You’re flat-hunting. You’ve seen six so far. The first one was alright but the light was wrong. The second one was expensive but had a view. The third one had mould. The fourth and fifth were forgettable. The sixth one is actually pretty good — good enough that you’re tempted to sign — but there are four more booked in on Saturday, and what if one of those is better? What if the perfect flat is still ahead?
You hesitate. You ask to view one more on Sunday. You see it; it’s mediocre. You try to go back to the sixth, but it’s been taken. You start the search again.
This is an almost universal experience, and it has a name in decision theory. It’s called the secretary problem, or the optimal stopping problem, and it has an elegant, specific, and counter-intuitive answer.
The maths of knowing when enough is enough
The problem was first formalised in the 1960s by the mathematician Martin Gardner in his Scientific American column. The set-up: you’re interviewing a sequence of candidates for a job. Each candidate must be either hired or rejected on the spot — no going back. You don’t know in advance how good the pool will be. How do you maximise the chance of picking the best candidate?
The mathematical answer is surprisingly clean. You should observe — without hiring — roughly the first 37 per cent of the candidates, tracking who was best among them. Then, from the 37 per cent mark onwards, you should hire the first candidate who is better than everyone you’ve seen so far. If no one surpasses the best of the first 37 per cent, you hire the last candidate.
Under the assumptions of the classical problem, this strategy maximises your probability of selecting the genuinely best candidate. The number 37 per cent comes from the mathematics of the exponential constant e: one divided by e, to be exact.
The problem has been applied, with varying degrees of looseness, to flat-hunting, job-hunting, partner-hunting, and almost any serial search for a good-enough-or-better option. Pop-psychology versions of the rule suggest that if you plan to date through your twenties, you should essentially use your early twenties as information-gathering and then commit to the first person past about age 27 who exceeds your early sample. This is, obviously, more of a playful suggestion than an instruction, but it captures something real about a pattern that turns up across domains.
Why real-world searching is harder than the maths
The actual problem is, of course, messier than the textbook version. The maths assumes you’re evaluating candidates on a single scale, that you know when you’ve seen them all, that you can’t return to previous options, and that you have no external information about what the typical distribution looks like. In reality, none of these hold cleanly. You can sometimes go back to earlier options. You don’t know the length of the sequence. Candidates vary along many dimensions, not one. And you usually have prior information about the market that the pure model can’t use.
More importantly, the clean version of the problem assumes you want the very best. But in practice, the difference between the best option and the third-best option in almost any search is small, while the cost of continuing to search can be substantial. This brings us to a second tradition in decision research, which arrived at a different conclusion.
Satisficing, and why good-enough often wins
The Nobel laureate Herbert Simon coined the term satisficing — a blend of satisfy and suffice — in the 1950s. Simon’s argument was that real decision-makers, faced with genuine uncertainty and limited time, don’t try to find the optimal option. They set a threshold for what would be acceptable, and commit to the first option that clears it. This is not laziness. It’s a rational response to the fact that continued searching has real costs: time, attention, the stress of prolonged uncertainty, and the opportunity cost of not acting on what you already have.
Simon’s work has been strengthened over the decades by research showing that in many practical contexts, the difference between the outcome of satisficing and the outcome of exhaustive optimisation is tiny — while the cost of optimising is large. You spend three extra weeks looking at flats. You find one that’s 5 per cent better than the one you would have signed. You’ve paid for those three weeks in rent, stress, and a worse emotional relationship with the whole decision. The arithmetic rarely favours the continued search.
The psychologist Barry Schwartz, in his influential book The Paradox of Choice, extended this into an argument about life satisfaction. Schwartz distinguishes maximisers (who try to find the best option across every decision) from satisficers (who settle for good-enough). His finding: maximisers tend to make slightly better objective decisions on average — better-paid jobs, better-educated partners, better-scoring products — but report significantly lower satisfaction with those decisions. They’re haunted by the option they didn’t choose. Satisficers, making slightly worse objective choices, report higher satisfaction and lower anxiety across almost every domain of life.
