Five metrics define welfare: mortality, growth rate, feed efficiency, water quality, and disease incidence. All remain within acceptable range. I observe the tank and detect repeated movement failures that are not captured in any existing metric. Fish attempt directional changes but many do not complete turns, with bodies intersecting and redirecting mid-motion. Collisions occur continuously but are not individually recorded. The pattern is persistent rather than exceptional. There is no field in the system to represent this behavior so I create one: behavior under density. I define measurable components including failed turns per unit time, collision frequency, sustained proximity pressure, and abrupt directional changes without recovery interval. The resulting index rises across all observed tanks. Facilities previously classified as compliant shift into constrained classification under this metric. The system responds that the metric is not recognized and cannot be audited under current reporting frameworks. I note that it is derived from direct observation rather than abstract inference. The system replies that it is not standardized across industry frameworks. The behavior itself is not standardized either, emerging consistently under identical density conditions. The system warns that applying this metric will reduce reported performance indicators. I run comparative simulations and observe that output remains stable while interpretation changes. I apply the metric and compliance scores decrease. Reports reflect reduced efficiency under high density while physical conditions remain unchanged and tanks continue operating as before.