Abstract
This paper analyzes the structural analogies and differences between the life cycles of measuring systems and machine learning (ML) systems for classification. Although measurement maps empirical quantitative properties to measured values and ML classification systems map information about entities to classes, both rely on parametric models that require optimization through external references: properties embodied in measurement standards and labeled training sets, respectively. Building on frameworks published in relevant literature, the paper proposes a unified three-stage - Context setup, System development, and System deployment - conceptual and procedural structured framework about the life cycles of measuring systems and ML systems. Feedback loops emphasize iterative refinement driven by new insights, evolving requirements, and environmental changes. By comparing the two life cycles, the paper highlights how metrology and ML can mutually contribute toward more robust, traceable, and reliable measuring and ML systems.