Why a quota changes the composition of output

A support team told to close thirty tickets a day will close thirty tickets a day. The number is met and the metric looks healthy. What changes is which tickets get picked, how long each answer is, and whether a difficult case is resolved or reclassified into something faster. The count is protected by moving everything the count ignores.

What everyone sees

Missed quotas are treated as effort problems and met quotas as evidence the system works. Both readings assume the output is a single quantity that rises and falls. Output is a bundle, and a quota prices only one attribute of it, which means the remaining attributes are available to be traded away at no measured cost.

What is actually happening

Holmström and Milgrom analysed exactly this in multitask agency: when an agent allocates effort across several activities and only some are measurable, strengthening the incentive on the measured activity draws effort out of the unmeasured ones. The optimal contract may therefore be weaker than it first appears, because a strong incentive on a partial measure distorts the bundle. Campbell had described the empirical version earlier: the more a quantitative indicator is used for decisions, the more it becomes subject to corruption of the activity it was meant to monitor. The quota does not fail. It succeeds on its own terms and pays for that success elsewhere.

Why it stays hidden

The substitution hides because the reported series is continuous. The count is stable or rising, so there is no moment where anything visibly breaks. Quality decline shows up as scattered complaints, repeat contacts and slow attrition, all of which arrive in different reports and at different times, and none of which is attributed to the number that caused them.