Why you stopped noticing your reminders

It is twenty to ten and the phone goes off on the arm of the sofa.

Pay the electricity bill.

You cannot. The login needs a code sent to an email account you can only access on the laptop, the laptop is upstairs on charge, and you are three minutes from the end of something you have been trying to finish all evening. So you swipe it away.

Tomorrow at twenty to ten, it does it again. And the night after that.

By the fourth night you have stopped reading them. You see the shape of the banner, you know roughly what it is, and your thumb clears it before your eyes finish the sentence.

This gets called a discipline problem

It is easy to describe what happened as you letting things slide. You set a reminder, you ignored your own reminder, and the bill is still not paid. It is easy to see that as a personal failing.

But look at what the reminder actually did. It arrived four times. On none of those four occasions could you have done the thing. It was not wrong about the bill. It was wrong about the moment, four times in a row, and then it trained you to ignore it.

So the problem lies in the design of the reminder rather than in your discipline, and its cost can be measured.

The ones you ignore are not free

You might think that ignoring a badly timed reminder costs nothing. Glance, dismiss, carry on.

In 2015, Cary Stothart, Ainsley Mitchum and Courtney Yehnert tested that directly. They gave people an attention-demanding task and split them into three groups. One group received calls during the task. Another received texts. The third got nothing.

The crucial part of the design: nobody in the notification groups was allowed to check or respond. They simply received the notifications.

Both notified groups did significantly worse than the control. Not because they were on their phones, because they were not. Receiving the notification was enough. The buzz raises an unanswered question (what was that?) and pulls some of your attention away from the task.

So the electricity-bill reminder did not merely fail. Each time it appeared, it also cost you a little attention.

What that costs at scale

The volume question is harder to pin down than most articles pretend, but there is decent evidence. In 2024, Nanne and colleagues published a study in Media Psychology that did not ask people to estimate how many notifications they got. It counted them, directly from the phones of 205 participants. During the baseline week, participants received an average of 154 notifications a day.

The study was also preregistered, meaning the researchers specified the design and analysis before collecting the data, which rules out picking the flattering result afterwards.

Treat that number cautiously, for reasons explained further down.

The clearest example of where this can lead is not from phones at all. It is from hospitals, where the stakes made it worth studying properly.

In 2020, Meghan Woo and Olivia Bacon reviewed the alarm literature for the US Agency for Healthcare Research and Quality. The volumes are enormous: one academic medical centre logged more than 59,000 alarms in twelve days. Another recorded almost 17,000 in eighteen days on a single ward.

Across the studies they reviewed, the share of alarms that were false ranged from 72 to 99 per cent.

Imagine a nurse's day. In some of those settings, ninety-nine alarms out of every hundred required no action. Clinicians in that environment stop responding to alarms, and the literature has a name for it: alert fatigue, the state of becoming desensitised to safety alerts and, as a result, failing to respond appropriately.

These are people who know the alarms are about patients. They are not careless. They are triaging, because when almost none of the alerts need action, ignoring them is the reasonable strategy.

Your thumb clearing the banner at twenty to ten is the same behaviour, learned the same way, over just four nights.

Why those figures need some context

The 154 figure comes with a standard deviation of 143. The average is nearly as large as the spread around it, which means it is not describing a typical day so much as an enormous range. Some people in that study were getting a handful; some were getting hundreds. Anyone quoting it as "people get 154 notifications a day" is doing something the data does not support, and this article has just quoted it, so treat it as evidence that the volume is large and variable rather than as a figure about you.

The clinical figures are a review of many studies rather than one measurement, and the 72 to 99 per cent is a range precisely because it varies enormously by ward, by device and by how the thresholds were set. The range is the finding. Picking a single number out of it would be a fabrication.

And none of this says fewer alerts is automatically better. A missed critical alert is also a failure, and a system that solves interruption by going quiet has just moved the cost somewhere you will not see it. The goal is not silence.

For anyone who designs reminders

A reminder is not just information; it is a request for action. If it arrives at a moment when you cannot act on it, it achieves nothing and still costs you attention. The first thing to get right is therefore the moment: send the reminder when the person can act on it, not simply when it was scheduled.

The same reasoning applies to how a system judges itself. If it reports how many reminders it delivered, it is measuring its own activity. What it should measure is what proportion arrived at a time when the person could actually act.

Most of what wants your attention today does not need it at a specific minute, so what is not urgent can be grouped. One evening digest can contain exactly the same information as eleven separate notifications spread across twelve hours. It just interrupts you once.

Finally, a reminder that has been ignored should not simply go out again, unchanged, at the same time. Firing the same alert at the same time on the fourth night is a system that has not noticed it is being ignored. That is the moment to ask for a better time, not to try again louder. What a design has to avoid above all is pushing people into turning everything off. People turn notifications off entirely when the ratio gets bad enough, and that decision is very rarely revisited. It is not a setting, it is an ending.

An app whose reminders you set in a sentence

This is why Nitka (heynitka.com), a capture-and-recall app, does not fix in advance how you are reminded. Some people want one digest in the evening and nothing in between. Some want a ping each time. You tell it which, in plain words, and it proposes the change for you to confirm.

It applies the rest of the list above to itself. A reminder that falls inside your quiet hours is moved out of them. After three reminders about the same thing, it stops and asks once whether the task is still on your list. And a kind of suggestion you keep dismissing stops being offered.

The same bill, on Saturday morning

The bill still needs paying. Nothing about better notification design does that for you.

But there is a version of that evening where nothing arrives at twenty to ten, because nothing could have been done at twenty to ten. The thing surfaces on Saturday morning instead, when the laptop is open and the code is one click away.

Same reminder, same person. The difference is entirely in the timing, and the timing was the only part the software was ever responsible for.

Disclosure: this article is published by Nitka, which is mentioned above. The work cited exists independently of us and the links below go to the original sources.

References

1. Cary Stothart, Ainsley Mitchum and Courtney Yehnert, "The attentional cost of receiving a cell phone notification", Journal of Experimental Psychology: Human Perception and Performance, 2015. Three groups, one receiving calls, one texts, one nothing. The notified groups performed worse without being allowed to check or respond. https://doi.org/10.1037/xhp0000100 Record and abstract, freely readable: https://pubmed.ncbi.nlm.nih.gov/26121498/
2. Nanne and colleagues, "Beyond the Buzz", Media Psychology, 2024. Preregistered trial, 205 participants, logging notifications actually received rather than self-reported: 154.00 per day in the baseline week, with a standard deviation of 143.28. https://doi.org/10.1080/15213269.2024.2334025
3. Meghan Woo and Olivia Bacon, "Alarm Fatigue", chapter 13 of Making Healthcare Safer III: A Critical Analysis of Existing and Emerging Patient Safety Practices, Agency for Healthcare Research and Quality, March 2020. Alarm volumes, and the finding that the share of false alarms across studies ranged from 72 to 99 per cent. A review of a body of research rather than a single measurement, which is why the figures above are a range. Full text, freely readable: https://www.ncbi.nlm.nih.gov/books/NBK555522/

Nitka, the app mentioned above, turns what you say, type, photograph or forward into dated reminders, and tells you when two things you told it do not fit together. It keeps only what you give it, and your data is never sold, never used for advertising and never used to train models.

Try it free for three months.* On the App Store now; on Android, ask for early access.

* Then $0.99 a month, cancel any time. More at heynitka.com.