Batching email into set windows does one thing well. It cuts how often you go to the mailbox. Then four things go wrong. A heavy Monday breaks it. The work it turns up has nowhere to go. It slows conversations faster than it slows messages. And it leaves you holding a question it can’t answer, which is what keeps sending you back in.
- The window is a fixed budget aimed at a queue other people fill. Size it from an average week and a bad week runs a deficit. The deficit carries forward.
- A window is priced as reading and paid for in writing. What you find in it is work, and the method stops at the point where the work starts.
- A delay is charged per exchange. Two windows a day can turn six minutes of typing into two days of calendar time, and the other side is the only party who feels it.
- You look between windows to answer one specific question. Opening the whole inbox is the only answer the method gives you.
What you installed when you installed windows
Batching bundles three things, and most people adopt them as one.
The first is a set of times. Mail gets handled at 11:00 and at 4:00. That’s the visible part, and the part with the name on it.
The second is silence in the gap. No badge, no preview, no glance at the phone in the elevator. Most people install about half of this, and half is worth much less than all of it.
The third is a promise about how fast you answer. You make it on behalf of everybody who writes to you, and you make it silently. Hardly anyone thinks of it that way at the start. It’s the part that eventually forces the practice open, because it’s the one piece of the three that another person experiences.
Three neighboring questions get folded into this one, and each has its own answer. How often to go to your mail at all, and where in the day to put the passes, is when to check email. Why you reach for the inbox when it’s empty, which is a pull rather than a schedule, is why you check your email constantly. Whether a system should hold its alerts and hand them to you on a schedule is a different lever, with better evidence behind it than anything you can do by hand: ten correct alerts. This page is about the method you already tried and the reasons it came apart.
One study watched people batch in the wild rather than instructing them to, and it found less than the reputation suggests. At CHI 2016, Gloria Mark and colleagues tracked 40 information workers for 12 workdays with computer logs, biosensors and daily surveys. They reported that “despite widespread claims, we found no evidence that batching email leads to lower stress.” On productivity they found one thing, in one place. Batching on its own came out insignificant, and batchers rated their days more productive than mixed-strategy people did when email time was high. Read that as narrowly as it’s written. Clustering appears to pay when there’s a lot to cluster, and it left the physiological stress measure untouched. The full account of that study, and of the checking-frequency evidence beside it, is next door in why you check your email constantly.
One detail from the same paper changes what you think you’re aiming at. A participant counted as a batcher if their three heaviest hours of email use accounted for half or more of their total. That’s a much looser thing than two sacred windows. Everyone in that study was holding a looser line than the one you tried to hold.
The window is a budget and the mail is not
Here’s the failure that feels like weakness of will. It’s arithmetic.
Two windows, 45 minutes at 11:00 and 30 minutes at 4:00, give you 75 minutes a day of capacity. You set that number by watching a normal week, and in a normal week it fits. Say your typical day’s mail takes 55 minutes to work through. That’s twenty minutes of slack, and it holds.
Then a Monday arrives that takes 120. You spend your 75 and carry 45 into Tuesday. Tuesday brings its own 55, so Tuesday is a 100-minute day against a 75-minute budget. You carry 25 into Wednesday. Wednesday is 80 against 75, and you carry 5. Thursday clears it. One bad day took three days to absorb, and it absorbed on the condition that the rest of the week stayed quiet.
Now put two bad days in the same week. That’s an ordinary week with a deadline in it. The residue survives to Friday. It sits over the weekend, and Monday opens with a backlog Monday didn’t make. From there the 11:00 window has to choose between ending on time and ending clean, which makes it something other than a window. A session that always ends badly is a session you start skipping.
The general shape is that a fixed server against variable arrivals holds and then gives way all at once. It’s fine, fine, fine, and then it’s over, and the turn is sharper than the rise in work that caused it. Why the wait climbs so much faster than the load, with the math written out, is in email overload when you own the firm.
The correction is dull and it works. Size your windows against your bad week. Track two weeks, take your fourth-worst day, and make your total window time clear it with room left over. The room is the drainage. A practice with no slack in it asks for a week where nothing goes wrong, which is a much bigger ask than discipline.
Deciding is cheap and doing is not
The second break is built into the advice itself.
Every version of the method describes a window as a processing session. You go through what arrived, you decide on each thing, you close the lid. That fits most of what’s in there. The few messages that decide your day want more than a decision. They want an answer written, a document read, a number checked, a call made.
So you open the 11:00 window and find three threads that each want 20 minutes of real work. Your window has 45 minutes. You have two moves, and each one costs you something.
You can run over. The window eats the block behind it, and you come out at 12:30 having done email and only email, which is the exact outcome you adopted batching to prevent. Do that twice and your calendar goes soft, because the entries stop predicting what happens.
