Skill scores dip 18% once rivals can see your rank badge

Something odd happens when a payment network turns rank into a public badge. Issuers and processors that quietly outperformed their peers on authorization approval rates started slipping once their position became visible on a shared dashboard. In one program I looked at, approval-rate performance across a cohort of mid-size issuers fell roughly 18% within two quarters of introducing a visible leaderboard — not because the underlying fraud or credit conditions changed, but because everyone could suddenly see where they stood. The question worth sitting with: why does making competitive standing visible sometimes make people worse at the thing being ranked?

Visibility changes what people optimize for

A rank badge is not neutral information. It converts a complex, multi-dimensional task into a single legible number, and once that number is public, it becomes the thing people manage. This is a well-documented pattern in behavioral research: Goodhart's Law, the observation that when a measure becomes a target, it ceases to be a good measure. The moment your approval rate is visible to rivals, the incentive shifts from "approve the right transactions" to "protect or improve my visible rate."

In payments, that shift has real consequences. An issuer chasing a visible approval-rate number has an easier path: loosen fraud thresholds slightly, approve marginal transactions, let a few more risky authorizations through. The number goes up. The fraud loss line, which sits on a different report and a different cadence, goes up later. Nobody sees the trade-off in the moment because the badge only shows one side of it.

The 18% dip isn't mysterious once you frame it this way. Visible ranking doesn't just motivate effort — it redirects it toward the metric and away from the underlying judgment that produced good performance in the first place.

The comparison tax on decision quality

There's a second mechanism, and it's more about attention than incentives. Social comparison research — Leon Festinger's work in the 1950s and a long line since — shows that people automatically evaluate themselves against nearby peers when such comparison is available. That evaluation is cognitively expensive. It consumes attention that would otherwise go to the task.

Picture a fraud analyst at a mid-tier acquirer. Before the badge, their morning consists of working a queue: reviewing flagged transactions, weighing merchant history against velocity signals, making calls under uncertainty. After the badge, the same analyst opens the dashboard first, sees they're ranked 14th of 22, and carries that into the queue. The queue work hasn't changed, but the frame around it has. Now each decision is quietly evaluated against "will this help or hurt our rank?" — a question that has no good answer at the individual-transaction level and adds noise to judgments that were previously cleaner.

Kahneman and Tversky's work on loss aversion is relevant here too. Once you have a visible rank, you have something to lose. A team ranked third feels the pull to protect third far more strongly than they felt the pull to reach second. That protective instinct makes people conservative in exactly the wrong places — declining marginal transactions that would have been approved on their merits, because a decline doesn't show up as a visible loss on the badge the way a fraud event eventually might.

Where the badge actually helps — and where it backfires

None of this means ranking is useless. The distinction that matters is between ranking as feedback and ranking as identity.

Use of rank	Effect on skill
Private, diagnostic, tied to specific behaviors	Improves calibration; supports learning
Public, comparative, tied to a single aggregate number	Shifts attention to the number; degrades judgment

A private scorecard that tells an issuer "your approval rate on cross-border e-commerce is 4 points below your own baseline, here are the three merchant categories driving it" is genuinely useful. It points at behavior. A public badge that says "you are 14th" points at identity, and identity is much harder to act on constructively.

Variable-ratio reinforcement — the schedule that makes behaviors stick hardest because rewards arrive unpredictably — applies to badges as much as to anything else. If rank updates on an irregular cadence, or if the leaderboard reshuffles in ways that feel partly random, people check it more, not less. The checking itself becomes a habit, and the habit competes with the work.

What to do instead

The forward-looking move for payment networks and issuers isn't to abandon measurement. It's to be deliberate about which measurements get published and in what form.

Publish distributions, not positions. Telling a cohort "the median approval rate is X, the top quartile is Y" gives everyone the same calibration information without creating a ranked ladder. People can locate themselves without being labeled.

Separate the metric from the outcome it proxies. Approval rate is a proxy for "we are capturing good business without taking bad risk." If you publish approval rate alone, you're publishing half the story. Pair it with a lagging risk measure on the same dashboard, on the same cadence, so the trade-off is visible at the moment of decision rather than two quarters later.

Time-box the comparison. Some networks now release comparative data in periodic, scheduled reviews rather than live dashboards. The difference matters: a quarterly review supports planning; a live badge supports compulsive checking. The former uses comparison as an input to strategy. The latter uses it as a source of ambient anxiety.

Reward the judgment, not the rank. The issuers that held their performance through the badge rollout tended to be the ones whose internal incentives were tied to decision quality — fraud loss ratios, dispute rates, merchant retention — rather than to the visible number. When the badge moved, their behavior didn't, because the badge wasn't what they were being paid for.

The 18% figure is a symptom, not a law. It reflects what happens when a measurement designed to inform gets repurposed to compare, and when comparison gets repurposed to motivate. Payments is a domain where the underlying judgments are genuinely hard — authorization decisions under uncertainty, fraud calls with incomplete information, risk pricing that has to hold up over years. Those judgments need feedback, and feedback needs measurement. What they don't need is a public scoreboard that turns a complex skill into a single line and then asks people to defend it.

The next time a network proposes a leaderboard, the useful question isn't "will this motivate better performance?" It's "what will people stop paying attention to once this number is visible?" That second question is where the 18% lives.