Monday, a little before eight. You’re back from four days at a client’s premises. Two chases are due. Three watches fired while you were gone. Six tasks look finished and wait for you to say so. Marcus wants an answer. The surveyor has confirmed. A filing came back accepted. And two hundred and forty unread messages sit underneath all of it.
The chapter before this one ended right there. Every line on that screen is true. Each one is a correct claim, produced by an arrangement doing exactly what the five chapters before it asked for. So here’s the failure that’s left. The system is right about everything, and the morning is still wrecked.
- The problem is delivery, not accuracy, and filtering can’t reach it. Drop true items to shorten the list and you’ve traded one loss for another.
- Being right too rarely loses somebody’s attention. Arriving all at once loses it too, and almost nothing is built against that.
- The interval between deliveries does more work than the wording or the ranking. A small number of deliveries a day is the useful setting.
- A day’s alerts arrive as one statement of where things stand. A short written list says what’s allowed to break it.
Ten true things at eight in the morning
Two separate failures sit on that screen. They need different fixes, so pull them apart first.
The first is a heap. A dozen unrelated matters, each with a real claim on you, all landing together. A dozen items means a dozen decisions, and you work out the order yourself. That ordering is a job. It’s the same job a flat inbox hands you. It has turned up in the one place that was meant to take it off you, in the worst hour of the week for it.
The second is quieter, and it’s what the title of this chapter is about. Some of those items are the same thing. The surveyor confirmed. That one event closes the task you set yourself. It satisfies the watch you put on the confirmation. It resolves the wait you recorded when you handed the file to Marcus. And it cancels the chase that was queued for Tuesday. Four records, one fact. Let each record speak for itself and you’re told about one thing four times. You read all four, because from the outside they look like four.
Good practice built that failure. Each of the five chapters before this one installed a kind of record that wants a voice. The missing piece is the step that checks a record against its neighbors. The better your capture, the more routes one event has to reach you.
There’s a third version, duller, and it doubles everything above. The same record arrives as a banner on the laptop and a buzz on the phone. Then a badge on an icon, a row on a list, and a summary email at six. That’s one claim delivered five times. Thinking harder about priority leaves it exactly where it is, because the priority was right. The channel count was wrong.
So the honest arithmetic of your Monday runs to about six matters. They reach you through four surfaces, in whatever order they land, at the hour you’re least able to rank them.
What happens when the alarms are right
Somebody has already run this experiment, at scale, with lives attached. Aim a large number of correct signals at a busy professional, then write down what happens.
In Sentinel Event Alert Issue 50, published April 8, 2013, The Joint Commission set out what medical device alarms were doing to hospital staff. Alarm signals can reach several hundred per patient per day. Across a unit that runs to thousands, and across a hospital to tens of thousands every day. And “it is estimated that between 85 and 99 percent of alarm signals do not require clinical intervention.”
What follows is the part worth reading twice. Clinicians “become desensitized or immune to the sounds, and are overwhelmed by information,” which the alert names alarm fatigue. Then they act on it. The alert says that “in response to this constant barrage of noise, clinicians may turn down the volume of the alarm, turn it off, or adjust the alarm settings outside the limits that are safe and appropriate for the patient.”
Those are trained people. Their job is one where missing a signal costs somebody their life, and they switch the alarms off. The alert counts the result. The Joint Commission’s Sentinel Event database “includes reports of 98 alarm-related events between January 2009 and June 2012. Of the 98 reported events, 80 resulted in death, 13 in permanent loss of function, and five in unexpected additional care or extended stay.” Alarm fatigue was the most common contributing factor named. The alert is careful about its own numbers, and anybody quoting them should be too. Reporting “is voluntary and represents only a small proportion of actual events,” so these “are not an epidemiologic data set.”
That governs hospitals, and your firm sits well outside it. The devices are ventilators and cardiac monitors. A task list that speaks a dozen times on a Monday costs you a morning, and that’s the whole of the stake. What transfers is the mechanism. A person’s willingness to attend to a channel is a budget. Every signal spends some of it, and the signals that turn out to have been worth it put some back. Overdraw the budget and diligence is the first thing to go. They start editing the channel, usually in the crudest way available.
