Every apartment hunt arrives at the same moment. You are standing in a place that is pretty good, the leasing agent says someone else is viewing it this afternoon, and you have no idea whether the next listing will be better. Sign now and you might regret it next week. Keep looking and this one is gone.
That moment has a name in mathematics, and an answer. The 37% rule is a strategy for choosing the best option from a sequence you see one at a time: spend the first 37% of your search looking without committing to anything, then take the first option that beats everything you saw in that look phase. For a ten-apartment search, that means touring three places purely to learn the market, then signing the first one that beats all three.
It is a real result, and it is also one of the most misunderstood rules on the internet. In a ten-viewing search it quietly fails three times in ten, and in a long search more than a third of the time. It assumes things no rental market does. And the version that works best for an actual renter is not the famous one. What follows is the machinery underneath it, so you can tell which version you are actually playing.
One thing to settle first, because it decides whether any of this works: the look phase only counts if you remember it. By the fourth viewing, it is easy to lose track of whether the second place had the bigger kitchen or the longer commute, and a benchmark you cannot recall is no benchmark. The Apartment Search Tool is built for exactly this: it logs every community, tour and review and scores each place the same way, so the bar your look phase set is written down rather than half-remembered. The method below works with a notebook too. It does not work with nothing.
What is the 37% rule?
The 37% rule is the solution to what mathematicians call the secretary problem, the best-known example of an optimal stopping problem: deciding when to stop searching and commit, when every option you pass up is gone for good. It first appeared in print in Martin Gardner’s Mathematical Games column in Scientific American in February 1960, and it has since grown into a whole field of probability research, as statistician Thomas Ferguson recounts in Who Solved the Secretary Problem? (opens in new tab) (Statistical Science, 1989).
The rule has two phases:
- Look. See the first 37% of your options. Commit to none of them, however good they are. Their only job is to set the bar.
- Leap. From then on, take the first option that is better than everything in the look phase.
Ferguson’s paper gives the result: for a long search, the best cutoff is a fraction 1/e of the way through, about 36.8%, and following the rule finds the single best option with a probability of about 36.8% too. That second number is the part that is easy to miss. The 37% rule is the best possible strategy for this problem, and it still misses the best option nearly two times in three.
How many apartments to see before you commit
Apply the rule to whatever unit your search is actually limited by: viewings if you know roughly how many places you will see, days if you have a hard move-out date.
By days, it is simple arithmetic. With a month before your lease ends, spend the first 11 days looking (30 × 0.37 = 11.1), then sign the first place that beats everything from those 11 days.
By viewings, the 37% is only an approximation for small numbers. We computed the exact best look phase for searches of different sizes using the formula in Ferguson’s paper:
| Viewings planned | Look at (commit to none) | Then leap | Chance you get the best one |
|---|---|---|---|
| 5 | 2 | from viewing 3 | 43% |
| 8 | 3 | from viewing 4 | 41% |
| 10 | 3 | from viewing 4 | 40% |
| 12 | 4 | from viewing 5 | 40% |
| 15 | 5 | from viewing 6 | 39% |
| 20 | 7 | from viewing 8 | 38% |
Two things stand out. For searches of ten or more, the look phase is a little under 37%: for ten viewings it is three, not four. And the success rate is slightly better than 37% for small searches, sliding toward 37% as the search gets longer.
The six assumptions the 37% rule is built on
The rule is only optimal under the conditions it was solved for. Ferguson lists six. Here they are, translated into apartment hunting, with how well each one holds in a real search:
- One position available. You are signing one lease. This holds.
- The number of options is known. You know you will see ten places. This roughly holds. Most people can estimate their viewings or their deadline.
- Options arrive in random order. This mostly holds. Listings appear when landlords post them, not in order of quality.
- You can only rank each option against the ones you have seen. You can say “better than the last three” but not “top 5% of the market.” This is the one you can change, and changing it is the biggest improvement available (see below).
- A rejected option cannot be recalled. This partly holds. Good places do get leased to someone else, but sometimes a place you passed on is still available a week later.
- Only the very best counts. Second best scores the same as worst. This almost never matches what a renter wants. Many renters would be happy with any of the top three.
Every assumption that fails changes the right cutoff. The next three sections deal with the ones that matter most.
The 30% problem: what happens when the best place was in your look phase
If the single best apartment shows up during your look phase, nothing afterwards can beat it. The strict rule then never fires, and you are left with whatever you see last, which on average is a middling apartment.
That happens more often than people expect. With a look phase of three out of ten, the best place lands in those first three 30% of the time, simply because it is equally likely to be any of the ten.
Assumption 6 makes it worse. The classical rule is tuned to maximize your chance of the single best place, and it accepts a lot of bad outcomes to get there. If what you actually want is a top-three apartment, a shorter look phase does better. We simulated 200,000 ten-viewing searches for each cutoff.
| Look at first | Get the best one | Get a top-3 one | Stuck with the last one |
|---|---|---|---|
| 1 | 28% | 62% | 10% |
| 2 | 37% | 71% | 20% |
| 3 (the 37% rule) | 40% | 71% | 30% |
| 4 | 40% | 67% | 40% |
| 5 | 37% | 61% | 50% |
Look at two instead of three and your chance of a top-three apartment stays at 71%, while your chance of being stuck with the last place drops from 30% to 20%. You give up three points of “the very best” to get that. For most renters, that is a good trade.
