Product

Why we use SMS and phone calls for the most important nudges

Push notifications are easy to dismiss. Here is the design philosophy behind escalation.

A phone call in 2026 is an intimate act. We treat it that way.

> KeptMind escalates reminders push → SMS → phone call only for tasks you mark critical and only on plans where you opted in. We prefer a missed nudge over training you to ignore us.

## Why push alone fails

Push notifications used to be powerful. In the early 2010s, a push from your todo app cut through your day. It was novel, it was rare, and it carried weight.

Now push is wallpaper. The average smartphone user receives between sixty and a hundred and twenty notifications a day. Most are dismissed without registering as anything more than a flicker on the lock screen. The brain has trained itself to treat the small visual flash as background noise. Inside two weeks of installing any new app, the user is already filtering its push notifications out of conscious attention. This is not a failure of the user. It is a rational response to volume.

For ADHD users this effect is amplified. Notification fatigue compounds with executive function variability. By Wednesday afternoon a person may have dismissed dozens of pushes that morning, and the relevant one — the medical appointment, the prescription pickup, the kid's thing — has become indistinguishable from the noise. The reminder system has failed silently.

The neuroscience here is worth understanding. ADHD brains have a documented deficit in prospective memory — the ability to remember to do something in the future. This is not the same as forgetting facts; it is forgetting intentions. A push notification is supposed to bridge that gap, but it only works if the notification reaches conscious attention at the moment the action is possible. When the notification arrives during a hyperfocus session, or while the user is driving, or during a conversation, it registers as a flicker and is gone. The intention was never retrieved. The brain did not fail to remember — it failed to be interrupted effectively. This is why channel diversity matters: different channels have different interruption profiles, and the right channel depends on what the user is doing when the reminder fires.

There is also a learned-helplessness dimension. After enough missed reminders, the ADHD user begins to distrust the entire reminder system. They stop marking tasks as important because "it won't matter anyway — I'll just dismiss it." This is not apathy; it is a rational response to a tool that has repeatedly failed to reach them. Breaking this cycle requires a reminder system that occasionally succeeds in a way the user notices. One SMS that actually gets through, one call that actually prompts action, rebuilds trust in the entire stack. The escalation ladder is as much about restoring faith in reminders as it is about delivering any single one.

## The escalation ladder

The only reminder model that actually reaches through is escalation: a push first, then an SMS if there is no response, then a phone call if the item is critical and the user has opted in.

Most reminders never need to escalate beyond push. That is the design point. The point of the escalation ladder is not to spam — it is the opposite. Because the ladder exists, push is meaningful again. The brain learns that a KeptMind push might matter, because most pushes will be the only step taken. The escalation possibility makes the first step worth listening to.

Push works for low-stakes prompts. Buy milk on the way home. Reply to that email when you have a minute. These can be missed without consequence; the push is a gentle suggestion. SMS works for medium-stakes items where missing matters but is not catastrophic — the appointment that was rescheduled, the form that needs to be sent today. SMS reaches the lock screen with a different visual treatment, often a different sound, and crosses the filter that has been built up against app pushes. Calls work for the small set of items where missing is genuinely a problem — the medication, the surgical pre-op, the international flight. They cross every filter.

## The silence between steps

The hardest design decision was the silence between push and SMS. We debated the interval for weeks. Too short and it feels like panic; too long and the window closes.

We tested intervals from two minutes to ninety. Two minutes felt aggressive — users felt watched, like the app did not trust them. Ninety minutes was effectively useless because the relevant moment had already passed. The sweet spot, both in user feedback and in actual completion rates, was between fifteen and forty minutes depending on the criticality flag and the time of day.

We landed on a default of twenty-five minutes between push and SMS, configurable per user, with smart adjustment for time-of-day (longer at night, shorter during work hours) and for energy state (longer on bad-energy days, when intrusion costs more).