Or you can stop at 45 minutes. That leaves three pieces of started work looking for a home. The window’s over, so they’re out of it. The calendar never got them. They live in your head, which is the one storage location that charges rent all afternoon.
The fix is to give one session one job. A window ends when every message has a disposition. For anything substantial the disposition is a slot on the calendar with a name on it, today or tomorrow, sized honestly. A heavy window then produces three appointments rather than three anxieties. The appointments compete with your other work in the open, where you can move them.
Two things keep this from turning into a second backlog. Anything genuinely small gets finished on the spot, with a real bound on what counts as small, and the bound is where that rule usually goes wrong: the two-minute rule. And whatever you’re deferring carries a time rather than a location, because a folder has no clock and a time brings the thing back: snoozed email.
The tax lands on conversations
This break stays out of the advice, because the cost lands on the other person.
A window delays a message by up to the gap between windows. With 11:00 and 4:00 that’s five hours at most. For a single message, five hours is fine. Most of the mail that matters is an exchange, and an exchange pays the delay once per turn.
Follow one through. A client’s question lands at 11:10, ten minutes after you closed the morning window. It waits until 4:00. You read it, write four sentences, and send at 4:20. They answer at 5:30, after your last window. You see it at 11:00 the next morning and send the confirmation that ends the thread. Total typing: about six minutes. Total elapsed: 24 hours. From where you sat, every message got handled in the very next window, so the whole day passed without you feeling a minute of it.
The largest measurement of email conversations we have makes this the common case. Evolution of Conversations in the Age of Email Overload went to the World Wide Web conference in 2015. Farshad Kooti and Kristina Lerman, with three colleagues at Yahoo Labs, looked at threads across millions of consumer mailboxes. They found a mean of 3.76 messages per thread against a median of 2. It’s consumer mail rather than professional-services mail, so take the shape and leave the value. The typical exchange is one message and one reply, and two windows a day stretch that across most of a business day. The average exchange is closer to four messages, and two windows a day stretch that across two.
Three consequences follow. Together they’re why batching feels fine for weeks and then produces a phone call from somebody who’s annoyed.
The threads that cost the most are the ones that matter the most. Mail that needs no reply pays no tax. Mail that needs one exchange pays a little. Approvals, negotiations and scope questions pay in full, and they pay on every turn. So does anything with a clock behind it, and anything where the other side needs your answer before they can move.
If they batch too, the schedules beat against each other. Two people on 11:00 and 4:00, an hour out of phase, can turn every single turn into a full day. Both of them are being punctual, and the thread takes a week anyway.
Your own sends move the other side’s replies. Sending at 4:20 makes an answer at 5:30 likely. 5:30 sits outside your day, so a reply that could have landed inside a window lands after one. The schedule partly creates the pattern it then has to absorb.
Both real remedies work on turns rather than times. Write so that a single line can close the thread. That’s a habit rather than a system, and it returns more than any tool does: email habits worth keeping. And take your highest-turn category out of email completely, which for most firms means finding a meeting time: scheduling without the email chain. What the gap between typing time and calendar time costs a firm across a year, counted properly, is how much time firms lose to email.
The look between windows is a question
The fourth break is the one people blame themselves for, and that blame is the least deserved of the four.
You check at 1:40 with the next window at 4:00. It feels inside the rules, because you only looked. Then you look again at 2:15. By the second week the gaps have quietly filled in, and the practice is gone without a day you could point to.
What’s happening is that you’re answering a question, and it’s a fair question. Is there anything in there that changes what I should be doing right now. In a firm where clients wait on you, that’s diligence. The trouble is that a window gives you exactly one way to answer it, which is to open everything, and opening everything is the thing the window exists to prevent. So the method sets your discipline against your professionalism, and in conscientious people professionalism wins.
Half-installing the silence charges you twice. With alerts still running you take the interruption anyway and still deliver the delay, which is the one arrangement that gets you both costs at once. What an interruption actually costs, counted from the logs rather than guessed at, is in when to check email.
The rule that survives contact with a real week governs what you do, not whether you look. A ten-second look against a written list of what would genuinely change your afternoon costs almost nothing, and it relieves almost all of the pressure. A five-minute answer to something that could have waited until 4:00 is what actually dismantles the block. The reply is what pulls you into somebody else’s thread. The glance leaves you in your own. Write the rule that way and you’re governing something you can keep.
Better still, let the exception reach you instead of going out to find it. A standing alert on two or three named senders or conditions does the same job as vigilance for a fraction of the commitment: setting your first watch in plain words.
Three patches and what each one costs
When windows start slipping, almost everybody reaches for one of three fixes. Here’s what each one actually does.
