When a streak counter disappeared from a specialty liquid flavour shop's loyalty interface, repeat orders fell 14 percent by the ninth trial run. The number is oddly precise for a phenomenon that most operators treat as folklore. What, exactly, is being lost when a visible tally of consecutive purchases vanishes — and why does its absence show up in order data with such consistency?
The Streak as a Reinforcement Schedule, Not a Game
B.F. Skinner's work on schedules of reinforcement remains the cleanest lens for understanding why a counter changes behaviour at all. A streak is not a reward in itself; it is a conditioned reinforcer — a neutral stimulus (a number) that acquires motivational force because it has been paired with things the buyer actually wants: discounts, free shipping thresholds, the mild social pleasure of being recognized as a regular.
What makes streaks behaviourally interesting is that they function as a variable-ratio schedule wearing a fixed-ratio costume. On paper, "buy ten times, get a reward" looks like fixed-ratio reinforcement, which produces steady, moderate responding followed by a predictable post-reinforcement pause. But real streak systems are rarely that clean. Bonus multipliers appear unpredictably. A shop runs a double-points weekend. A new flavour drops and suddenly the streak is worth more because the reward it unlocks is something the customer wants more than last month's reward. The schedule is nominally fixed; the value of each link in the chain is variable. That hybrid is what produces the persistent, low-level checking behaviour that keeps a customer returning — not daily, but reliably.
Remove the counter and you remove the conditioned reinforcer. The underlying discounts may still exist. The purchases still count toward something. But the buyer can no longer see the chain accumulating, and a chain you cannot see is a chain you do not feel you are breaking.
Loss Aversion Does the Heavy Lifting
Daniel Kahneman and Amos Tversky's central finding — that losses loom roughly twice as large as equivalent gains — explains why streak counters outperform almost any other loyalty mechanic per dollar spent. A customer with a 27-day streak is not primarily motivated by the reward waiting at day 30. They are motivated by the prospect of watching 27 days reset to zero.
This is the asymmetry that makes streaks work. The reward at the end is a gain, and gains are psychologically muted. The streak itself is a possession, and possessions are subject to loss aversion. The moment a customer mentally files the streak under "things I have," every subsequent purchase becomes partly an act of protection rather than acquisition. They are not buying a bottle of flavour concentrate to get something new. They are buying it to avoid losing something they already have.
When the counter is removed, the possession disappears. The customer still accumulates whatever the program offers, but it exists only as an abstraction — a number in a database they never see. Loss aversion has nothing to grip. And because loss aversion was doing most of the motivational work, the drop in repeat orders is disproportionate to the apparent size of the change. A 14 percent decline from removing a visible number is not surprising once you recognize that the number, not the reward, was the engine.
The Ninth Trial Is Where the Effect Stabilizes
The detail that the decline was measured "by trial 9" is worth pausing on. Early trials of any interface change are contaminated by novelty effects and by customers who simply haven't encountered the new state yet. A customer with an active streak who visits the shop on day one of the change may not notice the counter is gone at all — they came to buy, not to check their tally.
By the ninth trial, three things have happened. First, the population of active streak-holders has cycled through at least one full purchase interval, so nearly everyone who would have noticed has noticed. Second, the customers whose streaks were about to hit a milestone have hit the wall of not knowing where they stand, and some have disengaged. Third, the shop's own data has stabilized — the novelty-driven noise has washed out, and what remains is the steady-state behavioural difference.
This is a useful methodological point for anyone running loyalty experiments. A change that looks neutral at trial 3 can look catastrophic at trial 9, because the mechanism being disrupted (streak maintenance) operates on a slower timescale than the measurement window most operators use. The 14 percent figure is credible precisely because it was taken late enough to reflect the mechanism rather than the transition.
What a Liquid Flavour Shop Actually Sells
There is a specific texture to this niche that makes streaks unusually load-bearing. Liquid flavour concentrates are consumables with short replenishment cycles and high variety-seeking. A customer who buys a 30ml bottle of one flavour will finish it in weeks, not months, and will likely want a different profile next time. That variety-seeking is a natural enemy of loyalty programs, because it pushes customers to sample competitors.
A streak counter counters variety-seeking by introducing a cost to switching that has nothing to do with product quality. If buying from a competitor means breaking a 40-purchase chain, the competitor's new flavour has to be meaningfully better to justify the reset. The counter is, functionally, a switching cost made of nothing but a number and a small dose of anticipated regret.
Remove it, and the switching cost evaporates. The customer is free to follow their palate wherever it leads, and the shop's repeat-order rate falls. The 14 percent decline is not customers abandoning the brand in anger. It is customers reverting to their default variety-seeking behaviour once the artificial friction is gone.
Designing the Next Version
The obvious response — put the counter back — misses the more interesting question. Why did it get removed in the first place? Usually because it created support burden, or because a segment of customers found it stressful, or because the engineering cost of maintaining streak state across sessions and devices was higher than the team expected.
Each of those problems has a design answer that preserves the behavioural mechanism. A counter that displays only for customers who have opted in addresses the stress objection. A counter that survives a missed purchase with a "grace day" reduces the anxiety that drives support tickets while preserving loss aversion for the customers who are actually at risk of churning. A counter implemented as a simple server-side integer with a last-purchase timestamp is not an engineering challenge; it is a schema decision.
The forward-looking version of this problem is not whether to show a number. It is how many behavioural mechanisms a shop can afford to remove before its repeat-order rate looks like that of a shop with no loyalty program at all. Streaks, milestone rewards, and personalized recommendations all lean on the same underlying machinery — conditioned reinforcement and loss aversion — and they tend to fail together when that machinery is dismantled piece by piece in the name of simplification. The next experiment worth running is not a removal. It is a measurement: track repeat orders across a cohort that keeps its counter and a cohort that never had one, and see whether the 14 percent gap holds when neither group experiences a transition. That number would tell you what the streak was worth all along, rather than what it cost to lose.