There is no best pomodoro length in the sense of a number that beats the others in a controlled trial, because that trial has not been run. What exists is one arbitrary convention, one observational dataset, one rough physiological model, and a lot of people repeating all three as though they were the same kind of evidence.

The four common lengths

Work	Break	Origin	Blocks before a long break
25 min	5 min	Cirillo’s kitchen timer	4
50 min	10 min	Same 5:1 ratio, doubled	2-3
52 min	17 min	DeskTime usage data, 2014	2-3
90 min	20 min	Ultradian rhythm, as a model	2

What the evidence for each actually is

25/5 is arbitrary and Cirillo says so. He was a university student in the late 1980s with a tomato-shaped kitchen timer, trying to study for a short block he could commit to. He has described settling on the length by trying a few and picking one, not by measuring anything. The technique’s mechanisms are the indivisible block and the interruption log, both of which work at any length. The 25 is packaging. More on the method in pomodoro technique 25/5.

52/17 is observational, not experimental. In 2014 the time-tracking company DeskTime looked at its own users’ logs, took the most productive 10% by their own internal measure, and reported that those people averaged about 52 minutes of work followed by about 17 minutes away. That is a correlation in software-usage data from a self-selected population, published by a company that sells time tracking. It is a genuinely interesting observation and it is not a finding that 52 minutes causes productivity. People who are good at their jobs may simply be allowed longer uninterrupted blocks. The number is also an average, which means very few of those users actually worked in 52-minute chunks.

90/20 rests on a rough model. Ultradian rhythms are real: Nathaniel Kleitman’s sleep research described roughly 90-minute cycles, and later work suggested something similar continues through waking hours. The step from “there are cyclical fluctuations in alertness of roughly this period” to “therefore work for exactly 90 minutes and rest for 20” is an inference, not a measurement. Cycle length varies between people and across a day, and nobody is synchronising your work block to the phase of yours. Treat 90/20 as a reasonable shape for deep work rather than a prescription.

50/10 has no origin story at all. It is 25/5 doubled, which is precisely why it is the most sensible default for most knowledge work. Same ratio, half the transitions.

So the honest summary: the ratios all land around 4:1 to 5:1 work to rest, none of them are validated against each other, and the main real variable is how expensive it is for you to re-enter the task.

Choosing by task type

Re-entry cost is the thing to optimise. If it takes you eight minutes to get back into a piece of code, a 25-minute block spends a third of itself on warm-up.

Task	Length	Why
Email, admin, small discrete items	25/5	Near-zero re-entry cost, and the list is the win
Writing, coding, analysis, design	50/10	One warm-up instead of two
Long meetings-free stretch, mixed work	52/17	Longer break genuinely recovers you
Dense reading, proofs, hard debugging	90/20	Some problems need an hour before they open up
Something you have been avoiding a week	10/3	The task is starting, not sustaining

Two other rules worth more than the numbers.

Match the block to the smallest finishable unit. If a task naturally takes 40 minutes, a 25-minute block guarantees you stop mid-thought and a 50-minute block means you finish and have ten spare. Pick the one that ends on a boundary.

Longer blocks need honest breaks. A 90-minute block followed by five minutes of scrolling is worse than 25/5. The break scales with the work or the whole thing degrades by block three.

Presets: 50/10 for general deep work, 52/17 if you want the DeskTime shape, 90/20 for the long version. Or type your own on the pomodoro timer.

How to find yours

Run one length for a full week, not a day. Log two things at the end of each block: whether you finished what you sat down for, and whether you wanted to stop before the beep. Then run a different length for a week.

Wanting to stop early means the block is too long. Consistently working past the beep means it is too short and you are being interrupted at a bad moment. Both are more informative than any study, because they are about you.

For study specifically there is a wrinkle around spacing and recall that the pomodoro framing does not cover, and study timer vs pomodoro gets into it.

Try it: start a pomodoro.