Picture a shared workspace at ten to four. A leaderboard sits on the wall, updated hourly, and the number beside your name has slipped two places since lunch. At four o'clock it comes down. The next task you attempt is short, tightly defined, and requires sustained attention for about ninety seconds. Does the leaderboard's removal make you better at it? The claim in the title — a 19% accuracy gain — is the kind of figure that should make anyone pause, because it bundles together three separate questions: whether the leaderboard was doing harm, whether removing it does good, and whether either effect is large enough to survive contact with a real working day.

What the leaderboard actually does to attention

A public ranking is a feedback device, and feedback is rarely neutral. Research on the audience effect — Norman Triplett's original observation that cyclists ride faster in company, and the later refinement by Zajonc that the presence of others raises arousal and therefore strengthens whatever response is dominant — suggests the mechanism clearly. On a well-learned task, being watched helps. On a task that requires care, checking, or the suppression of a quick wrong answer, arousal tends to push people toward their fastest, most habitual response. That is usually not their most accurate one.

A leaderboard is a permanent audience. It is also a comparative one: the relevant information is not how well you did but how well you did relative to the person above you. That distinction matters. Absolute feedback ("you completed 42 of 45") invites a correction. Relative feedback ("you are 11th") invites a status response, and status responses consume exactly the cognitive resources that a ninety-second accuracy task depends on.

The 19% figure, if it holds, is large enough to be interesting precisely because ninety-second tasks are so short. There is little room for learning, strategy, or fatigue to intervene. What changes between a ranked and an unranked condition is mostly the state the person arrives in — and a 19% swing is roughly what you would expect if a meaningful share of working memory were being spent on something other than the task.

Why the timing matters more than the removal

The detail that does the work in the headline is "at 4pm," not "removed." A leaderboard taken down at the start of the day would be a different intervention entirely. Taken down mid-afternoon, it interrupts an established monitoring habit at the point where self-regulation is already under strain. The removal is not just the absence of a stimulus; it is a small discontinuity that people notice, and noticing is itself a reorientation.

There is a useful parallel in the implementation-intention literature. Peter Gollwitzer's work showed that specifying when and where an action will occur substantially raises follow-through, and much of that effect comes from the cue doing the work rather than conscious effort. A leaderboard is a cue, and a very loud one. Remove it and the cue-driven behaviour — checking, comparing, recalibrating — loses its trigger. The person is left with the task and nothing else to look at.

Loss aversion does not switch off when the board comes down

The most commonly cited explanation for leaderboard anxiety is loss aversion: losses loom larger than equivalent gains, per Kahneman and Tversky's prospect theory. On a ranked board, every refresh is a potential loss. You can only stay where you are or fall. The reference point moves with you, so the frame is permanently tilted toward threat.

But the interesting part is what happens after removal. Loss aversion is not a property of the board; it is a property of the person's reference point, and reference points are sticky. Someone who has spent the morning in eleventh place does not simply become unranked in their own head at four o'clock. They carry a private ranking into the next task. This is the most plausible reason a removal effect would be smaller in practice than in a controlled trial: the intervention removes the display, not the internal comparison.

There is a practical implication here that most workplace experiments miss. If you want the accuracy benefit, the removal has to be legible. People need to know the board is gone and that its absence is deliberate, not a glitch. Otherwise the monitoring behaviour persists and you have removed the information while keeping the vigilance.

The variable-ratio problem

Any discussion of leaderboards and attention has to reckon with variable-ratio reinforcement, the schedule B.F. Skinner identified as the most resistant to extinction. Rewards delivered after an unpredictable number of responses produce steady, persistent behaviour — and, crucially, behaviour that continues long after the reward stops.

A leaderboard is a variable-ratio schedule wearing a spreadsheet. You do not know when your position will change, so you check. Each check is a response; occasionally it pays off with a rise. That is precisely the structure that makes a habit hard to break, and it explains why "we took the board down" so often fails to change anything. The checking behaviour was never really about the board. The board was the slot the behaviour was aimed at.

This is where the 19% claim becomes more interesting than it first appears. If accuracy improves by that margin within ninety seconds of removal, the effect is probably not the extinction of a habit — that takes far longer — but the temporary release of capacity while the habit is suspended by confusion. The person does not know what to check, so for a short window they check nothing. Short windows are exactly what a ninety-second task measures.

What the evidence from elsewhere suggests

The closest well-documented analogue is not in workplace research but in surgical and aviation checklists. Atul Gawande's widely cited work on checklist implementation found that structured prompts improved outcomes most in environments where hierarchy and social comparison were strong, and that a significant part of the benefit came from removing the need for individuals to hold everything in mind simultaneously. The checklist does not make people smarter. It removes a monitoring burden.

The same logic applies here. A leaderboard is a monitoring burden that masquerades as motivation. When it is present, part of the person's attention is permanently allocated to tracking their position; when it is absent, that allocation is freed. The nineteen per cent is best read not as a motivational effect but as a reallocation of a finite resource.

There is a counterpoint worth taking seriously. Competitive framing genuinely does improve performance on simple, well-practised tasks, and removing it can flatten effort in teams where the board was the only visible signal that anyone was paying attention. The effect is task-dependent, and ninety-second accuracy tasks sit at the far end of the spectrum where competition hurts. A ninety-second speed task would likely show the opposite result.

Designing for the ninety-second window

The forward-looking question is not whether to have leaderboards but when they should be visible. Three design moves follow from the mechanics above.

First, decouple ranking from the working window. If comparison is useful, it belongs at the boundary of a work period, not inside it. A board that appears for ten minutes at the end of a session and disappears for the rest gives you the motivational signal without the permanent monitoring load.

Second, make removal explicit and predictable. An unannounced disappearance leaves the vigilance intact. A scheduled one — "rankings are visible until four, then the board comes down for the rest of the afternoon" — converts the removal into a cue in its own right, and cues are what actually change behaviour.

Third, measure the right thing. A 19% accuracy gain on a ninety-second task is a real signal, but it will not show up in monthly output metrics, because it is a within-task effect that gets absorbed by everything else. If you want to know whether removing the board helped, you need short, repeated, tightly-scored tasks — and you need to run them both with and without the board, in the same week, with the same people.

The wider point is that attention is not a motivational quantity. It is a capacity, and it is spent on whatever the environment makes salient. A leaderboard is salient by design. The question worth asking about any such display is not whether it motivates, but what it costs in the ninety seconds after someone looks at it — and whether the look was ever necessary in the first place.