1. Monitor & Refresh
Monitor Model Performance
Determine When Refresh Is Required
Refresh & Update Deployed Model

Monitoring and refreshing models after deployment is a frequently overlooked but critical step to protecting your investment and forecast returns.

 All predictive models experience ‘drift’ over time.  This reflects the fact that the world is in constant change around your business and that what was true of your customers, prospects, products, competitors, pricing, or any other factors impacting your business change over time.  At some point in the future your model will need to be refreshed.

 It is also possible that some of the same errors found to exist in your data have recurred after deployment, degrading performance.  From the Data Preparation phase you will remember that this may be missing values, incorrectly entered values, or outliers.

 The key to knowing when a refresh is necessary is the on-going monitoring that is part of every PreMo modeling engagement.  During the Readiness Assessment we will discuss and agree how often this is to be accomplished.  Periodically we will report to you on the status of your model and whether the change in external circumstances or internal data inconsistencies has degraded the accuracy of your model.

 When the time comes to refresh, the formal approach we have taken to development will pay off with quick and cost efficient model updates which are then redeployed.