Model Performance Monitoring

The deployment and integration of a predictive model into one’s business processes often requires a significant investment from both a time and labor perspective. As more businesses rely on predictive models within their day-to-day operations, it is imperative to protect that investment and ensure the models are working as they were originally intended. The ongoing health and sustainability of a deployed model is essential to ensuring critical business decisions are made using the most accurate information. Different than model system monitoring, which examines more obvious issues like model failures and records processed, Model Health and Performance Monitoring detects more subtle signals that a model may not be delivering its intended value and may be degrading.

Whether it’s our own custom-built models or your existing models, we can help ensure their longevity and ensure your key business processes and decisions are made using the most accurate predictions and classification. Our proactive monitoring solutions measure and detect subtle changes that may impact model performance and ensure that your production models are performing optimally.

Key Areas monitored:

• Data Completeness
• Feature Drift
• Prediction Drift
• Tests of Model Accuracy (when labelled data is available)