IMPROVED THE DEMAND FORECAST SYSTEM'S ACCURACY AND REDUCED MANUAL INTERVENTION WITH AUTOMATIC MODEL SELECTION & ML OPS ON AWS

Know More

THE CLIENT

One of the world’s largest pharmaceutical company.

BUSINESS CHALLENGES

With increasing complexity, velocity, and volatility in today’s markets, predicting drugs’ future demand is getting more difficult every year. Changing government policies and regulations can change drug sales in different geographies. The client was looking for a solution to deal with below challenges:

Different demand patterns for drugs portfolio and drug combination for each country
Below par results from the existing demand prediction system
Predicted demand being modified by local sales teams based on their experience and market knowledge

SOLUTION

Mphasis' solution utilizes ML pipeline built on AWS cloud, using Amazon Sagemaker and leveraging additional AWS services such as Amazon API Gateway, AWS Lambda, Amazon S3 Bucket, Amazon Redshift, AWS CloudTrail, AWS IAM. The solution classifies the global portfolio of SKUs into different category based upon their demand patterns. We helped with -

Setting up the pipeline for seamless production deployment and faster deployment of new changes/models into production
An automatic model selection framework that identifies the best model combination (out of 40 time series, machine learning and deep learning models) for each demand category. The framework accounts for changes in demand due to regional and market factors and identifies abrupt changes in demand patterns

BENEFITS

Reduced forecast error (Root Mean Square Error) by ~15%
Reduced costs by ~2.5%
Boosted sales revenue by 1.5% (~$600 million)
Project completed before time as ML pipeline enabled rapid pushing of new approaches into production