Now the part that makes your case harder. The hospital’s problem is precision. Nearly all of those signals could have stayed quiet, so there’s enormous room to raise fewer. Every item on your Monday screen earned its place. Every one is a genuine obligation with your name on it. The standard remedy is built for noise, and your list is signal. Apply it anyway and you lose information. A list that gets quieter by dropping true items is your memory again, with an interface on it.
Check this rather than assume it. It takes two weeks. Every time a prompt reaches you, note whether you acted on it that day. If fewer than about half turn out to be worth acting on, the problem sits upstream, with the earlier chapters. Look at what’s being captured, whether the dates are right, and whether finished work is closing itself out. If most of them are worth acting on and the mornings are still bad, this chapter’s problem is yours, and delivery is the lever left.
The interval is the finding
That lever is stronger than it sounds, and there’s decent evidence about where to set it.
The clearest is a randomized field experiment at Duke and Georgetown, reported as Batching smartphone notifications can improve well-being. It’s by Nicholas Fitz, Kostadin Kushlev, Ranjan Jagannathan, Terrel Lewis, Devang Paliwal and Dan Ariely, published in Computers in Human Behavior in 2019. They built an app and recruited 237 people. For two weeks they ran four conditions: notifications as usual, notifications batched three times a day, notifications batched hourly, and no notifications at all.
Three results, and the combination is what matters.
Three batches a day helped. “Compared to those in the control condition, participants whose notifications were batched three-times-a-day felt more attentive, productive, in a better mood, and in greater control of their phones.” They also unlocked their phones less often, and the app measured that rather than asking about it.
Hourly was a wash. “The hourly batching condition did not differ from the control, except for a single effect on the daily measure of feeling interrupted by notifications.” A batch delivered often enough is an interruption with a delay on it.
Turning everything off was worse. “In contrast, participants who did not receive notifications at all reaped few of those benefits, but experienced higher levels of anxiety and ‘fear of missing out.’” Silence buys you anxiety instead. You spend the day wondering what landed, and the wondering costs you in the background.
That’s a study of consumers and phones, rather than firms and obligations. Most of the outcomes are self-reported, and two weeks is two weeks. Take the direction and leave the size. The direction is that when a system speaks is a real variable with real effects. A small number of scheduled deliveries beats a continuous drip. And an hourly batch behaves like the drip.
One distinction has to stay straight here, because the evidence pulls apart along it. There’s how often a system reaches you. That’s what the study manipulated, and it’s what this chapter is about. Then there’s how often you go and look. That’s a different behavior with its own literature, including a field study that found no measurable stress benefit at all for people who arranged their own email checking into clusters. Both halves of that argument are set out in why you check your email constantly. Read it if you were planning to fix your Monday with a rule about when you open your mail. The short version is that clustering your own checking leaves the wondering in place, and the wondering sends you back between clusters. Scheduling the delivery settles it, because you know when the next one comes and what it holds.
The drip feels free, and you pay for it anyway. That’s a separate line of evidence again, and it’s priced properly in the real cost of email overload. The one-sentence version is that interrupted work gets done faster and thinner. The charge lands in effort and strain rather than in hours, which is exactly the account nobody audits.
The short list that is allowed to interrupt
A batch that holds everything is the wrong reading of all this. Let a digest swallow the one thing that had to move that hour and it’s worse than the noise. It only has to happen once. After that you check between deliveries, and the arrangement is over.
So a schedule needs an exception list, and the list has to be written rather than felt. Four kinds belong on it.
A clock that runs out today, set by somebody else. A filing deadline, a board that meets at ten, a court date, a wire that has to go before the cutoff. A deadline you set yourself can move by definition, so it belongs on the other list.
The thing you set a watch on because you’d change your plan the moment it happened. That’s the whole reason a watch exists rather than a reminder, and it’s the case setting your first watch in plain words is built for. If it can wait four hours, it was a task with a date on it.
Somebody stopped until you answer. Blocked, rather than waiting. The junior whose next step starts with your reply. The client whose closing is held on your one line. The test is whether your delay costs them the day.
A person in front of you, or about to be. A meeting in twenty minutes changes what you need to have read. Everything else about that client can wait until the afternoon.
Then the harder half, which is what stays off. A message from an important person about a matter with no date. A chase that would go out just as well at four as at eight. Anything you’d handle identically three hours later. The blunt test is whether you’d interrupt a client meeting for it. Everything else waits for the next delivery, whatever it feels like at the time.