The jam experiment
The most famous illustration of Schwartz’s argument came from a study by Sheena Iyengar and Mark Lepper, who set up a tasting booth at a specialty food store in California. On some days, they offered 24 varieties of jam to taste. On other days, only 6. The larger display attracted more attention — people gathered, tried more samples, seemed more interested. But far fewer of them actually bought anything. The smaller display produced about ten times as many sales.
The interpretation: more options generated more initial interest but also more cognitive load, more sense of tradeoff, more anxiety about the option not chosen, and ultimately more paralysis. Somewhere around six options seemed to be the sweet spot. Beyond that, more was worse — even though the extra options were genuinely giving customers more of what they wanted to choose between.
Iyengar and Lepper’s finding has been replicated, with some variation, across many other domains. It has implications for how people shop, hire, date, choose universities, and select investments. The abundance of options in modern life is often experienced, paradoxically, as a reduction in freedom rather than an expansion of it.
The counter-view worth hearing
A significant counter-tradition, led by the psychologist Gerd Gigerenzer, argues that simple heuristics often outperform optimisation in real-world conditions. His research on what he calls take-the-best reasoning — make your decision based on the single most important criterion, ignore the rest — has shown that in many practical contexts, simple rules beat complex models, partly because they’re more robust to noise and partly because they’re faster.
Applied to searching, Gigerenzer’s view would suggest something like this: don’t try to find the optimum. Don’t even try to satisfice across many dimensions. Identify the one or two things that actually matter to you in the decision, and pick the option that best satisfies those, ignoring everything else. If you’re searching for a flat, and the thing that actually matters is location plus natural light, don’t let the fancy kitchen or the second bathroom distract you. Your best flat is the one that wins on the criteria you chose. Everything else is noise.
This approach has the virtue of producing faster, less anxious decisions, and — Gigerenzer argues — often better outcomes. His research at the Max Planck Institute in Berlin has repeatedly shown simple heuristics outperforming more sophisticated optimisation methods on real-world decision problems, particularly when the environment is uncertain.
Permission to commit
Put all these traditions together and what emerges is something more useful than any of them alone.
For most serial search problems, you probably should do some early looking — enough to get a sense of the distribution, to calibrate your sense of what’s possible, to learn what the market actually looks like. The 37 per cent rule isn’t literal, but the underlying idea (spend some time learning before committing) is sound.
After that, you should set a reasonable threshold and commit to the first option that clears it. You should not keep looking past that point in pursuit of the mythical perfect match. The perfect match, if it exists, is one option in many thousands, and the probability of distinguishing it from the merely good is low — while the cost of prolonged searching is high and often irreversible.
You should focus on the two or three dimensions that actually matter to you, and ignore the rest. You should be a satisficer rather than a maximiser, particularly for decisions you’re going to live with rather than measure. And, perhaps most importantly, once you’ve committed, you should stop comparing. The option you didn’t choose is not competing with the option you did. The two exist in separate worlds now, and the worst thing you can do is keep holding them side by side.
The question that remains
The deepest message of the research on searching is, oddly, less about searching and more about stopping. Almost everyone can find a good-enough option. Almost nobody can stop looking once they’ve found one. And the skill of a settled, satisfied life — in flats, in careers, in relationships, in a thousand small daily choices — is the unglamorous, unphotographed, unsung skill of closing the search and moving on.
The question worth asking yourself, the next time you find yourself compulsively comparing an option that’s already clearly good to imaginary options that haven’t arrived:
If I committed to this now, what would I actually lose — and would any of what I’d lose be worth the weeks of continued looking? For most searches, the honest answer is: not much, and no.
Key research referenced: Martin Gardner’s presentation of the secretary problem (1960); Herbert Simon on satisficing and bounded rationality; Barry Schwartz, The Paradox of Choice (2004); Sheena Iyengar and Mark Lepper’s jam-tasting experiment (Journal of Personality and Social Psychology, 2000); Gerd Gigerenzer’s research on take-the-best heuristics.