Choice architecture in the liquid flavour shop is rarely treated as a serious design problem. Most operators assume that giving a customer more options is a courtesy, and that a smaller menu is a compromise forced by limited stock. The question worth asking is narrower and more testable: when a returning customer reorders at roughly session 35 — deep into a habit, past novelty, past the exploratory phase — does a two-option flavour wheel actually outperform a seven-option one on completion, satisfaction, and repeat rate? The behavioral evidence on option count and decision quality suggests it might, and the reasons have less to do with taste than with how the choosing mind behaves under repetition.
The paradox of choice, restated for repeat buyers
Iyengar and Lepper's 2000 field study remains the cleanest demonstration of what happens when you widen a menu. Shoppers encountering a display of 24 jams were more likely to stop than those encountering six, but only about 3 percent of the large-display group made a purchase, against roughly 30 percent of the small-display group. The finding is often oversimplified into "fewer options are better," which is not what the data show. Six outperformed twenty-four; it did not follow that two would beat six.
The relevant distinction for a flavour shop is between first purchase and the thirty-fifth. Iyengar and Lepper were measuring acquisition under low prior knowledge. A session-35 reorder is a different cognitive task entirely: the customer already has a stable preference, a history, and a working model of what they want. For that customer, a seven-option wheel is not an invitation to explore — it is a re-decision of something already decided, and every additional option imposes a small cost of re-evaluation. The question is whether that cost is trivial or whether it compounds into abandonment.
There is a useful parallel in the work on choice overload by Chernev, Böckenholt, and Goodman (2015), whose meta-analysis found that the effect is strongly moderated by the chooser's prior preference. Choice overload is most acute when the decision is complex, the options are hard to compare, and the chooser has no established preference. Remove the last condition — as repetition does — and the overload effect shrinks. But it does not vanish, because a repeated menu still forces comparison, and comparison is where reorders stall.
Why two options may dominate at the reorder stage
Two options do something a seven-option wheel cannot: they convert a selection problem into a confirmation problem. A binary choice lets the customer ask "is it this one or the other?" rather than "which of these seven is closest to what I had?" The first question draws on memory and habit; the second draws on working memory and comparison, which are more expensive and more error-prone under time pressure.
Kahneman and Tversky's work on framing and reference points is relevant here, though it is usually applied to risk rather than retail. The core insight is that people evaluate options relative to a reference state, not in absolute terms. A returning customer arrives with a reference: last session's flavour. A seven-option wheel presents six alternatives to that reference, each of which must be weighed against it. A two-option wheel presents one. The reference point does most of the work, and the two-option design lets it.
There is also a reinforcement argument. Variable-ratio schedules — the intermittent reward structure long associated with persistent behavior — produce high and steady response rates, but they require the response to be cheap. If each reorder demands a fresh seven-way comparison, the response cost rises, and the schedule's effectiveness falls. A two-option wheel keeps the response cheap while preserving enough variability (two is not one) that the choice still feels like a choice. The customer is not being told what to buy; they are being asked a question they can answer in two seconds.
The counterargument: seven options and the exploration dividend
None of this argues that seven-option wheels are wrong everywhere. The case for breadth is strongest at acquisition, where novelty is the product and the customer is still building a preference map. A seven-option wheel functions as a sampling instrument: it lets a new customer locate their position in the flavour space, and that positioning has real value downstream.
The risk of a two-option wheel at acquisition is premature narrowing — the customer anchors on the first pair they see and never discovers the flavour that would have become their staple. This is a well-known failure mode in recommendation systems, where aggressive narrowing produces short-term conversion and long-term preference stagnation. The design implication is not "two everywhere" but "two at the point where the preference is already formed." The session-35 threshold is a rough proxy for that point, and it should be treated as a hypothesis to test rather than a law.
There is a second counterargument worth taking seriously: a two-option wheel can feel like a restriction, and restriction can read as disrespect to a customer who has earned the right to browse. The mitigation is framing. If the two options are presented as a curated pair — "these two, based on what you've been choosing" — the design reads as service rather than limitation. If they are presented as the only options available, it reads as a stock problem. The same architecture produces opposite reactions depending on whether the narrowing is explained.
What a session-35 test would actually measure
The claim in the title is empirical, and it should be treated that way. A defensible test would randomize returning customers at the session-35 mark into a two-option wheel and a seven-option wheel, then measure four things: reorder completion rate, time to decision, post-purchase satisfaction, and the rate at which the customer returns for session 36.
The first two are the mechanism; the second two are the outcome. If the two-option wheel raises completion and cuts decision time without depressing satisfaction or the session-36 return rate, the design is doing what the behavioral literature predicts. If satisfaction drops even as completion rises, the narrowing is being experienced as a cost, and the framing needs work before the architecture does.
A concrete precedent worth borrowing from is the menu-design research on restaurant ordering, where smaller, well-curated menus have been shown to reduce decision fatigue and increase average satisfaction — but only when the curation is legible to the diner. The lesson transfers directly: the two-option wheel succeeds or fails on whether the customer can see why those two.
Where this points next
The interesting frontier is not two versus seven as a fixed rule but as a function of session depth. A shop that tracks reorder history can shift the wheel's width dynamically: broad at acquisition, narrow at consolidation, and — this is the part most operators miss — occasionally re-widened to test whether the customer's preference has drifted. Preference is not static, and a two-option wheel applied indefinitely becomes a cage rather than a courtesy.
The next round of testing should therefore ask a sharper question than the one in the title. Not "does two beat seven at session 35," but "what is the optimal width at each stage, and how often should it be re-tested?" The behavioral literature gives a strong prior that narrow beats wide once preference is formed. It gives almost no guidance on when to widen again. That is the open question, and it is answerable with the same randomized design — provided the shop is willing to treat its flavour wheel as an instrument to be calibrated rather than a menu to be filled.