not a biological attention span.
Two figures are left out on purpose. The "eight-second attention span, less than a goldfish" line comes from a 2015 marketing report and has been widely debunked. The "23 minutes to get back to a task" figure is often credited to a study, but it comes from interviews and fieldwork summaries rather than a published result.
Alarm fatigue: where the stakes are highest
Notification fatigue on a personal phone is an annoyance. The same effect in a hospital or an operations team is a safety problem, and that is where it has been studied hardest.
Hospitals. The Joint Commission, which accredits US hospitals, estimated that 85% to 99% of hospital alarm signals do not need clinical intervention. Between January 2009 and June 2012 it logged 98 alarm-related sentinel events, 80 of them deaths, and alarm fatigue was the most common contributing factor. When nearly every beep is a false alarm, staff learn to tune beeps out, and the real one is lost among them.
On-call engineering. Google's Site Reliability Engineering book treats paging volume as a design limit. It estimates one incident takes about six hours to handle properly, so an on-call engineer should face at most two incidents per 12-hour shift, with a likely median of zero. The point is not that problems are rare. It is that a page has to stay meaningful, so anything that does not need a human now should not page one.
Security teams. Vendor surveys of security operations staff show the same pattern, with the caveat that they are commissioned by companies selling a fix. In Imperva's 2018 survey, 66% of security professionals said they had ignored an alert because of a previous false positive. In Orca Security's 2022 survey, 43% said more than 40% of their alerts were false positives.
How the professionals answered it
The tools built for people who must respond did not answer fatigue with more notifications. They changed the shape of the alert.
- Escalation until someone acknowledges. In PagerDuty, an escalation policy notifies the first level at once; if nobody acknowledges before the timeout, the incident escalates to the next level, and acknowledging stops it. incident.io's escalation paths work the same way, setting who is paged at each level and how long to wait.
- A budget for interruptions. Google's two-incidents-per-shift guideline is a cap, not a forecast.
What follows for anything that must reach a person
Take the research together and four rules fall out. They apply as much to a friend's flight as to a server.
- Fewer. Every unnecessary alert lowers the value of the next one. The hospital and security evidence points at false alarms as the problem, because they are what teach people to ignore the true ones.
- Louder, and different. If something must wake a person, it cannot arrive looking like the 56th notification of the day. It needs its own channel: a ring that sounds like an alarm, comes through silent where the phone allows it, and fills the screen.
- Chosen. An alert the person asked for, for something specific, is one they have already decided to care about. Reminder research backs specificity: a Cochrane review of randomised trials found text reminders raised healthcare attendance from 67.8% to 78.6%, and two NHS trials found that a reminder stating the cost of a missed appointment cut missed appointments from 11.1% to 8.4%, at no extra cost.
- Shared, and answered. Acknowledgement is what makes paging work, and other people are what make acknowledgement stick. In a randomised trial of 166 adults, 95% of those who joined a programme with friends completed it, against 76% of those who joined alone (Wing and Jeffery, 1999). A meta-analysis of 241 studies found that the presence of others slightly improves performance on simple tasks (Bond and Titus, 1983).
Where LFG Alarm fits
LFG Alarm is built on those four rules for personal life rather than for operations. An alarm rings only for what you set, either a time or a moment you described: a post on X, a stream going live, a token price, a big wallet trade, an event from a Luma or Partiful invite, or a call to a private webhook link. Rocco, the cat inside the app, does the watching, so you are not the one checking.
It rings as an alarm, not a banner, though the strength depends on the phone. Time and event alarms are real alarms on both phones: on iPhone they use Apple's AlarmKit and ring through silent mode and Focus, and on Android they take over the screen. Post, price, trades and webhook alarms ring full screen on Android, even on Do Not Disturb; on iPhone they arrive as a time-sensitive alert with the alarm's own sound, which respects the silent switch, so keep the ringer on.
And it is shared: one link, lfg.gm/a/<alarm-code>, rings every phone that joined, each person taps I'm up, and the wake board shows who did. It has no escalation policies or schedules, so it is not an on-call tool. See what an AI alarm clock is, an alarm for when someone posts and alert apps compared.
Common mistakes
- Quoting "the average person gets X notifications" from a 15-person study.
- Using the goldfish attention-span figure. It does not hold up.
- Treating one push open rate as universal. It varies more than tenfold by app category.
- Fixing missed alerts by adding more alerts. The evidence says remove the ones nobody needs first.
Sources and further reading
- An in-situ study of mobile phone notifications, Pielot, Church and de Oliveira, MobileHCI 2014
- Pielot and others, MobileHCI 2018
- Teens are bombarded with hundreds of notifications a day, Common Sense Media, 2023
- Mobile app push notification benchmarks for 2026, Airship
- You're distracted even when you ignore that text, Association for Psychological Science, on Stothart, Mitchum and Yehnert, 2015
- Silence your phones, Kushlev, Proulx and Dunn, CHI 2016
- The cost of interrupted work: more speed and stress, Mark, Gudith and Klocke, CHI 2008
- If you can't pay attention, you're not alone, UC Irvine, and Our 47-second attention span, Steelcase, with Gloria Mark
- Sentinel Event Alert 50: medical device alarm safety in hospitals, The Joint Commission, 2013
- Being on-call, Google, Site Reliability Engineering
- 44% of security professionals spend more than 20 hours a week responding to alerts, Security Boulevard, on Imperva's 2018 survey
- 2022 alert fatigue report, Orca Security
- Escalation policies, PagerDuty, and Escalating, incident.io
- Mobile phone text messaging for appointment reminders, Cochrane, Gurol-Urganci and others, 2013
- A zero-cost way to reduce missed hospital appointments, UK government, on Hallsworth and others, PLOS ONE, 2015
- Benefits of recruiting participants with friends, Wing and Jeffery, 1999
- Social facilitation: a meta-analysis of 241 studies, Bond and Titus, 1983
LFG Alarm is not affiliated with Airship, PagerDuty, incident.io, Imperva, Orca Security, Luma or Partiful.
