The incentive paid for the pilot, not the outcome

Meta description: Teams optimize for the metric someone chose when the real incentive is activity that is easy to explain and buy.

Miklos Roth

The scorecard was designed sensibly, and I want to say that clearly, because the failure I am describing did not come from carelessness. The leadership team wanted AI adoption, so they paid for adoption: use cases launched, teams onboarded, tools in production. The metrics were achieved — exceeded, in most of the companies I am thinking of. The outcomes the metrics were meant to serve did not arrive, because the scorecard paid for the launch and the outcome lived after it, unmeasured and unrewarded. The team was not irrational. The team was precise. They did exactly what the incentive said, and the incentive was wrong.

[AWAITING MIKLOS — scene: a bonus tied to use-cases launched, and a quiet penalty for killing one]

What does the current reward actually pay?

The forensic question I ask of scorecards is: what behavior, repeated all year, maximizes this person’s reward? The answers in these companies are consistent. Launching pays; operating does not. Starting pays; stopping costs — a killed project is a visible failure, a zombie project is an invisible one. Activity pays; outcomes are attributed so diffusely that chasing them is bad personal economics. When I lay the scorecard’s actual payments next to the program’s stated goals, the two documents describe different strategies. The organization is executing the scorecard’s strategy. It usually does not know what the scorecard’s strategy is.

Why does an added outcome metric lose to the old one?

The incentive I see distorting behavior is the preference for the quick, visible, easy-to-explain activity over the difficult foundation. Organizations choose the marketing activity that looks easier to buy and explain, and avoid the work whose success is invisible when it succeeds. The result is a scoreboard full of motion — campaigns launched, tools adopted, content published — while the unmeasured constraint stays exactly where it was.

The false explanation is balance: add an outcome KPI beside the activity KPI and the incentives will correct. I have watched the outcome metric get added and lose, quarter after quarter, to the activity metric it was meant to discipline. The mechanics favor the old metric: activity is countable now, outcomes arrive later and contested; activity is attributable to the team, outcomes are shared with the market; activity is in the team’s control, outcomes are hostage to everything. In a contest between a metric that is immediate, attributable, and controllable and one that is not, the scorecard’s weighting is decoration. The immediate metric runs the year.

What behavior is rational under this scorecard?

Under these scorecards, the rational portfolio is many launches, shallow operating, no endings. I have watched capable, ethical managers build exactly that portfolio while privately knowing better — the pilot that should be killed is kept alive because killing it costs standing; the use case that should be deepened is left shallow because depth is unpaid and the next launch is. The managers are not cynics. They are bounded: they have a year, a team, a mortgage, and a scorecard. The scorecard’s rationality and the program’s needs diverge, and the scorecard is the one that pays.

What pattern marks distorted incentives?

The marker I trust is the gap between the review narrative and the reward history. Listen to what the program celebrates in reviews, then read what the last three bonus cycles actually paid for. In distorted systems, the celebration language has drifted toward outcomes — the program talks value, impact, transformation — while the payments remain attached to launches and activity. The language updates faster than the incentives, and the drift is itself diagnostic: the organization knows what it should be paying for, says it, and continues paying for something else.

A good scoreboard question is whether each metric names a business result or an activity. “Use more AI” is not a business result. A precise objective reveals which measure will indicate whether the approach was worthwhile. If the incentive system rewards activity, the team will produce activity; the optimization is rational, and the metric is the cause.

[AWAITING MIKLOS — scene: second scene]

What question exposes the paid behavior?

Incentive conversations ask what we want to encourage — an intention question, answered with values. The question that exposes the distortion is behavioral: what did this scorecard actually purchase last year? The answer requires no philosophy, just an audit of payments against behaviors. I have watched leadership teams go quiet over that audit, because the purchased behavior was on the table and it was not the behavior in the strategy deck. The scorecard had been faithfully executed. That was the finding.

The question that audits the incentives: which of the numbers this team is rewarded for would survive being restated as a business result — and which are activity wearing a result’s clothing?

PR excerpt:[PR — awaiting Miklos material; cut 80–120 words from his paragraphs once available.]

By Miklos Roth — AI marketing consultant, Budapest. More: his field notes.