Deep work in practice: how long a session should be, and what to bind it to

You blocked two hours for the hard thing. You opened the file at 9:02, and by 9:40 you had written one paragraph, checked Slack four times, and read most of a thread about a bug that belongs to someone else. The block was real. What happened inside it was not the work you booked the block for.

Deep work is Cal Newport's term, from the 2016 book of that name: concentration pushed to its limit, without interruption, on something that genuinely needs it. The definition is easy and it is not where people come unstuck. What decides whether a block works is a handful of small decisions you make before you sit down, and nearly all of them are about how long the session runs and what it is for.

This article is about those decisions. How long a session should be for the kind of work in front of you, why ninety minutes is a ceiling rather than a goal, what changes when the session has a name before it starts, and how to fit any of this into a week that has other people in it.

What deep work is, and the honest ceiling

Newport draws the line between deep work and shallow work. Deep work is cognitively demanding, hard to replicate, and creates something that was not there before: the design of the migration, the argument in the middle of the report, the bug that three people have already failed to find. Shallow work is the rest of the job, and it is not worthless, it is just work you can do while half-attending: expenses, scheduling, the pull request that bumps a version number.

Most working days are mostly a third thing, which nobody has a good name for. You are at the desk, technically working, attending to nothing in particular. That third category is where the hours go, and it is why people who feel busy all week can be honestly unable to say what they made.

The ceiling is lower than the internet suggests. Newport's own estimate is modest: a few hours a day is the practical limit for someone who has built their working life around this, and someone new to it should expect considerably less. If you are getting one properly focused block on four days a week, you are ahead of the week you imagine everyone else is having.

The other thing worth being honest about is the cost of an interruption, which is mostly the return trip rather than the interruption itself. A colleague asking a thirty second question does not cost thirty seconds. It costs thirty seconds plus however long it takes to rebuild the part of the problem you were holding in your head, and that rebuild is the expensive half.

How long a session should be

The usual answers are twenty-five minutes, because that is what the Pomodoro Technique says, or ninety, because someone read about ultradian rhythms. Both are answers to a question that depends entirely on the work in front of you. A session length is a guess about how long it takes to load the problem and how long you can hold it once loaded, and those two numbers are different for drafting than they are for triage.

15 to 25 minutes: the block that breaks inertia

For the task you have been carrying for nine days, set fifteen minutes. The output is not the point. The point is that the file is open and the first sentence exists, which removes almost all of the resistance you have been paying every morning since.

This length also suits genuine shallow work done deliberately: the email backlog, expenses, the four small approvals. Doing them in one named block is much better than letting them leak across the day and interrupt everything else.

25 to 45 minutes: reading, review and editing

Work with a natural unit fits here. A chapter, a pull request, a section of a draft that already exists. You are working on material that is already in front of you, so the loading cost is low and you can start producing in the first minute.

Editing is the one people get wrong. It feels lighter than drafting, so it gets pushed into the gaps, and then it takes three times as long because you keep starting over.

45 to 60 minutes: debugging and learning something new

Anything where you have to hold a whole system in your head wants at least this. You pay a real cost to load the problem, and that cost is the same whether the session is twenty minutes long or sixty. A twenty minute debugging session is mostly loading.

Learning something unfamiliar behaves the same way. The first fifteen minutes are you remembering what you knew last time.

60 to 90 minutes: drafting and design

The first pass at something that does not exist yet is the most expensive kind of attention, and it is the one worth protecting hardest. Give it the longest block you can defend, put it where your head is clearest, and accept that some of these sessions produce something you throw away. That is what a first pass is.

Why 90 minutes is a cap and not a target

Past about ninety minutes, most people are running on recovered attention rather than sustained attention: small breaks you take without noticing, a paragraph reread three times, a slow drift into the browser. You can extend the clock and you cannot extend the thing the clock was measuring.

There is a second reason, which is more practical. An overlong session is how you learn to dread the calendar entry. Two ninety minute blocks you actually keep are worth more than one three hour block you start avoiding on Wednesday.

The focus timer in Loggd runs from 15 to 90 minutes for exactly this reason, and the top of the range is the top on purpose. The server measures the session from its start time and clamps it to the length you planned, so a tab left open overnight is recorded as the session you meant to run rather than as ten hours of heroic concentration.

Name the session before it starts

"Work on the report" is not a session. It is a two hour negotiation with yourself about which part of the report, conducted at the exact moment your judgement is worst. "Draft the two paragraphs about Q3 churn" is a session, and the difference shows up in the first ten minutes rather than the last.

The naming has to happen before the timer starts, and preferably the day before, because deciding what to work on is itself demanding work and it should not be paid for out of the block you set aside for something else. Writing tomorrow's first session down at the end of today costs about twenty seconds. This is the same principle that runs through self-discipline as a system: the decisions that matter get made once, in advance, by a version of you who is not tired.

A focus session in Loggd can be bound to a task or to a habit, which is the mechanical version of naming it. You pick the thing first and the timer second.