The energy-state adjustment deserves emphasis. On a day the user has logged as "bad," the cost of an intrusive reminder is higher than on a good day. A phone buzzing with an SMS on a low-energy afternoon can feel like an accusation rather than a helpful nudge. So we widen the silence interval on bad days — not because the task is less important, but because the user's capacity to receive the reminder gracefully is lower. This is the same principle as the energy-aware Today list applied to the nudge system: match the intervention to the day the user is actually living, not the day you wish they were living.

## When to call

A phone call is the most invasive thing an app can do. We call only for items the user has explicitly marked critical, on plans where call escalation is enabled, and only after both push and SMS have been ignored.

The call is short. It plays a calm pre-recorded message naming the task and asks the user to press 1 to dismiss or do nothing to mark it as still open. There is no live agent. The user is not interrupted by a person; they are interrupted by a recording that they themselves authorized minutes earlier.

In the first six months of the call feature, the average user received fewer than three calls per month, and the average critical task that escalated to call was completed within ten minutes of the call. That is the entire point: a small number of moments where the system did the unusual thing, the user took the action, and the rest of the time the system stayed quiet.

The psychological mechanism behind call effectiveness is novelty within a trusted system. A phone call from an app is unusual enough in 2026 that it breaks through attentional filters that have learned to ignore everything else. But novelty alone would wear off — what sustains the effect is that the user authorized the call, knows exactly what it means, and trusts that it would not have happened unless the earlier steps failed. The call is not an interruption from an aggressive app; it is the final step of a contract the user wrote. That framing makes the difference between a call that feels invasive and a call that feels like the system working as designed.

## What we will not build

People ask why we do not just make all reminders louder. Because louder does not mean heard. When everything is loud, the signal is gone. Escalation only works if most reminders stay quiet.

We also will not gamify reminders. No streak counters tied to "responding within X minutes." No public dashboards. No "you ignored 7 nudges this week" summaries. The reminder system is not a behavior-modification tool aimed at the user; it is a service the user has hired, and it should behave that way.

Another request we decline regularly is location-based nudges — "remind me when I arrive at the pharmacy." Location tracking adds a surveillance dimension that conflicts with our privacy posture, and the accuracy of geofencing on mobile is inconsistent enough that it creates false confidence. The user thinks the system will catch them at the right moment; the system fires two blocks too early or not at all. We would rather offer a time-based nudge that reliably fires than a location-based one that works sixty percent of the time and silently fails the rest. Reliability is the foundation of trust, and trust is what makes the entire escalation model work.

## Frequently asked questions

### When does KeptMind call me?

Only after push and SMS for items you marked critical, on Pro or AI+ plans with call nudges enabled. Per-task opt-in is required. We do not call for low or medium criticality items.

### Can I disable SMS or call entirely?

Yes — both are off by default. Push is on; SMS and call require an explicit toggle in Settings → Nudges, and call additionally requires marking individual tasks as critical.

### What does the call sound like?

A calm pre-recorded voice naming the task. Short. No live agent. You can press 1 to dismiss, otherwise the task stays open and may escalate again later according to your settings.

### Does this cost extra?

SMS is included in Pro. Call escalation is part of AI+ because the per-call cost is meaningful. We do not pass per-message charges to the user — pricing is flat-rate.

### What if I am in a meeting?

Quiet hours and calendar-aware silencing both apply. We respect your "do not disturb" settings and your calendar busy state by default, with overrides only for items you have explicitly flagged as urgent enough to break through.

## Related reading

If this article was useful, these related guides cover adjacent ground and are worth reading next:

- [ADHD Productivity Apps 2026](/blog/adhd-productivity-apps-2026) - [Voice To Task ADHD Guide](/blog/voice-to-task-adhd-guide) - [Executive Dysfunction ADHD Guide](/blog/executive-dysfunction-adhd-guide)

Each of the linked articles approaches the topic from a slightly different angle, and reading two or three of them together usually produces a more complete picture than any single article can. The shared underlying neurology means that improvements in one area often unlock progress in others, which is why the topics interconnect even when they appear separate at first glance.