Background
The client is a global pharmaceutical company specializing in brain health, with a large sales force across Canada. They needed to modernize their sales quota-setting process, especially for two high-performing drugs. Their existing process was manual and based mostly on historical sales data
Challenge
The traditional forecasting approach had several limitations: • Manual effort: Time-consuming • Multiple influencing factors: Sales quotas were shaped by historical sales, competitor activity, physician prescribing behavior, and market dynamics. • Data fragmentation: These variables were stored across different tables, making integration complex. The client needed a data-driven, automated solution that could integrate multiple data tables and provide accurate, granular forecasts
Approach
Implemented best AI/ML model approach in the forecasting solution using the following steps: 1. Data Integration: Pulled data from IQVIA, including: -Historical sales -Physician data -Market trends -Competitor performance 2. Data Preparation: o Cleaned and structured the data o Engineered drug-specific features to improve model relevance 3. Modeling Techniques: o Used a combination of SARIMA (for time series), Random Forest, and XGBoost (for pattern recognition and prediction) - selected the best fit model o Trained models to forecast monthly sales at the MSA level
Key results
Benefits
By integrating various data sources, including historical sales, physician data, and market trends, and employing a combination of advanced AI/ML models like SARIMA, Random Forest, and XGBoost, Keyrus developed a solution that achieved approximately 70% forecast accuracy at the MSA and monthly levels. This automated, data-driven system significantly reduced manual workload and human error, enabling more timely and effective sales planning while also being scalable for future use on other drugs.
Technology partners
Python
Share this key play