More windows. This is the honest response to residue, and it works right up to the point where it has dissolved the thing you were doing. Four windows is a schedule. Seven is checking email with a timetable written over it, and you keep all of the overhead for none of the protection.
Longer windows. This works on residue too, and it takes the payment from somewhere you weren’t watching: the block of uninterrupted work the windows were built to protect. There’s a second cost people underrate. The last stretch of a long email session is where the tired decisions get made, and a tired decision on a client thread costs more than a late one.
A stricter rule. The most common thing tried, and almost always aimed at the wrong thing. When your budget falls short of your bad week, resolve is pointed at arithmetic, and arithmetic wins every time. It also leaves you believing the failure was yours, which is why people abandon the method rather than resize it.
The fourth move is the one that actually holds, and it works on the load rather than the window. Reduce what has to pass through one. There are three routes:
- Fewer arrivals. Worth doing, and it takes you a short way, since most of what lands came from people you chose to work with.
- Fewer turns per thread. That’s the habits and the scheduling fix above.
- Somebody other than you doing the sorting. That’s a change of a different kind.
Whether your mail is the shape for this
How well windows suit you comes down to the shape of your mail rather than your job title.
They hold well on three conditions. Most of what arrives needs no reply or one reply. Your volume fits your budget with slack in it. And the thing that can’t wait has a route around you. A solo consultant with 60 messages a day, and a phone number their three live clients know, is close to the ideal case.
They tax you heavily when your work is made of exchanges. If your week is approvals, scope conversations, document requests and finding times, the delay compounds against you on precisely the threads that decide whether things move. That compounding is visible from their side only.
They’re risky when you’re a routing point. If mail comes into a shared or firm address and you’re one of the people who works it, your window is somebody else’s block. The failure shows up as a message everybody assumed somebody else had. The cure there is visibility: one inbox for a two-person company, and at practice scale running the firm inbox.
They’re the wrong instrument when the mail is the work. If answering is the service itself, a window is just a name for the hours you were going to spend anyway, and the arithmetic in the second section keeps beating you.
And some weeks are for suspending the practice rather than surviving it. A filing week or a live matter runs different rules for ten weeks and then goes back to normal. Treating that as a failure of the practice is how people quit something that was working in March: inbox zero during busy season.
If you came for a schedule to adopt rather than a diagnosis of the one that broke, when to check email is the page built for that. It has times, a job per pass, and what to do about the urgent exception.
A windows practice that survives a heavy week
You can keep the method. Keep it in this shape, where every item below repairs one of the four breaks.
- Give each window a written job and a legal ending. Something like this: the 11:00 window ends when every message has a disposition, and the 4:00 window ends when nobody is blocked overnight. A session with an open ending runs until you’re tired.
- Size the total against your bad week. Two weeks of honest measurement, take the fourth-worst day, and leave slack above it. Slack is the one thing that drains residue, and a practice with none is one heavy Monday from over.
- Separate deciding from doing. A window produces dispositions and calendar slots. Anything substantial leaves the window as an appointment with a name and a length.
- Give the overflow a clock. Whatever you missed comes back at a time you chose. This single change is what keeps residue temporary rather than letting it settle into a permanent floor.
- Name the threads that are exempt, in advance and by name. Usually two or three: the live matter, the approval you’re blocking, the client mid-transaction. Exempting by feel exempts everything by Thursday.
- Publish the window and give one bypass. Much of the pull to check comes from wondering whether somebody’s sitting there expecting you. A stated turnaround plus a phone number turns that wondering into a fact, and it holds better as a firm-wide commitment than as a personal habit: setting client response-time expectations.
- Write the degraded version now, while the week is calm. One window, at 2:30, with one job: anything blocking another person. A method with a reduced setting has somewhere to go on a bad week. A method with two states spends the second one abandoned.
- Count the windows you finished, not the windows you started. You can keep every time and lose ground the whole month, so the times make a poor measure. If three windows in four end with residue, the budget is wrong, and resolve has never yet fixed a budget.
That version bends. You can fall behind on Tuesday, notice, and be level by Thursday, and none of it feels like starting over. That’s the property the original lacked.
When the window stops having to be a dam
Every problem above traces back to two facts holding at once. Everything that arrives passes through you, and finding out what’s in there means opening it. A window is what you build when both are true. It’s a dam, so it has to be sized for the worst case, and the not-knowing keeps sending you back to it between times.
Point is an email app layered over the mailbox you already have, and works on those two facts rather than on your resolve.
Point weighs mail as it arrives, for what it’s worth to you and for when it’s actually due, and every thread carries a plain summary. So a window opens on the few things that decide your day, in that order, and you decide from a line rather than opening messages to find out whether they matter. That moves the number in the second section directly. A window that costs 25 minutes instead of 45 is a window with slack in it, and slack is what drains a backlog. How email triage works is the mechanism, and why importance and lateness are two different measurements is important versus urgent.
