We have updated the forecasting engine to interpret absent data (implicitly zero data) as explicitly zero data when generating a linear regression model used to forecast measures. This adjustment minimizes the impact of a small number of irregular charges on the model.

Before

After