Written by Dr. Leo Maren · Co-founder of Fwocus · Behavioural scientist · PhD, Imperial College London
Thirty-seven is a special number in decision science. It comes from the "37% rule," a result from the mathematics of optimal stopping. In a 100-minute search, the rule says: look for the first 37 minutes and commit to nothing. After that, take the first option that is better than everything you have seen.
This page uses the timer for the look phase. Before you use it, you should know when the rule works, and when it gives bad advice.
Where 37 percent comes from
The rule is the solution to a classic puzzle called the secretary problem. You interview candidates one at a time, in random order. After each one, you must hire or reject. You cannot call a rejected candidate back. Your only goal is to hire the single best one.
The best strategy is to reject roughly the first 37 percent of candidates, whatever their quality. Use them only to learn what "good" looks like. Then hire the first candidate who is better than all of them. The exact fraction is 1/e, about 0.368. Thomas Ferguson's paper Who Solved the Secretary Problem? gives the history.
In real life, you seldom know how many options will come. You often know how much time you have. In 1984, the mathematician F. Thomas Bruss showed that the rule also works in time. If options arrive at random times over a fixed period, you look for the first 37 percent of the period, then choose. This gives at least a 37 percent chance of choosing the best one, even when you do not know how many options will come.
That is why a 37-minute timer fits a 100-minute search window. For other windows, calculate 37 percent:
60 minutes: look for about 22 minutes. Use the 22 minute timer.
90 minutes: look for about 33 minutes. Use the 33 minute timer.
100 minutes: look for 37 minutes, on this page.
2 hours: look for about 44 minutes. Use the + button to add 7 minutes here.
A 37 percent chance is not a guarantee
Most articles about the rule skip this point. The best strategy wins only about 37 percent of the time. In about 63 percent of searches, you do not get the single best option.
Sometimes the best option came during the look phase, and you rejected it. Then nothing later beats it, and the rule tells you to take the last option, whatever it is. That is the price of the rule. It is the best you can do under these strict conditions. It is not a way to be sure.
So use the rule to stop a search that could run forever, not to promise yourself a perfect result.
Check the conditions before you trust the number
The 37 percent answer is correct only for a specific problem. Real searches often break one or more of its conditions.
You cannot go back. In the classic problem, a rejected option is gone. If you can return to an option later, the rule does not apply. You can look at everything and then choose the best. Online shopping often works this way.
You want only the best. The classic problem counts second place as a total loss. If a very good option is almost as good for you as the best one, you should look for less time. Mathematicians have studied this version too. In it, the look phase becomes a much smaller share of the options: about the square root of the number of options, not 37 percent of them.
Options come in random order. If the best options usually appear first, as in a search engine ranked by quality, the look phase wastes your best options.
You can rank each option against the others. You must be able to say "better than all so far" at a glance.
A good fit is a live search in a fixed window: an evening of apartment viewings, a hiring round with a fixed interview schedule, a first-come rental market where good places are gone the same day.
What people do in practice
People tend to stop too early. In laboratory versions of the secretary problem, Seale and Rapoport (1997) found that participants usually stopped searching before the optimal point. Later experiments found the same pattern.
The reason is probably not only impatience. In real searches, every option has a cost in time, effort and attention. When evaluation is expensive, an earlier stop can be reasonable. The laboratory problem has no such cost.
I used the rule when I looked for a flat in London during my PhD. Good flats were gone in a day, and I could not go back to one I had declined. I gave myself a fixed window of viewings and decided before the first viewing where the look phase ended. The difficult part was not the math. It was the second viewing, which was very good, and which the rule told me to decline. I declined it. The flat that I finally chose was better, but I know that I could easily have lost that bet.
When you should not be optimizing at all
The 37 percent rule is a tool for people who want the best option. That goal has a cost. Schwartz and colleagues (2002) compared "maximizers," who try to find the best choice, with "satisficers," who stop at a choice that meets their standard. Maximizers reported more regret and depression, and less happiness and life satisfaction. The study was correlational, so it does not prove that maximizing causes these results.
For small decisions, I now use a simpler plan. I write my standard in the task line before I start, such as "under £1,200, two bedrooms, 30 minutes to work." Then I take the first option that meets it. The timer still helps, because it puts a limit on a search that could take all evening.
How to run the look phase with this timer
Decide the full search window and calculate 37 percent of it.
Write what you will compare in the task line above the digits, for example "rent, light, commute."
Press Start. During the look phase, commit to nothing. Note the best option so far.
When the chime sounds, your benchmark is fixed. Take the first new option that beats it.
The timer does not count the second phase. If you want a limit for the whole window, run the 100 minute timer in a second tab from the start.
Two tests that last exactly 37 minutes
Some visitors need 37 minutes for a simpler reason: a test section. Two admissions tests use this exact length.
ATI TEAS, English and Language Usage. This nursing-school test gives 37 questions in 37 minutes, so one minute per question. Four of the 37 questions are unscored, but you cannot tell which ones.
UCAT, Decision Making. The UK medical and dental admissions test gives 35 questions in 37 minutes, plus separate instruction time.
For both, a single countdown works well if you write checkpoints first. For the TEAS English section, the question number and the minutes used are the same, so check at 27:00 left for question 10 and at 12:00 left for question 25. Do not pause the timer during practice, because the real test does not pause.
Timer controls
Press Space to start, pause and resume. Press R to reset to 37:00. Press F for fullscreen. The − and + buttons, or the Down and Up arrow keys, change the length by one minute.
At zero, one soft singing-bowl chime plays. The sound follows your device volume, and some phones mute web audio in silent mode. The browser tab title shows the time left while the timer runs, and on supported browsers the screen stays awake.
The task line is editable but not saved when you reload. The timer is free, needs no account and saves nothing. The Fwocus app adds backgrounds, ambient sounds, focus stats and streaks.
FAQ
What is the 37% rule?
It is the best strategy for the secretary problem: look at the first 37 percent of options, or of your time, without choosing. Then choose the first option that is better than all of them. It finds the single best option about 37 percent of the time.
Does the 37% rule work for dating?
Only in a loose sense. The math assumes that you cannot go back, that options arrive in random order, and that only the best one counts. Real relationships break most of these conditions, so treat the rule as a reason to explore early, not as a precise instruction.