Sorting a given thing into a category is the two-axis question worked through in important versus urgent. Read it once before you write your list. The item that lands on an interrupt list wrongly, more reliably than any other, is the important one with no deadline attached.
There’s a piece of institutional design worth borrowing here, from the same place as the alarm evidence. A whole profession was switching its alarms off, and The Joint Commission’s answer was to make each hospital do the deciding. It issued National Patient Safety Goal NPSG.06.01.01, effective January 1, 2014 for hospitals and critical access hospitals. During 2014, hospitals had to “identify the most important alarm signals to manage.” The stated criteria include “risk to patients if the alarm signal is not attended to or if it malfunctions” and “whether specific alarm signals are needed or unnecessarily contribute to alarm noise and alarm fatigue.”
The second phase went further. As of January 1, 2016, the policies had to address “who in the organization has the authority to set alarm parameters,” “who in the organization has the authority to change alarm parameters,” and “who in the organization has the authority to set alarm parameters to ‘off.’”
That last clause is the one to borrow. The interrupt list is a permission, and permissions grow unless somebody owns them. In a firm of two or four, let anybody promote a category to “reach me anywhere” and the list covers most of the day’s traffic inside a quarter. One entirely reasonable addition at a time. Decide once who may add to it, and put a date on revisiting it. That’s most of what keeps it short. Where two people share the work, that decision has to be joint to count, which is a version of the problem when you stop having to ask who has it takes up from the other end.
Be careful how far you carry the analogy. That goal governs accredited hospitals, and it covers physiologic monitors and infusion pumps. The Joint Commission’s own explanatory report draws the line at software. In its words, “in general, this does not include items such as nurse call systems, alerts from computerized provider order entry (CPOE), or other information technology (IT) systems.” So it’s a rule for hospital equipment. What you take from it is the shape, which is a well-tested answer to who gets to decide what’s allowed to shout.
The ten that are really one
Back to the second failure. It’s the one that belongs to a system that tracks well.
The requirement is easy to state, and most tools miss it. An alert should be about a matter, not about a record. When the surveyor confirms, you get one line. It says the confirmation landed and names what it settles: the task, the wait, the watch, the chase queued for Tuesday. One line, in place of four notices that share a date. Your question was always “is the Hallett file with the surveyor,” rather than “which of my records changed.” A system holding four records about that has everything it needs to answer in one sentence.
The same collapse works in the other direction. A client’s reply answers two of the three things you asked. It reaches you once, saying which two, with the third still open. The first chapter made the point that one message routinely opens two loops pointing opposite ways. One message routinely closes several as well. A system that catches the openings and misses the closings is the one that produces a Monday like this.
Then the channel problem, which is yours to fix rather than anybody’s to build. The previous chapter but one settled where a prompt about a message should land, which is where the reply would land. The extension is the word “only.” Pick one surface per class of thing and turn the duplicates off. That includes the helpful daily summary email that repeats a screen you already read. Two surfaces showing the same item give you two places to check whether the other one is out of date. That’s the whole of what they give you.
It’s fair to say what the mainstream instruments offer, because it’s real. Gmail’s mobile app carries a notification level Google’s own help page describes plainly: “You can set notifications for only high priority emails” (checked September 6, 2026). That’s a real reduction, and for many people it’s worth having. It answers a different question from this chapter’s. High priority there is a judgment about a message, formed from signals like who sent it and how you’ve treated similar mail. It ranks the message. What you owe somebody by Thursday sits outside what a mail app knows. Ranking messages and consolidating obligations are different jobs, and a setting solves one of them.
A statement of position, not a log
Here’s the shape the delivery should take, in a vocabulary you already have.
A transaction listing tells you everything that happened. A trial balance tells you where you stand. Both are correct, and one of them is any use before a meeting. What arrives after four days away is a transaction listing. The surveyor confirmed on Thursday. A chase fired Friday. A watch fired Friday afternoon. A reply landed Saturday. And it’s now Monday. What you want is the balance.
The difference is a practical one. In a listing, an item that fired on Thursday and resolved itself on Friday appears twice, and you read it twice. In a statement of position it’s either gone, or it’s one line about things that closed themselves. Follow the series this far and a decent share of any four-day gap is exactly that: loops that opened and shut while you were away.
