social share alt icon




An American multinational pharmaceutical and biotechnology corporation.


The client has 27,000 product portfolios catering to 14 markets across the world. The client's existing ERP systems required the manual intervention of the regional sales team to avoid supply chain disruptions and accurately predict the market demand for products. The client aimed to improve their demand forecasting to minimize manual intervention by the sales team, optimize their supply chain operations, and enable faster time to market. In addition, they wanted to create a prediction pipeline using time series and machine learning-based models that outperform the existing forecasting system's benchmark and improve the demand forecast for the supply chain by 20% overall and reduce over-fitting models to optimize the performance.




The client approached Mphasis to improve existing demand framework algorithms and increase accuracy by 5% every quarter for a better prediction pipeline of commercial solution. Mphasis Next Labs, which focuses on disruptive & break-through innovations, analyzed the data to understand the problem and design new algorithms to improve the projections. Our team proposed The Front2Back (F2B) approach to showcase the direct impact on critical business KPIs (forecasting accuracy) and implemented the following solution :

  • Based on analytical time series (Drug lifecycle, type, demand patterns, and other) products data was divided into 17 unique categories.
  • Incorporated 'Automatic Model Selection' framework into the intelligence layer of solution to eliminate manual processes for product selection and further automated the entire process flow, leveraging ML Ops methodology to analyze and effectively provide the most contextual and accurate decision metrics from structured and unstructured data.
  • Using DeepInsights™ & Pace-ML, and code optimization, Mphasis developed 40 separate models for time series and machine learning models, which help cut production code running time by 30% (from 18 hours to 12.5 hours 27000+ products with 40+ algorithms)


Our innovative implementations delivered the following outcomes:

Increased forecast accuracy from 38% to 65% for the worldwide portfolio within eight months

Enhanced forecast accuracy from 45 % to 72 % for top 5 markets (contributing 60% revenue)

Reduced overall costs by 2.5 percent by transforming business processes and boosting accuracy

The local sales team was manually changing the demand prediction for 40% products at the outset of the project, and after eight months, further lowered it to only 22%

1.5 % increase in revenue due to the improved forecast's downstream impact by modernizing inventory and lowered inventory management expenditures by 2.5% while upgrading customer service standards

40% reduction in development life cycle time and a 30% reduction in production uptime (from 18 hours to 12.5 hours) through ML Ops Principles application

Improved production planning and minimized manual intervention through DeepInsights™-Cognitive AI-ML Solutions