The second thing worth deciding in advance is where to stop. Ending a session at a natural finish is satisfying and expensive: tomorrow you have to find a new starting point. Ending mid-paragraph, with a sentence in the file saying what comes next, means tomorrow's first ten minutes are already spent.

Fitting sessions into a week with meetings in it

Most advice about deep work is written as though you own your calendar. If you do not, the useful question is narrower: which single block in the week can you actually defend, and what goes in it.

Start with one. Put it where your attention is best, which for most people is the first two hours after they properly wake up rather than the first two hours after they arrive. Defend that one block for a fortnight before adding a second, because a second block you keep breaking teaches you that blocks get broken.

The twenty minutes after a meeting is one of the most wasted slots in a week. It is too short to load a hard problem and too long to spend reading the thread you were already reading. Give it a standing job: the email backlog, tomorrow's session named and written down, the one small task that has been sliding.

Making the block recur is where a habit is more useful than a calendar entry. A calendar entry is a note about the future, while a habit carries a schedule, a check-off and the rest of the habit loop with it. A habit set to weekdays, with the focus session bound to it, is a much smaller commitment than a daily one you abandon every Saturday, and how to build good habits works through choosing a schedule you can actually keep. If the sessions are in service of something larger, put the something larger where you can see it: goals you will actually track covers the split between the goal, the habit underneath it and the task underneath that.

What usually goes wrong

The block with no name. An unnamed block gets filled by whatever asks loudest, which is almost never the thing you booked it for.

The three hour block. Booked in optimism, abandoned at ninety minutes, remembered as a failure. Book ninety and finish it.

Treating focus as a personality. People who get a lot of deep work done are usually people with a defended block and a named task, not people with unusual brains. The system is doing the work, and it is available to you.

Stopping without a note. Thirty seconds at the end of a session writing where you got to and what comes next saves ten minutes at the start of the next one. Nobody does this and everybody should.

Counting hours instead of finished things. Hours are easy to count and easy to fool yourself with. What matters is whether the migration notes exist. Two focused hours that produced a draft beat five hours of measured presence.

Tracking a session without turning it into a report

You do not need software for any of this. A kitchen timer works. A phone timer face down works. If you like the structure, a Pomodoro app gives you the twenty-five and five rhythm that Francesco Cirillo described, and there are dozens of good ones. Forest ties the session to a tree that only finishes if you stay in the app, which some people find genuinely effective and others find irritating within a week. A blocker such as Freedom or Cold Turkey removes the sites rather than asking you to resist them, which is the more reliable design.

If what you actually need is a record of hours for billing or for a client, use a time tracker built for that. Toggl and similar tools exist precisely for this and they are better at it than any habit app will be.

What a focus timer inside a habit tracker adds is that the session sits next to everything else you are keeping track of. In Loggd, a session runs from 15 to 90 minutes, bound to a task or a habit, and the minutes convert to XP at one a minute, up to the ninety minute cap. Badges arrive at thresholds for sessions completed and minutes accumulated. That is the whole of it: a count and some badges.

It is worth being plain about what is not there. Loggd does not draw charts of your focus time, there is no report to go and generate, and nothing tells you your afternoons are worse than your mornings. If you want that kind of picture, a dedicated time tracker is the honest recommendation, and running one alongside is perfectly reasonable.

The free plan includes three habits, two goals and thirty open tasks, which is enough to run one defended block a day with a habit and a task behind it. What each plan includes is on the pricing page, and how the timer, tasks and habits fit together is in the guide.

Deep work FAQ

How many hours of deep work can you do in a day?

Fewer than you would like. Newport's estimate is a few hours a day for people practised at it, and less for anyone starting out. One defended ninety minute block on most days is a realistic target that compounds faster than an ambitious schedule you keep abandoning.

How long should a deep work session be?

Match it to the work. Fifteen to twenty-five minutes to get started on something you have been avoiding, twenty-five to forty-five for reading and editing, forty-five to sixty for debugging or learning, sixty to ninety for drafting and design. Ninety is the practical ceiling for most people.

Is deep work the same as the Pomodoro Technique?

No. Pomodoro is one specific timing pattern, twenty-five minutes of work and a short break, from Francesco Cirillo. Deep work is a description of the kind of attention rather than a schedule, and it often wants longer blocks than a single pomodoro allows.

What is the difference between deep work and shallow work?

Deep work is cognitively demanding and hard for someone else to replicate. Shallow work is logistical and can be done while distracted. Both need doing. The mistake is letting the shallow work set the shape of the day so the deep work has to fit in the gaps.

Can you do deep work in an open office?

Yes, with worse odds. Noise-cancelling headphones, a booked room, or an hour before the office fills up all help more than willpower does. Where you have any say in it, move the block rather than trying to concentrate harder in the same seat.

Should I listen to music while doing deep work?

For most people, instrumental music or nothing is easier to work with than music with words in it, and the same track on repeat stops being interesting quickly, which is the point. This is one to test on yourself rather than take on authority.

Pick tomorrow's first session now, while you still remember what needs doing: one task, one length between 15 and 90 minutes, written down somewhere you will see it in the morning. If you want the timer to sit next to the habits and tasks it belongs to, the guide shows how the pieces connect.