A Metric Is Easy to Game and Someone Is Gaming It: Decision Tree

Why this matters

Every incentive gets tested against its cheapest workaround eventually, and someone on your team will find it before you do. A close rate that improved because estimates only go out on sure things, a completion count that climbed because jobs got marked done before they were finished, an on-time metric that looks great because the clock stops the moment the truck parks, not when the customer's problem is solved. Left unaddressed, a gamed metric is worse than no metric at all: it tells you everything is fine while the real thing it was supposed to track quietly gets worse. Catching it is a specific skill, and confronting it well matters as much as catching it.

Start here: confirm gaming before you confront anyone

A metric that looks suspiciously good is not proof of gaming. It could be genuine improvement, a data-entry error, or a change in what is being counted. Jumping straight to an accusation on a hunch damages trust for nothing. Before any conversation, pull the underlying records behind the number and look for the specific pattern that separates gaming from a real win.

If the underlying records show real work behind the number (jobs are actually complete, calls actually resolved the issue, estimates genuinely reflect the full scope), the metric is telling the truth. Stop here, this is a real improvement, not gaming.

If the underlying records show a gap between what the metric claims and what actually happened, move to the next step.

Step 1: Identify which of the two gaming patterns you are looking at

Gaming almost always falls into one of two buckets, and they call for different responses.

• Definition gaming: the person is hitting the letter of the metric while missing its point. A job gets marked complete when it is 90 percent done, a call gets logged as resolved when the customer will clearly be calling back, an on-time arrival gets counted from the moment the truck is on the street rather than at the door. Nobody lied about a fact, they exploited an ambiguity in how the metric is defined.
• Selection gaming: the person is choosing which work goes into the pool the metric measures, rather than changing how the work itself is scored. Only sending estimates for jobs you already expect to close, avoiding difficult customers who might lower a satisfaction score, cherry-picking easy jobs to keep a per-day count high while a harder backlog sits untouched.

Knowing which one you are dealing with tells you where the fix belongs: definition gaming means the metric's definition is too loose, selection gaming means something in how work gets assigned or scored needs to change.

Step 2: Decide whether this is one person or a systemic pattern

If only one individual shows the pattern, this is likely a direct, one-on-one conversation, not a policy change. Assume good faith on the first pass. Most people who found a shortcut did not set out to deceive anyone, they responded rationally to what was being rewarded. Ask them to walk you through how they are hitting the number, without accusing, and the gap usually becomes obvious to both of you in the conversation itself.

If multiple people independently show the same pattern, this is not a person problem, it is a metric-design problem. When several different people converge on the same shortcut without coordinating, the incentive itself is inviting it, and fixing one person's behavior while leaving the metric as-is just produces the next person doing the same thing. Go to the fix at the metric level, not the individual level.

Step 3: Fix the metric, not just the behavior

Whichever pattern you found, patching the individual instance without changing the underlying setup guarantees a repeat.

• For definition gaming, tighten the definition so the loophole closes. Redefine "complete" to require a specific verifiable condition (customer sign-off, a photo, a checklist item), not a status flag anyone can flip. Redefine "on time" from the moment of arrival at the door, not the moment the truck entered the neighborhood.
• For selection gaming, add a second metric that exposes the selection bias, mirroring the fix for a bad target: track total estimates sent alongside close rate, track job difficulty or type alongside completion count, so a rising primary number against a shrinking or easier pool becomes visible instead of hidden.
• Where possible, pair the gamed metric with a metric that catches the actual outcome it was standing in for. A completion count paired with a callback rate. An on-time rate paired with a customer satisfaction score. Gaming one number usually shows up as a decline somewhere else; make that somewhere else visible on the same screen.

Step 4: Address the conversation itself with the person or team, separately from the fix

The metric fix and the human conversation are two different jobs and both need to happen.

• Frame it as "I noticed a way this number can be hit without really moving the thing it's supposed to track, and I want to close that gap" rather than "you gamed the system." This keeps the conversation about the incentive, not an accusation of dishonesty, and it is usually the truer read of what happened anyway.
• If the gaming was deliberate and known to be dishonest (falsifying a record, lying about a completed step), that is a separate, more serious conversation about integrity, not a metric-design issue, and should be handled as such.
• After the fix goes in, watch the metric for a full cycle to confirm the real pattern underneath it, not just the number, actually improved.

The recap

Confirm the gap between the metric and reality before assuming gaming. Sort it into definition gaming or selection gaming, they need different fixes. If it is one person, have a direct, assume-good-faith conversation; if it is several people independently, fix the metric itself. Tighten the definition or pair it with a metric that exposes the shortcut, and separate the technical fix from the human conversation about how it happened.

References

• U.S. Small Business Administration (SBA), small business performance management practices
• Trade-standard practice for KPI design and integrity in field-service operations
• See related: Setting a Target for a Metric Without Inviting the Wrong Behavior; A Metric Was Wrong Because of Bad Data Entry