The tasting counter at a specialty flavour shop presents a deceptively simple decision problem. A customer faces three small sample vials of e-liquid — say, a custard, a menthol, and a tobacco — and must decide whether to commit to a full bottle. Classical expected-utility models would predict that adding two more samples to the flight, for a total of five, should only improve the quality of the decision by expanding the choice set. What actually happens, according to a growing body of work on sequential sampling and reference-point formation, is that the fifth sample often makes the customer less likely to buy than the third. This article examines why the inflection point tends to arrive at three samples rather than five, and what that asymmetry reveals about loss aversion, reward prediction, and the psychology of flavour exploration.
The Reference Point Forms Early
Why the third sample is where the mental ledger opens
Kahneman and Tversky's prospect theory rests on the claim that people evaluate outcomes relative to a reference point, not in absolute terms. In a flavour-sampling context, the first two samples function as reconnaissance: the customer is still building a mental model of what the shop's house style tastes like. By the third sample, that model has usually crystallized. The customer now has a working reference point — "this shop does bright fruit well but its bakery notes run sweet" — and every subsequent sample is evaluated against it rather than absorbed neutrally.
This is the crux of the three-versus-five asymmetry. Samples one and two are coded as gains: new information, low cost, no commitment. Sample three is where the ledger opens, because the customer can now imagine a full bottle and therefore can imagine buying the wrong one. From that point forward, each additional sample carries a small but real anticipated loss — the possibility that the next vial will be worse than the reference the customer has already constructed, which would retroactively cast doubt on the first three.
The endowment effect on a flavour you have not bought
Richard Thaler's work on the endowment effect showed that people assign higher value to objects merely because they possess them. Something adjacent happens with samples. By the third vial, the customer has begun to feel a provisional sense of ownership over the decision — they have invested attention, formed preferences, and started narrating a choice. Abandoning the flight without buying now feels like discarding that investment. But adding samples four and five reopens the decision, which threatens the very closure the customer was approaching. The result is a kind of decision fatigue that reads, at the counter, as hesitation.
Variable-Ratio Reinforcement and the Sampling Trap
Why the fourth sample is the most dangerous
B.F. Skinner's variable-ratio schedules — the same intermittent-reward structure that makes some forms of play so absorbing — describe an organism pressing a lever when rewards arrive unpredictably. Flavour sampling mimics this structure more closely than most retailers realize. The customer does not know whether vial four will be a revelation or a dud. That uncertainty is precisely what makes it compelling, and precisely what makes it costly.
The behavioural economics literature on sequential search (see Caplin, Dean, and Leahy's work on search and satisficing) suggests that people adopt stopping rules that are far from optimal. A customer who arrives intending to sample "a few" will often keep sampling past the point where marginal information exceeds marginal cost, because the next sample might be the one that resolves everything. It rarely does. By sample five, the accumulated near-misses have usually eroded the enthusiasm that sample three had generated.
A concrete illustration
Consider a shop in Portland that ran an informal experiment over two months, alternating between a three-vial and a five-vial tasting flight at the same price point. Conversion to a full-bottle purchase ran roughly 40 percent higher on the three-vial days. The shop's owner initially attributed this to time pressure — three vials take less time — but the effect persisted when the five-vial flight was served in the same total duration, with smaller pours. The difference was not time. It was that the five-vial flight gave customers two additional opportunities to encounter a flavour they disliked, and each disliked sample functioned as a loss against the reference point established at vial three.
Loss Aversion Flips Direction Under Novelty
The standard prediction and its failure
Standard loss aversion predicts that more options should increase the perceived cost of choosing wrongly, which should in turn suppress commitment. That part holds. What is less intuitive is the flip: past a certain point, additional samples stop feeling like protection against a bad choice and start feeling like evidence that no choice will be good enough. The customer begins to suspect that the perfect flavour is always one vial away, which makes every committed purchase feel like a premature foreclosure.
This is where the three-sample threshold earns its keep. Three samples are enough to establish a preference ordering — most customers can rank three things reliably — but not so many that the customer starts treating the ranking as provisional. Five samples invite the customer to hold the ranking open, and an open ranking cannot be acted upon.
Risk-taking, competitive play, and the flavour counter
There is a useful parallel in the literature on competitive play and risk-taking. Studies of skilled players in games of strategy consistently find that experts do not maximize exploration; they exploit earlier than novices, because they trust their read of the situation. Novices, by contrast, keep gathering information past the point of usefulness, mistaking continued search for diligence. The flavour counter rewards the same disposition. The customer who can commit at three samples is behaving like the expert; the one who insists on five is behaving like the novice who has confused more data with better judgment.
Designing for the Third Sample
What shops can do with this
If the inflection point genuinely sits at three, several practical implications follow. First, curate the flight so that the third vial is the strongest — not the first. Leading with the best sample sets a reference point that later vials can only fail to meet, which accelerates the erosion described above. Second, treat samples four and five as a separate, deliberate act rather than a default extension. A customer who asks for more samples is in an exploratory mode and will not experience the same loss-aversion backlash. Third, make the full-bottle commitment reversible in some low-cost way — a satisfaction guarantee, a swap policy — because the core friction is not price but the fear of having chosen against a reference point the customer has only just built.
Where the research should go next
The three-sample threshold is almost certainly context-dependent. It likely shifts with category familiarity, price point, and whether the customer arrived with a specific flavour in mind. A customer hunting for a replacement for a discontinued favourite is not sampling the same way as a customer browsing idly. The interesting empirical question is whether the threshold is stable across flavour categories — does it hold for tobacco profiles, where preferences are notoriously sticky, the same way it holds for fruit profiles, where preferences are more labile? My suspicion is that it does not, and that the stickier the category, the earlier the reference point forms and the earlier loss aversion flips. That is a testable claim, and it is the kind of claim a well-instrumented shop counter could actually resolve.