The Hard Part of a Reminder Is Not the Task. It Is the Switch.
Stopping one thing to start another is its own executive demand, separate from whatever you are being asked to do. That is why a warning before the alarm is worth more than a louder alarm.
Here is a thing that sounds like an excuse and is not.
You can want to do a task, know exactly how to do it, have the time to do it, and still be unable to start it, because you are currently doing something else. The obstacle is not the task. The obstacle is the gap between the two.
If you have ADHD you already know this from the inside and have probably been told, repeatedly, that it is laziness. It is not. Shifting from one activity to another is a demand on executive function in its own right, and it is separate from the demands of either activity.
Why switching costs something
Russell Barkley's model of ADHD puts behavioral inhibition at the center of everything else. In his framing, the ability to stop an ongoing response is the foundation that working memory, self regulation and planning are all built on. Everything downstream depends on being able to interrupt what is currently running.
You do not need to accept any particular model to see the consequence. Switching tasks requires you to stop one set of actions, hold the new goal in mind long enough to act on it, and overcome the pull of the thing you were doing. Each of those is a separate operation, and all three are exactly the operations that are hardest in ADHD.
This is why the experience of being interrupted mid flow is so disproportionately awful. It is not that you dislike the new task. It is that the mechanism you need in order to leave the old one is the one that works least reliably.
The three places a reminder goes wrong
It arrives with no warning
An alarm that fires at exactly the moment it is due is asking for an instant switch, at whatever point in your activity it happens to land. Sometimes that is a natural gap and it works. Often it is three sentences into a thought, and the honest options are to lose the thought or ignore the alarm. Most people ignore the alarm and intend to come back. Most people do not come back.
It demands something at the worst possible moment
Alarm apps built on friction make the alarm itself into a task: solve this, photograph that. At the exact moment when your capacity to switch is most strained, the app adds a second thing to switch to. For a morning wake up, where the user is barely conscious and the friction is the point, that is a reasonable trade. During the day it is counterproductive.
It does not say what it is for
An alarm that goes off and does not say what it is for forces you to work it out. If you have eight alarms a day, that is eight small acts of retrieval, each one happening at the moment your working memory is already loaded with whatever you were doing. Looking at the screen and reading a label is more effort than it sounds when you are mid flow.
What a warning actually buys
The fix for the first problem is boring and extremely effective: warn first.
Not a second alarm. A notification, a few minutes before, that says what is coming and how long there is. "Meds in five minutes." That is it.
What this changes is the shape of the request. Instead of "stop now", it is "find a stopping point in the next five minutes". The second is a dramatically easier thing for an executive system to deliver, because it lets you choose the moment. You finish the paragraph. You get to the end of the function. You put the bookmark in. By the time the actual alarm arrives, you have already done the hard part.
There is a second effect, less obvious and arguably more valuable. A warning that lets you act early means you often never hear the alarm at all. If the warning carries a single button meaning "doing it now", the fastest possible path through the whole thing is: get the nudge, finish the sentence, tap once, go. No alarm, no interruption, no dismissal.
The rules a warning has to follow
A warning done badly is just a second alarm, which is worse than one alarm. Three things keep it on the right side of the line.
It must demand nothing. It plays once and stops on its own. It does not loop, it does not need dismissing, and ignoring it has no consequence whatsoever. The moment a warning requires an action, it is an interruption rather than a preparation for one, and you have doubled the problem you were trying to halve.
It must be unmistakably not the alarm. Different sound, different wording, different behavior. If the two converge, the distinction collapses and you are back to two alarms. In practice this means the warning should be short, soft, and harmonically resolved. Rising or unresolved sounds read as unfinished and create exactly the low grade urgency a warning is supposed to avoid.
It must be independent. Whether you saw the warning, ignored it, or were in a meeting with your phone face down has no bearing on whether the alarm fires. They are two separately scheduled things. A warning you can accidentally use to cancel an alarm you still needed is a trap.
The lead time is personal, and it varies by task
Five minutes is a reasonable default and it is wrong for about half of all alarms.
Closing a book takes ten seconds. Closing out a thought in code takes twenty minutes on a bad day. Leaving a conversation politely takes however long the other person needs. The right warning time is the time it takes you to reach a stopping point in that activity, and the only person who can estimate it is you.
A practical approach: start at five, and then pay attention to the alarms where the warning consistently feels too late. Those are the activities worth a longer lead. You will usually find it is one or two specific things, not everything.
One risk worth knowing about
If every alarm has a warning, the warning gradually becomes the real alarm, and the alarm becomes the thing that happens when you ignored the warning.
That is arguably the best possible outcome, because acting at the warning is the cheapest path. But it does mean the warning ends up carrying real load, and anything that carries real load is eventually subject to the same habituation that broke your alarm in the first place.
Worth watching in your own use. If the warnings have stopped registering, that is useful information: it means they have become the alarm, and the alarm behind them has become a backstop. Which is fine, as long as the backstop still works.
Frequently asked questions
Why not just set two alarms, five minutes apart?
You can, and it is better than nothing. Two things make it worse than a purpose built warning. The first is that both alarms behave like alarms, so both demand dismissal and the first one is an extra interruption rather than a gentler one. The second is admin: two alarms per reminder means twice as many things to create, edit and keep in step, which is friction in a different place.
Does a warning make me more likely to procrastinate the task?
In practice it works the other way round, because the warning gives you a cheap way to act early and finish with it. The failure mode to watch is not procrastination, it is treating the warning as the alarm and then needing a warning for the warning.
How is this different from a calendar notification?
Mostly in what it does afterwards. A calendar notification fires once and that is the whole system; if you miss it, nothing follows. A warning paired with an alarm is a two stage thing, where missing the first stage is completely fine because the second stage still happens.
What if I am in a meeting and cannot act on it?
Then ignore it. That is the design. A warning that is ignored costs nothing and changes nothing, and the alarm still arrives on time. It is specifically built so that the common case of not being able to respond carries no penalty at all.