Why MLOps practices matter for optimization algorithms – ANWB Case Study

In high-stakes, time-sensitive environments like the ANWB alarm center, manually finding the optimal solution for complex problems, such as assigning replacement cars, can be a challenging. This presentation introduces 'Hanna', our optimization model designed to identify the best replacement cars, significantly reducing agent workload and improving efficiency. This session will shine a light on two key learnings: the surprising difficulty in gaining consensus on the exact objectives and constraints for an optimization function across an organization, and the absolute necessity of rigorous monitoring to ensure the model remains performant over time. We demonstrate why applying MLOps practices – even for non-ML models – is crucial for operational stability and long-term value.