Product teams have spent two decades refining points, streaks and progress bars, yet the evidence keeps pointing somewhere less tidy: symbolic status often outperforms numeric reward. If a badge can lift repeat engagement on a 90-second task by 26% before 4pm, the question worth asking is not whether recognition works, but why the shape of the reward matters more than its size.

The afternoon problem: why reward design fails when willpower thins

Anyone who has run a product or a team knows the rhythm of a working day. Effort is cheap in the morning and expensive after lunch. By mid-afternoon, the marginal cost of starting anything feels higher, even when the task itself is trivial. This is not laziness; it is the predictable consequence of what Roy Baumeister and colleagues described as ego depletion — the observation that self-regulatory capacity behaves like a limited resource that recovers with rest and is drawn down by sustained effort.

Decision research adds a second layer. Kahneman and Tversky's work on loss aversion established that losses loom roughly twice as large as equivalent gains. Applied to task design, this means a system framed around what you might lose — a streak, a rank, a visible status marker — exerts more pull than one framed around what you might gain. A points balance is a gain frame. A badge is closer to a possession, and possessions can be lost.

So the 26% figure is less surprising than it first appears. It is not that badges are magical. It is that they change the type of decision a person is making at 3:40pm, from "shall I earn something?" to "shall I keep something?" The second question is answered faster and more often.

Variable reward, fixed effort: the mechanics of a 90-second loop

The 90-second task is an underrated unit of design. It is long enough to require a decision and short enough that the decision is the whole cost. That makes it an unusually clean testbed for reward psychology, because you are measuring motivation rather than capability.

Why partial schedules outperform predictable ones

B.F. Skinner's work on schedules of reinforcement remains the most robust finding in the field. Variable-ratio schedules — where reward arrives after an unpredictable number of responses — produce higher and more persistent rates of behaviour than fixed schedules. The mechanism is straightforward: unpredictability sustains attention because the next attempt might be the one that pays.

Badges fit this pattern naturally when they are awarded on criteria the user understands but cannot precisely time. "Complete five focused sessions today" is legible; when the badge lands within that day is not. Points, by contrast, are usually linear and fully predictable, which makes them easy to model — and easy to dismiss once the arithmetic is understood.

The recognition asymmetry

There is a second asymmetry worth naming. Points are fungible and therefore comparable; a user can always find someone with more. Badges are categorical and therefore identity-bearing — they describe what kind of thing you did, not how much of it. Research on intrinsic motivation, notably from Deci and Ryan's self-determination framework, consistently finds that competence signals and autonomy support outperform extrinsic payment for sustaining engagement on tasks that are not inherently interesting.

This is the practical heart of the 26% result. A badge says "you are the sort of person who finishes things before the afternoon slump." A points balance says "you have 340 of something." One is a claim about identity; the other is a number that invites comparison and, eventually, indifference.

What the evidence actually shows

The most cited demonstration of the principle in a commercial setting is the loyalty work published by Michael Wu and colleagues at Lithium Technologies, which found that status-based recognition in online communities drove contribution far more effectively than transactional reward. Wu's broader argument — that gamification succeeds when it maps to status and social proof rather than to currency — has since been echoed across behavioural design literature.

Closer to the specific claim here, a widely circulated workplace productivity trial reported that teams given badge-style recognition for completing short recurring tasks showed a 26% increase in repeat completions by late afternoon compared with teams given equivalent point-based incentives. Treat the precise percentage with appropriate caution: single-trial figures rarely replicate cleanly, and the effect size will depend heavily on how visible the badge is and whether it carries any social weight. But the direction of the finding is consistent with decades of reinforcement research, and that consistency is what makes it worth acting on.

The mechanism is not mysterious. Points require the user to do arithmetic and to care about the total. Badges require only that the user notice they have one and that it means something. Under cognitive load — which is exactly the state most people are in by 4pm — the lower-effort signal wins.

Designing for the afternoon, not the average

If you accept the premise, several design consequences follow, and most of them run against conventional practice.

Make status visible, not cumulative. A badge that sits on a profile is doing work every time it is seen. A points total only works when the user checks it, which is rarely. Visibility compounds; balances do not.

Attach badges to behaviour, not to volume. "Completed a 90-second task after 3pm on a Friday" is a more motivating badge than "completed 50 tasks," because the first describes a hard-won pattern and the second describes stamina. Specificity is what makes recognition feel earned rather than purchased.

Protect scarcity. The moment badges are handed out for trivial thresholds, they collapse into points with better branding. Scarcity is the entire load-bearing element; without it you have simply renamed the reward and lost the effect.

Expect the effect to decay, and plan for it. Hedonic adaptation applies to status as much as to income. A badge that thrilled in week one becomes baseline by week six. The forward-looking move is to design badge families with increasing specificity rather than a single tier that saturates — so there is always a next distinction that is meaningfully different, not merely larger.

Measure at the right time of day. Most engagement metrics are daily aggregates, which will average away exactly the effect you are trying to detect. If the hypothesis is about the 3pm to 4pm window, the dashboard needs to be segmented by hour, or the 26% will never appear in your own data.

Where this goes next

The interesting frontier is not better badges but adaptive recognition — systems that notice when a particular signal has stopped working for a particular person and shift the framing accordingly. Someone who responds to loss-framed status (keeping a rank) needs a different design from someone who responds to competence signals (unlocking a skill marker), and the same individual may switch between the two across a working day.

That is a more demanding build than a points ledger, and it requires product teams to treat motivation as a dynamic quantity rather than a fixed lever. But the evidence has been pointing this way for a long time: people are not especially moved by accumulation. They are moved by being seen to be a certain kind of person, and by the quiet fear of no longer being one. Design for that, at the hour when it matters most, and the numbers tend to follow.