Why clean processes still produce bad outcomes

The British administration in colonial Delhi wanted to reduce the cobra population. They offered a bounty for every dead cobra. The bounty worked at first, then it stopped working, because locals began breeding cobras to collect the reward. When the administration realized what was happening and cancelled the program, the breeders released their now-worthless snakes, and the wild cobra population ended up higher than it had been at the start.¹

The story is told often enough that it has become a cliché, but the underlying mechanism is worth taking seriously. The administration designed a measurement that perfectly captured the activity they wanted (dead cobras delivered to a counting station) and missed the goal they actually cared about (fewer cobras in Delhi). The measurement was clean and the bounty process worked exactly as designed. The outcome was the opposite of what anyone intended.

This is the failure mode that most process efficiency projects do not address. The work of redesigning a process happens at the level of what people do. It rarely happens at the level of why they do it. A clean process running on top of misaligned incentives does not produce better outcomes. It produces worse outcomes faster.

What the research actually shows

The economist Charles Goodhart formulated the principle in 1975: when a measure becomes a target, it ceases to be a good measure.² The reason is structural rather than moral. People respond rationally to the incentives they are given, and if the incentive measures the wrong thing, the rational response is to optimize the wrong thing.

The pattern is well documented in research on sales compensation. According to the Alexander Group, only 21% of companies express satisfaction with their sales compensation plans.³ Drivetrain’s analysis of compensation plan effectiveness identifies the most common failure pattern directly: when reps are incentivized solely on revenue booked, there is little disincentive to discount heavily, and a rising discount rate is a sign that the plan is rewarding the wrong behavior.⁴ The deals close and the targets are hit. What disappears is the margin.

This is not a sales-specific problem. It shows up in three places in mid-sized companies more often than anywhere else.

1. Sales targets that hit and miss the business

The most common version. A sales team is compensated on gross revenue, with no margin component and no clawback for cancellations. The team is professional and responds to the structure they have been given. Deals get pushed through with heavy discounts, terms get loosened to accelerate closes, and contracts that should not have been signed get signed because the commission is paid on booking, not on what controlling later writes off.

The sales process itself can be excellent. Clear stages, fast handoffs, good CRM hygiene. None of that matters at the outcome level if the incentive points away from profitable revenue. The fix is not in the process. It is in the compensation plan. Either margin enters the formula, or clawback provisions catch the deals that should not have been counted, or both.

2. Operational metrics that get gamed once they become bonus criteria

A manufacturing operation introduces a defect-rate target as part of the plant manager’s bonus. Defect rates drop the following quarter. Everyone is pleased. What has actually happened, in many cases, is that the definition of defect has quietly tightened. Marginal issues that used to be flagged are now classified differently. The plant is producing the same quality of work and reporting better numbers.

This pattern is most acute when the metric is self-reported and the bonus is large relative to the work it would take to game the definition. The failure here is in the metric design, not in the people responding to it. The solution is not to punish gaming, but to design measurement systems where gaming is either impossible or indistinguishable from genuine improvement. The simplest version: the metric should not be defined and reported by the person whose bonus depends on it.

3. Customer metrics that improve as customer relationships deteriorate

A company tracks customer satisfaction through a quarterly NPS survey. The score climbs over four quarters. The board is pleased. Customer churn also climbs over the same period, but more slowly, and the two trends are not put next to each other until someone asks why.

The mechanism is usually that the dissatisfied customers have stopped answering the survey. The denominator shrinks faster than the numerator. The score reflects a more loyal subset of customers, not a more loyal customer base. The metric is technically accurate. It is also telling the company something close to the opposite of what it appears to say. When numbers become the primary goal, people start chasing the metric rather than improving the underlying reality.

What the process work should include

A serious process efficiency project does not stop at how the work flows. It also asks how the work is rewarded, and whether the rewards point in the same direction as the redesigned process. Three questions worth answering before any process improvement is approved:

For each role affected by this process, what gets that person’s bonus, recognition, or promotion? If the redesigned process performed perfectly, would that change what gets rewarded? And what is the simplest way for someone to game the new process to hit their existing target, and what stops them from doing it?

These are not HR questions. They are process questions, because a process is not an output, it is an outcome, and outcomes depend on what the people in the process are paid to achieve. A clean process running on top of misaligned incentives is the corporate version of the Delhi bounty program: well-administered, internally consistent, and producing the wrong result faster than the broken process it replaced.