How long should a focus session be?

The honest answer

Somewhere between 25 and 90 minutes, depending on the work and on you. That sounds like a dodge, but it is the only answer the evidence supports. The question hides an assumption worth rejecting: that there is one magic number, and that productivity comes from finding it. There is no magic number. There are shallow tasks that finish in twenty minutes, deep tasks that barely warm up in forty, and a personal rhythm you can only learn by paying attention.

What you can do is stop picking a length at random. Match the block to the work, protect the block from interruption, and end it on purpose instead of letting an alarm decide. The rest of this post is about how to do that.

Where the famous numbers come from

Most of the numbers people quote are conventions, not findings. The 25-minute Pomodoro started as one student's kitchen timer in the late 1980s; Francesco Cirillo picked a duration that fit the study sessions of a distracted undergraduate, and it stuck because a small number feels easy to say yes to. The 52-and-17 rule came from one app's data on its own most productive users, and its maker has since revised the numbers; it is an observation about one sample, not a law of nature. The 90-minute claim traces to real physiology, biological cycles observed in sleeping infants; the few direct tests of that rhythm in waking adults found nothing governing concentration.

None of this makes these numbers useless. A default you can start without negotiating is genuinely valuable, and 25 minutes is a fine default. Many people who keep the habit treat it the same way: start the 25, and if the work is moving, keep going; the timer is a floor, not a ceiling. The mistake is treating the default as the answer, so that a session ends because the timer rang rather than because the work reached a natural edge.

The floor: what interruption costs

Whatever length you choose, it has to clear a hard minimum: the cost of being interrupted. There is a research floor under any session length: the interruption research found people need about 23 minutes to fully return to a task after an interruption, so a 15-minute block barely outlasts one distraction. It is not that 15 minutes is too little time to do anything; it is that one knock on the door, one notification you glanced at, can consume the entire block in recovery. If your real environment interrupts you often, the fix is longer blocks with fewer openings, not shorter blocks that fit inside the interruptions.

Length should match the work

The deepest problem with one fixed number is that tasks differ in how long they take to get into. Batching email has almost no warm-up; you are productive in the first minute, so a short block wastes nothing. Writing, coding, or thinking through one hard problem has real ramp-up: the first stretch is spent reloading context, and the good part starts once the model of the problem is fully in your head. Creative and design work often needs the longest stretch of all, because the first half of the block is spent seeing the problem at all.

So size the block to the task. Roughly: 15 to 25 minutes for shallow batches, 50 to 60 for one sharp problem, 90 or more for building and making. Length follows the intention, not the timer: the one intention per day practice already assumes this, working in uninterrupted stretches of roughly 25 to 90 minutes depending on the work. If you chose one meaningful thing this morning, the question how long should a session be becomes much easier, because the work itself tells you.

What a long block actually feels like

The complaint about long sessions is that they feel impossible, and that complaint is usually about the first ten minutes. Getting into demanding work feels like wading into cold water; the urge to check something is strongest right at the start, when nothing interesting has happened yet. Sit through it once and you learn the shape of it: discomfort, then gradual absorption, sometimes the merged-attention state psychologists call flow, which arrives on its own schedule and cannot be talked into arriving sooner.

This is the other reason short defaults quietly cost you. A 25-minute cap means you exit at the exact moment the water stops being cold. If you have never worked a 60 or 90-minute block, try one this week on a task you care about, late enough in the day that nobody expects a reply from you. The point is not to suffer. The point is to feel where your attention actually wants to end, which is information no default can give you.

One ceiling is worth respecting, though. Even the strongest advocates of long blocks cap deliberate depth at around three or four hours a day; past that, the hours stop being deep and start being performed. Length is a tool, not a virtue, and a third 90-minute block is not twice as good as two honest ones.

Why Focus uses soft boundaries

When I built focus blocks into Focus, I kept the length adjustable and made the boundary soft. A block in Focus does not end with an alarm; it ends with a wind-down: the timer eases, the screen asks what you were in the middle of, and you choose to continue or close. The point of a boundary is to protect the session from drifting forever, not to chop your attention mid-thought because a round number arrived.

The countdown-as-drill-sergeant model has it backwards. If the timer is allowed to interrupt you, the timer is the interruption. I write about these choices as I make them in the public build journey; the soft-boundary block is one of them.

Try this week

Pick one task you keep not finishing. Give it one block, sized to the task rather than to a habit: 25 if it is shallow, 60 if it is one hard problem, 90 if it is something you are building. Phone in another room, notifications off, one opening. When the block ends, write one line about where you were when it ended: past the point of wanting to stop, or relieved to stop.

Do that three times and you will have something no article can hand you: your own number, for your own work, on your own evidence. Then the famous numbers become what they should have been all along, starting points instead of rules.