FORECAST ERROR REDUCED BY 15% THROUGH AUTOMATED MODEL SELECTION COMBINED WITH MLOPS PIPELINE

Know More

An American life sciences enterprise was facing below mentioned challenges -

Demand forecasting required for 27 thousand SKUs with a forward outlook of 36 months
Forecasting to be repeated monthly, with the latest monthly data incorporated into the univariate time series
Limited features and data descriptions available
Need for accurate forecasting for proper planning

Through digital transformation we enabled demand forecasting for SKUs using smart analytics and machine learning. We helped the client in -

Improving forecasting accuracy through causal forecasting methods
Classifying the best performing models for each of the SKU groups by using “model win”
Identifying seasonality by using an unsupervised learning approach
Identifying SKU groups by leveraging statistical parameters

15% forecast error reduction through automated model selection led forecast along with MLOps on AWS Cloud

45% reduction in overall manual effort by model selection and demand correction

2.5% decrease in overall cost through improved production planning and reduced manual intervention

1.5% increase in sales revenue through optimal inventory volume and enhanced customer service level