How many e-liquid options can a person hold in working memory before the act of choosing stops feeling like a reward and starts feeling like a chore? The question is not idle for anyone running a flavour shop, because the answer appears to shift with repetition. A customer facing thirty bottles on a first visit behaves nothing like the same customer on a ninth visit, and the direction of the shift is not intuitive: more choice helps early and hurts late. The pattern is worth examining closely, because it sits at the junction of two literatures that rarely get read together — assortment research in consumer behavior and reinforcement learning in experimental psychology.
The Classic Assortment Finding and Its Hidden Assumption
Iyengar and Lepper's 2000 jam study remains the standard citation: shoppers confronted with 24 varieties were far less likely to buy than shoppers shown six, despite the larger display drawing more initial attention. The mechanism usually invoked is cognitive load — evaluating 24 options consumes attention that could otherwise go toward commitment. But there is a detail in that design that gets glossed over: participants were one-shot visitors. Nobody returned to the table on day nine.
That matters because choice overload is not a stable trait of a product category. It is a function of how many evaluations the chooser has already performed. A flavour shop is a repeated-exposure environment in a way a grocery jam table is not. Regulars return weekly. They are not naive choosers; they are experienced ones, and experience changes the cost structure of comparison.
Why Trial 9 Is Not Arbitrary
Nine is not a magic number, but it is a reasonable marker for the point at which a customer has sampled enough of a menu to have formed stable preferences. By the eighth or ninth visit, most people have a mental shortlist. They know that custard-forward profiles run sweet for them, that certain menthol concentrations read as medicinal, that a particular brand's "blue" is reliably a candy blue rather than a fruit blue. The comparison work that made a 30-option wall exhausting on visit one has largely been amortized.
What remains is a different problem: the wall now competes with a preference the customer already holds. And here the literature on decision-making under uncertainty becomes relevant in a way that the jam study never anticipated.
Variable-Ratio Reinforcement and the Sampling Urge
B.F. Skinner's work on schedules of reinforcement established that behaviour maintained on a variable-ratio schedule — where a reward arrives after an unpredictable number of responses — produces high, steady rates of responding and considerable resistance to extinction. This is usually discussed in contexts involving uncertainty of outcome, but it applies just as cleanly to flavour sampling.
A customer working through a twelve-flavour set is not on a fixed schedule. Some bottles deliver exactly what the description promised. Others are duds. A few are unexpectedly excellent — the third variant of a profile they thought they had already mapped. That unpredictability is what keeps the sampling behaviour alive past the point where a purely rational chooser would stop. If every bottle were reliably good, the customer would settle on the first acceptable one. If every bottle were reliably bad, they would leave. The variance is the engine.
This creates a genuine tension with choice overload. Early on, a large set is a liability because it demands too much evaluation. Later, a large set is an asset precisely because it preserves the possibility of a pleasant surprise. The same twelve options that overwhelmed a novice support continued engagement in a regular.
The Loss-Aversion Complication
Kahneman and Tversky's prospect theory adds a wrinkle. Losses loom larger than equivalent gains, and in a flavour context, the "loss" is a wasted purchase — a bottle that gets abandoned after two tanks. Experienced customers are acutely aware of this asymmetry. They have a graveyard of half-used bottles somewhere.
This predicts something specific: as customers accumulate experience, they should narrow their consideration set, not widen it. And they do, in terms of what they will actually buy. But the menu they want to see does not shrink. They want the full twelve on display even as they purchase from a subset of three. The display serves a different function than the purchase decision — it maintains the option value of surprise while the transaction stays conservative.
What This Means for Set Construction
If the overload curve genuinely inverts with repetition, then a static twelve-flavour set is mismatched to its own customer base. The same wall is doing two jobs at once, and it does neither well.
Segmenting by Tenure, Not Demographics
The more useful segmentation is visit count, not age or device. A first-time customer benefits from a curated six — enough variety to signal range, few enough to permit a decision. A ninth-visit customer benefits from the full twelve, ideally with clear indicators of what is new since their last visit. Novelty flags matter more than breadth for this group, because they have already internalized the breadth.
Some shops already do this implicitly through staff recommendation. The clerk who says "you've probably tried the standard line, here's what came in Tuesday" is performing tenure segmentation manually. Making it explicit in the shelf layout costs almost nothing and removes the guesswork.
Preserving Variance Without Preserving Chaos
The reinforcement argument suggests that the surprise needs to remain genuinely uncertain. If a shop's "new arrivals" section is reliably excellent, it stops functioning as a variable-ratio schedule and becomes a fixed one — and the sampling behaviour it sustains will decay accordingly. This is not an argument for stocking bad product. It is an argument for not over-curating to the point where every option is a safe bet. A menu where everything is a 7 out of 10 is, behaviorally, a menu of one option.
Where the Next Round of Evidence Should Come From
The honest position is that the trial-9 inversion is inferred, not measured. The jam study did not test repeated exposure. The reinforcement literature does not typically use consumer choice sets as its operant. The gap between them is exactly where flavour retail operates, and it is a gap that could be closed with data most shops already possess: purchase histories tied to visit counts.
The tractable question for the next year is not whether choice overload is real — that is settled — but where the crossover sits for a given category. Is it visit nine for e-liquid, visit four for coffee, visit fifteen for craft beer? The answer likely depends on how much comparison each option demands and how much genuine variance exists within the set. A shop that tracks when its regulars stop browsing the full wall and start heading straight for a section has already begun collecting the evidence. The useful move is to treat that pattern as a design input rather than a curiosity, and to let the menu shape itself around the customer's accumulating experience instead of around a single, static ideal of variety.