Knowing when to stop looking is a separate decision from what to pick

The problem with an open-ended search

When options arrive one at a time and disappear if refused — flats, job candidates, second-hand instruments, places to live — the decision has two parts that get confused. One is judging whether the option in front of you is good. The other is deciding whether to keep looking at all. The second question is the one that determines the outcome, and it is almost never answered explicitly.

Without a stopping rule, search ends for the wrong reasons: exhaustion, an arbitrary deadline, or a run of poor options that makes an average one look excellent. None of those is related to whether the remaining search would have been worth its cost, which is the only thing that should end it.

What the idealised version says

There is a clean mathematical treatment of this, usually posed as choosing the best candidate from a sequence you must accept or reject on the spot, with no returns. The optimal policy has a distinctive shape: look at a fixed initial fraction of the candidates without accepting any of them, note the best you saw, then take the first subsequent candidate who beats it. The classic result puts that observation phase at roughly the first third of the sequence.

The assumptions behind that answer are strict, and they matter more than the number. It assumes you know how many candidates there are, that you can rank any two but have no absolute scale, that rejected options are gone for good, and that only getting the very best counts as success. Change the last assumption to something sane — that a near-best outcome is nearly as good — and the optimal look-first phase shortens considerably.

So the model does not hand you a rule for renting a flat. What it hands you is the structure: a deliberate calibration phase in which you are not buying, followed by a threshold you commit to. That structure survives the loosening of the assumptions even when the specific fraction does not.

Calibration is what the early phase is for

The reason not to accept anything early is not superstition about waiting. It is that you do not yet know the distribution. Until you have seen a handful of options, you cannot tell whether an asking price is high, whether a feature is standard, or whether the thing you were told is unusual is in fact universal. Any judgement made before that is a judgement against an imagined market.

This also explains a common regret: the person who takes the first option and then spends the next fortnight discovering what was available. Nothing was wrong with the first option. What was missing was any basis for knowing that.

Setting a threshold you will actually honour

The threshold has to be written down before the options get emotional, and it should be expressed in terms you can check on the day rather than in adjectives. Two or three requirements that would genuinely make you decline, plus a rough level of overall quality drawn from what you saw in the calibration phase, is enough. More criteria than that will not be applied under pressure.

Then the rule is unromantic: take the first option that clears the bar. Not the best of everything remaining, which you can’t see, and not the one that arrives after another fortnight of looking, which costs you a fortnight and often the earlier option too.

The costs people leave out

Search consumes time, attention and goodwill, and in competitive markets it consumes the options themselves while you deliberate. It also degrades judgement: a long search tends to raise standards as you accumulate a composite ideal assembled from the best feature of each option seen, none of which exists in one place.

There is a further cost that only shows up afterwards. Extended searching increases the salience of the alternatives you rejected, which is the raw material for regret. This is one of the more plausible readings of the mixed literature on having many options — the difficulty is less in choosing than in living with the choice once a large field of foregone alternatives is vividly in mind.

When to keep looking anyway

The rule flips when the options don’t disappear, when they arrive quickly and cheaply, or when the difference between good and best is large relative to the search cost. Buying a commodity item online has almost no stopping problem, because nothing is lost by looking and the alternatives remain available.

The genuine stopping problems have three features together: options expire, looking is expensive, and quality varies enough to matter. When you notice all three, stop optimising the choice and start writing the rule, because the rule is where the outcome is actually decided.