The 1:40 look gets a cheaper answer than a look. You say in one sentence what you’re waiting for, and Point tells you when it lands, which is what the exemption list in step five is really asking for. Every standing watch is visible, and you can stop it.
Things you set aside come back at the time you chose, so the residue from a heavy window has somewhere to sit with a clock on it. What moved while you were out of the mailbox arrives as one read, on a cadence you set. That’s the honest answer to what happened since 11:00: a briefing on your schedule.
And the highest-turn category, the one the tax lands hardest on, stops being turns. Point runs the scheduling back-and-forth, so finding a time stops costing four messages spread across two days of windows.
How much of this happens without you is set separately for each kind of work, on a range from suggest-only through review to fully handled. Each kind starts at review, so the preparing happens and the acting waits until you move that particular kind of work up. There’s a plain log of what was done, and undo covers what Point did. The whole inventory is on the benefits page, and only what needs you, at the top is the part of it this piece is about.
The limits, stated plainly, because this is a page about a method and Point supplies the mailbox rather than the method. Your times stay yours to choose and yours to keep. How much arrives answers to who writes to you and through which channel, and software installed in a mailbox leaves that where it is. Client patience works the same way, and the turnaround you promise is still a promise you’re making. And if your two windows have held for a year, they’ve held, and this page has no argument with that.
Common questions
Does batching email actually work?
For one thing, yes. It cuts the number of separate trips you make to the mailbox. On the thing it’s usually sold on, go carefully. The one field study that observed batchers rather than instructing them found no evidence that batching lowers stress, and found a benefit on rated productivity only for people already spending a lot of time on email. Treat a window as a container for your attention rather than as a treatment. The evidence on checking less, which is a separate claim, is in why you check your email constantly.
How many windows a day should I have?
Enough that your total time in them clears your bad week with slack, which is a number you get by measuring rather than by choosing. Most small-firm owners who make this work land on three or four, and two fails on arithmetic rather than on ambition. If you want a specific schedule with a job written for each pass, that’s when to check email.
What do I do when the window ends and the inbox is not clear?
Decide that in advance, because the version you invent at 11:45 with a meeting starting is always the wrong one. The workable rule has three parts. The window ends on time. Everything left gets a disposition in the last five minutes rather than an answer. Anything needing real work leaves as a calendar slot. Residue with a time on it is a plan. Residue in a folder is a floor you’ll still be standing on in three weeks.
Will my clients notice that I am slower?
They notice on threads rather than on messages, and that’s the part to manage. One reply held five hours passes unseen. A four-message exchange stretched over two days is very visible, especially when they’re waiting to act on your answer. Publish a turnaround you can actually keep. Give one route that skips the window. And take your highest-turn work out of email rather than trying to be quick inside it.
Should I turn notifications off between windows?
Yes, and it’s a smaller fix than it looks. Silence removes the interruptions somebody else started and leaves every one you start yourself. It also leaves the question that sends you back in, which is whether something in there changes your afternoon. The durable version is a rule about what you’re allowed to do between windows, plus a standing alert on the two or three things that genuinely can’t wait. What holding alerts and delivering them on a schedule does, which is a lever worth more than muting them, is ten correct alerts.
Does batching work when several of us share an inbox?
Poorly, and it fails differently from the solo case. Personal windows in a shared stream leave each person’s own view looking handled while a client waits four days for a message everybody assumed somebody else had. Fix the visibility before you fix the timing. Agree who takes what, and make it possible to see that every message has a hand on it. One inbox for a two-person company is that problem in full.
Is this the same failure as inbox zero?
They’re cousins. Inbox zero fails on its scoring rule, which registers one value only and gives you nothing for a day you handled well, and that argument is why the inbox zero method rarely lasts. Windows fail on capacity and latency instead. You can run both at once and hit both walls, which is roughly what a bad October feels like.
The short version
A window is a fixed budget aimed at a queue that other people fill, and the method stops at the point where the budget runs out. Size it against your bad week rather than your average one, and leave slack, because slack is the only thing that drains a backlog.
A window is priced as reading and paid for in writing. Let it end with dispositions and calendar slots rather than finished work, and give anything you’re deferring a time rather than a folder.
The delay you accepted is charged per exchange. The median thread is a message and a reply, and the average is closer to four. Two windows a day can turn six minutes of typing into two days of elapsed time, and the other person is the one who feels it. Work on the number of turns rather than the length of the gap.
And the checking you do between windows is a question. Give the question a cheaper answer than opening the inbox, write your rule about what you may do, and the block will hold on the days it actually matters.