That suggests the order to read in, and the first item is the one most briefings put last.
What’s gone. Start with what’s already handled. It’s the fastest cut to the size of the problem, and it’s the only part of the reading that shrinks as you go. Two chases resolved themselves. Three tasks were confirmed finished. One meeting canceled. A dozen items become half that before you’ve made a single decision.
What expires today, and can’t move. Short, dated, and first among the live items.
Who’s blocked on you. People, rather than tasks. These cost somebody else their day, and you can clear them in four minutes each.
What moved while you were away. The filing accepted, the confirmation received, the invoice paid. Facts to absorb rather than work to do. Skip this one and a plan quietly goes wrong.
Everything else, counted. A number, rather than a list.
That last line is what makes the whole arrangement usable, and it’s the part most digests leave out. A briefing you can trust declares its own scope. Six things need you today, thirty-eight others are open, none of them expire this week. Give the reader that count and a short list reads as a complete one. Leave it out and a short list and a broken list look identical. So you go and check the full list, and the check costs you everything the summary saved. State the exclusion and the reader can stop. That’s the entire product.
Two practical notes on sizing. A briefing that takes twelve minutes to read is a meeting with yourself, and it gets skipped in the week it matters most. And an overnight briefing and a four-day briefing are genuinely different objects. Overnight, what’s new is the story. After four days, what’s new is mostly noise. The story is what moved, what went stale, and what’s now late. A system that hands you the same format for both has built one of them.
How Point gets it down to one
Point is the mail client, rather than a notifier attached to one, and that’s what makes the consolidation possible. Mail and your team’s messages arrive in a single feed. The tasks, waits and watches this series has been building sit on that same surface, so one system holds the whole account of what happened while you were away.
The feed puts today’s real work at the top, ahead of the newest arrival. The ordering job from the top of this chapter is done before you sit down. Each thread carries a line of its own, so a long night’s mail reads as a list.
The briefing is the delivery. A short one gathers what needs you today. You choose when it comes and how often, which is the interval question above turned into a setting. Come back from time away and one digest tells you what happened and what still needs a decision. That’s the balance rather than the listing.
You hear about a thing once. Say in plain words what you want to hear about, and that’s what reaches you, at the moment it happens. A watch that fires speaks once, rather than becoming a queue of reminders about itself. Which categories earn an interruption, and which hold until the next briefing, is yours to decide. That’s where the written list from three sections ago goes.
The list itself gathers across every conversation and each business you run. That matters here, because a second list is a second alert stream, and the Monday is where it goes unread.
Point’s reach is set for each type of action, rather than once for everything. At the bottom of the dial Point suggests and stops. In the middle Point prepares something and waits for you. At the top Point acts. Every action type starts in the middle, so anything above that is somewhere you moved it deliberately. What Point did is written down in plain sentences, in the order it happened. Undoing an action starts from that record. Mail already delivered to another company’s server is the one thing that stays sent. The benefits page has the rest.
Choosing which of six real obligations deserves your morning stays with you. Point puts them in one place, in an order, with the resolved ones already struck out. Choosing among what’s left is the part that was always yours.
What six chapters were actually for
Be accurate about the trade this series proposed. It was a promise about the weight being visible, rather than a promise of less work.
You started with obligations held in memory. In there they’re invisible, unrankable and permanent, and the only feedback is a client mentioning something in October. Six chapters moved them somewhere they can be counted. They get lifted out of the paragraph they arrived in. They’re given a day on which silence starts to mean something. A handoff is held with a name and a return leg. A chase is prompted early enough to leave room. A task is retired when the work is done. And the whole lot arrives in one piece rather than twelve.
The week weighs the same. What changes is that you see the weight before it lands, which is a different and better thing, and it’s the most any arrangement of this kind can honestly offer.
What’s left is smaller and harder than what it took away. Six items, all of them real, and a Monday with room for four. That call stays yours. What these chapters change is that the two you drop are dropped deliberately, in front of you, at eight o’clock, with the cost in view. The alternative is a Thursday you were too busy to open.
That’s the whole of it. To see what the first version of that morning looks like in practice, before any of this is installed, start with the first sorted morning.