Dabcity Warehouse

▸ LIQUID FLAVOUR SHOP

▸ Featured ·

Skill-Tier Badges at Three Levels Lift Flavour Reorders 10%

Three-tier skill badges lift flavour reorders 10% by combining reinforcement schedules, competence signalling, and choice under incomplete information

5 MIN READ · 1114 WORDS

The claim is narrow and testable: three tiers of skill badges, no more and no less, produce a 10% lift in flavour reorders. The question worth asking is not whether the number replicates, but why a tier count should matter at all to someone choosing between two bottles of the same profile they bought last month. The answer sits at the junction of reinforcement schedules, competence signalling, and the small anxieties of choosing under incomplete information.

The Tier Count as a Design Variable

A badge is a compressed claim about the buyer. It says: this person has done this before, at this level. When a shop attaches such a claim to a flavour profile rather than to a person, it changes what the profile means. A flavour stops being a product and becomes a record of the buyer's own history with it.

The behavioural literature on goal gradients gives a first handle. Atkinson's work on achievement motivation distinguished approach from avoidance tendencies, and the essential finding is that people persist longest when success is probable but not certain. Two tiers give a binary — you have it or you don't — and binaries saturate. Once the second tier is reached, the gradient flattens and there is nothing left to approach. Five tiers, at the other end, fragment the gradient into steps so small that no single step feels like progress; the earlier finding that two-tier labels outperformed five-tier ones is consistent with this, and so is the observation that two-tier badges beat five-tier badges in the same direction. Three tiers sit between: a first step that is easy, a middle step that requires a repeat, and a top step that requires a pattern.

The 10% figure is modest next to the 16% lift reported for three-tier badges measured at trial twelve, and the gap is informative rather than contradictory. The larger effect appeared on a longer horizon; the smaller one appears sooner and decays less. Tier design is not a single lever but a family of levers with different time constants.

What the Middle Tier Actually Does

The middle tier carries most of the psychological weight, and it is the one most often designed badly. In a three-tier system the middle badge is the first one that cannot be earned on a single purchase. It requires a second order of the same profile within a window the shop defines. That makes it a commitment device rather than a reward.

Commitment devices work because of the gap between intention and action that Kahneman and Tversky's framing research made legible: people weigh losses more heavily than equivalent gains, and they will accept a small certain cost to avoid a larger uncertain one. A middle badge converts an open-ended decision — should I reorder? — into a bounded one — am I two orders from the next tier? The loss framing is doing the work. The buyer is not chasing a gain; they are avoiding the quiet erosion of a status they already hold.

This is where variable-ratio reinforcement gets invoked carelessly. A ratio schedule rewards after an unpredictable number of responses, and it produces high, persistent rates of behaviour. But badges are not ratio schedules. They are fixed-ratio schedules with a visible counter, and the predictability is the point. The unpredictability that matters here is not in the reward timing but in the flavour outcome: will the third profile in the set suit me as well as the first two? That uncertainty is what makes the badge feel like information rather than a coupon.

A concrete illustration from adjacent work: sticker-based badges, which are visually distinct from ordinary UI chrome, lifted repeat orders by 19% by trial six, as reported in the sticker-badge trial. The badge's material form mattered as much as its tier structure. A three-tier system rendered as flat text will underperform a three-tier system rendered as something the buyer can see accumulating.

Risk, Competitive Play, and the Cost of Being Seen

Flavour choice is low-stakes, which is precisely why badge systems can operate on it without triggering defensive reactions. High-stakes domains produce stereotype threat and self-handicapping; people avoid being measured when the measurement might confirm something unflattering. A flavour profile carries no such threat. The badge is a claim about taste, and taste is forgiving.

Competitive play introduces a second mechanism: social comparison. Tiered badges make a buyer's position legible relative to an implicit standard, and the middle tier is where comparison bites hardest, because the top tier is remote and the bottom tier is universal. This is the same dynamic that makes ranked shortlists effective when they are short — three-item ranked shortlists outperformed ten-item ones for first reorders, and the reason is cognitive rather than motivational: three options can be held in working memory and compared, ten cannot.

There is a risk worth naming. Tier systems reward the buyer who already repeats. If the badges are calibrated so that the top tier is reachable only by a small fraction, the middle tier becomes the effective ceiling and the system stabilises at two tiers in practice. The design lesson from the confidence-badge work, where two-tier confidence badges lifted cart adds by 13%, is that the label's meaning must be legible without explanation. A third tier that requires a paragraph of copy to justify itself is not a tier; it is a footnote.

Where the 10% Comes From, Mechanically

Decompose the lift. Some of it is frequency: badge holders order slightly more often. Some is concentration: they order the same profile rather than rotating, which raises the reorder metric without raising total volume. Some is selection: buyers who would have reordered anyway now do so through the badged path, and the measurement counts them.

Only the first component is genuinely new behaviour, and it is the smallest. This is not a reason to dismiss the effect; it is a reason to instrument it properly. A shop that reports a 10% reorder lift without separating frequency from concentration from selection is reporting a number it cannot act on. The forward-looking move is to treat the tier count as one parameter in a small design space — tier number, tier spacing, badge materiality, window length — and to run the space rather than the single point. The next useful experiment is not another three-tier trial but a spacing trial: does the middle tier bind earlier or later when the second-order window is fourteen days rather than thirty? That question has a clean answer, and it is the one the current evidence leaves open.