Whatever cutoff you choose, set a fallback before you start: the date by which you will sign the best place still available, whatever it is. The rule tells you when to stop looking; the fallback tells you what to do when the rule never fires.
How to adjust the 37% rule when apartments get snapped up
In a fast rental market, deciding to take a place does not mean you get it. Someone else’s application gets in first, or the landlord picks another tenant. That breaks a hidden assumption: in the classical problem, the option you choose always says yes.
We simulated the same ten-viewing search with a 50% chance that any place you decide to take is already gone. If it is gone, you keep searching, and the place you lost becomes the new bar:
| Look at first | Get the best one | Get a top-3 one |
|---|---|---|
| 1 | 27% | 57% |
| 2 (best cutoff) | 30% | 58% |
| 3 | 29% | 55% |
| 4 | 28% | 51% |
The best cutoff shrinks from three to two, and every outcome gets worse. The lesson generalizes: the more often good places disappear before you can act, the shorter your look phase should be. If you can apply the same day you view, use the classical cutoff. If you are always one application behind, look less.
This is also the honest case for being ready before the leap phase starts: documents gathered, deposit available, references lined up. Every hour of delay makes it likelier that the place is gone when you act; the simulation above held that chance at 50%.
The better game: score every place on a fixed scale
The biggest improvement comes from breaking assumption 4. The classical rule assumes you can only say “better or worse than what I have seen.” If you can instead score each place on a fixed scale, you are playing a different, easier game.
Mathematicians call this the full-information version of the problem. Ferguson’s paper reports that the best strategy for it, worked out by Gilbert and Mosteller in 1966, finds the single best option with a probability of at least 58% however long the search, against 37% for the classical rule. You can do much better simply by knowing how good a place is, not just whether it beats the last one.
In our computation, the full-information strategy has no pure look phase. It uses a bar that drops as your options run out: early on you only take something exceptional, and late in the search you accept something merely good. We computed those bars for a ten-viewing search and simulated it:
| Viewing | Take it if it beats at least |
|---|---|
| 1 to 5 | 86 to 92% of comparable listings |
| 6 to 7 | 78 to 82% |
| 8 | 69% |
| 9 | 50% |
| 10 | Take it (it is the last one) |
In our simulation, that strategy found the best of ten places 61% of the time and a top-three place 85% of the time, against 40% and 71% for the classical rule.
You will never know exactly what share of the market a place beats. But you can get much closer than “better than the last one,” and that is where a scorecard earns its place. Weight what matters to you (rent, commute minutes, light, laundry, the landlord’s reviews), score each place against the same criteria, and your look phase stops being a vague impression. It becomes a set of numbers you can set a bar against. If you have never built one, weighted decision scoring explains the method, and our guide to how to choose between two apartments runs it in five minutes for a final pair. Score what you see on each visit with the apartment walk-through checklist, so every place is judged on the same items.
A worked ten-viewing search
Here is how that plays out, with illustrative scores out of 100. The renter plans ten viewings over three weeks, can apply the same day, and wants a top-three place more than the single best one, so they use a look phase of two.
| Viewing | Score | Decision |
|---|---|---|
| 1 | 71 | Look phase: note it, do not commit |
| 2 | 78 | Look phase: the bar is now 78 |
| 3 | 74 | Below the bar, pass |
| 4 | 69 | Below the bar, pass |
| 5 | 83 | Beats 78: apply today |
Viewing five is the leap. Because the renter wrote down scores rather than impressions, they could see within minutes that 83 cleared the bar. Without them, it would have been a gut call against two half-remembered places.
Does the 37% rule work for house hunting?
Yes, with two adjustments. If your house search is shorter, say five serious viewings, the look phase is smaller: two of five, per the table above. And if offers in your market are often outbid or turned down, the snapped-up adjustment applies with more force. A shorter look phase is again the right response.
The scoring advice carries over unchanged, and matters more with a bigger price tag. The House Search Tool does for viewings what the apartment tool does for rentals. If the question is where to live rather than which house, the Neighborhood & Relocation Decision Helper scores whole places, with a weight column for each person in the household.
Common questions about the 37% rule
Is the 37% rule really the best way to choose an apartment?
It is the best strategy for a narrow version of the problem: you see places one at a time, you can never go back, you can only rank them against each other, and you only count finding the single best one as a win. Real apartment hunting usually breaks two or three of those conditions, and each break moves the best cutoff. Treat 37% as a rough guide for your look phase, not a target.
Should I count the 37% in days or in viewings?
Either works, as long as you use the one you actually know. If you know you will see about ten places, count viewings. If you know you have a month before your lease ends, count days and look for the first 11.
What if every place after my look phase is worse?
Then the best place was in your look phase, and the strict rule leaves you with whatever you see last. Set a fallback before you start: the date by which you sign the best place still available, whatever it is.
The short version
- Look, then leap. Spend the first part of your search setting a bar, then take the first place that clears it.
- Look at less than 37% in a real market. Two of ten viewings keeps your top-three odds and cuts the chance of being stuck with the last place.
- Shorten the look phase further when good places vanish fast.
- Score every place on the same scale. It moves you from the 40% game toward the 61% one.
The math rewards whoever remembers their look phase most precisely. The Apartment Search Tool was designed for exactly that: one file you own, holding every place you toured, scored the same way, with the bar written down before the agent says someone else is viewing it this afternoon.
