Definition
The ratio of realized outcome improvement from an automated retrain cycle to the full cost of that cycle, including re-evaluation, cache invalidation, downstream re-tuning, and review load.
The Measurement Gap
There is no metric that tells you the net yield of an automated retrain against its churn cost, which means retrain frequency is tuned by thresholds rather than value.
Formula
RNEY = value_improvement / (compute_cost + eval_cost + invalidation_cost + review_cost)
Proposed formula, pending validation pilots.
At a Glance
Decision Supported
Whether drift-triggered retraining frequency is creating or destroying value, and where retrain thresholds should move.