Predictive models are at the heart of pricing decisions -- but even the best model can easily deteriorate as things change in the market.
Knowing when this happens and how to quickly fix models is essential. We have developed a model monitoring framework for this purpose. This includes metrics developed specifically for monitoring as standard model fit metrics are not appropriate for monitoring future data.
We also use machine learning techniques to identify poorly fitting segments, enabling efficient model updating. Rather than manually checking model fit across all possible variables, we use segmentation algorithms to find important segments to save on time and effort.