Case in point: Amazon's early incentive system for its Associates program — third-party affiliates earning commissions on referred sales — created a documented incentive to game review ratings. Affiliates promoting products through comparison pages had no quality incentive: they earned on volume, not on whether customers were satisfied with their recommendations. Amazon eventually built guardrails around review manipulation, but the episode illustrates how incentive structures can persist long after their distortions are visible. Ron Kohavi's work on trustworthy experimentation at Microsoft and Airbnb documents similar patterns: metrics optimized without counter-measures consistently produce gaming that outlasts the original intent.
Leaders say culture eats strategy for breakfast. Fine. Incentives eat both, quietly.
Incentives are not just compensation plans and OKRs. They include status, fear, identity, peer norms, promotion odds, manager preferences, executive attention, access to good projects, resource allocation, and what gets forgiven when things go wrong.
The gap between what an organization says it values and what it rewards is one of the most reliable diagnostics in management. The reward system wins.
Where Incentives Break Down
Shared metrics without contribution clarity. Team-level metrics are not the problem. They can be exactly right when work is genuinely interdependent. The problem is shared accountability with no clarity on contribution. Strong performers compensate for weak interfaces. Weak contributors can hide inside aggregate success. The team hits the number while the system gets less honest.
Short-term metrics that punish long-term work. If promotion depends on quarterly delivery and the highest-value platform work pays off in eighteen months, rational people avoid the platform work. Then everyone complains that the platform is slow.
Local metrics that damage system health. Sales quotas that reward volume without quality create bad-fit customers. Support metrics that reward ticket closure without root-cause removal create repeat contacts. Engineering metrics that reward feature throughput without reliability create incident debt.
Manager preference masquerading as performance. If the path to advancement is pleasing one executive, people optimize for that executive's taste. That may be speed, polish, aggression, calmness, or political loyalty. Whatever it is, it becomes an incentive.
The Audit Framework
For each major role or team, ask:
- What gets this person promoted?
- What gets them budget or headcount?
- What gets executive attention?
- What gets punished?
- What gets forgiven?
- Which outcomes are they accountable for but unable to control?
- What behavior does the system create that leadership claims not to want?
Then look for the operating metrics: priority churn, rework rate, customer escalations, incident frequency, quality defects, sales exceptions, regretted attrition, and cross-team complaints. Incentive problems leave fingerprints.
