Your Best People Are Optimizing Against You

There is a moment in almost every audit where an executive lowers their voice and says some version of the same sentence: our top performer is technically hitting every number and somehow making the company worse.

They say it like a confession, as though naming it is disloyal. It is not a confession. It is a design report. The person in question is doing exactly what the organization asked for, at a level of skill nobody else has reached yet. That is what being a top performer means.

Your best people are not gaming the system. They are the first ones to finish reading it.

Gaming is a competence, not a character flaw

Leadership almost always reaches for a moral explanation first. They are cutting corners. They lack integrity. They care more about the bonus than the mission. This explanation is comforting because it localizes the problem inside one person, and people can be coached or removed.

The trouble is that the behavior reappears with the replacement. If you have fired a sandbagger and watched their successor start sandbagging within two quarters, you have already run the experiment that disproves the character theory. What persisted was not a personality. It was a payoff.

Skilled operators find the efficient frontier of whatever you measure. That is the same capability that makes them valuable when the measurement is well designed. You cannot hire for relentless optimization and then be surprised by relentless optimization.

The four moves, and what each one is telling you

Metric gaming is not random. It shows up in four recognizable forms, and each one points at a specific defect in the structure above it.

• —Sandbagging — pulling deals or output into the next period. Tells you the target ratchets, so exceeding it is punished with a harder future.
• —Cherry-picking — serving only the easy segment of the work. Tells you the metric counts volume without counting difficulty.
• —Cost-shifting — hitting the number by pushing work onto support, ops, or the customer. Tells you the metric has no owner downstream of the handoff.
• —Definition drift — quietly widening what counts as a qualified lead, a resolved ticket, a completed milestone. Tells you nobody outside the incentivized team controls the definition.

Why it starts at the top of the performance distribution

Average performers follow the process because the process is the safest available route to an acceptable outcome. They are managing downside, and the brain filters hard for pain avoidance before it goes looking for upside.

High performers operate with more slack. Their track record buys them the safety to experiment, and their experiments quickly locate the gap between what the organization says it wants and what it actually pays for. Once that gap is found, exploiting it is not even a decision — it is just the obvious way to work.

This is why the earliest warning sign of a broken incentive is almost never a scandal. It is a quiet divergence between a rising metric and a flat or falling outcome, driven by the people you trust most.

A metric that rises while the outcome it proxies stays flat is not a success. It is a receipt.

The redesign: pair the metric, then move the pain

The instinct is to add a rule. Rules invite compliance theater and a new search for the next gap. The durable fix has two parts, and neither of them is a bonus.

First, pair every performance metric with a counter-metric that only moves if the shortcut was taken. Bookings pair with net revenue retention at twelve months. Ticket closure pairs with reopen rate. Hiring speed pairs with ninety-day survival. A single metric can be gamed cheaply; a pair usually cannot, because the shortcuts that inflate one deflate the other.

Second, relocate the pain. We change to avoid loss far more reliably than we change to chase gain, so the highest-leverage edit is usually removing a threat rather than adding a reward: stop ratcheting the target off last quarter's overperformance, and sandbagging loses its purpose within a cycle. Give the downstream team a formal veto on the definition, and definition drift stops without a single policy memo.

• —Name the outcome the metric is a proxy for, in one sentence.
• —Ask your best performer how they would hit the number if they stopped caring about that outcome. They will tell you, in detail.
• —Pair the metric with the counter-metric their answer implies.
• —Remove the retroactive penalty that made the shortcut rational in the first place.
• —Re-measure in one quarter against the outcome, not the proxy.

What this looks like when it is working

The tell is boring: the metric grows more slowly and the outcome starts moving with it. Your top performer keeps their standing, because their advantage was never the shortcut — it was the speed at which they find the efficient path through whatever structure exists. Give them a structure worth optimizing and they will optimize toward the thing you actually wanted.

That is the entire discipline. Not better people. A payoff sheet that agrees with the strategy deck.