Data Science to Unlock $45M Opportunity to Improve Revenue for a Biopharma Organization
A leading fortune 500 pharmaceutical company headquartered in the United States was experiencing a decline in the sales volume of its blockbuster drug, which usually accounts for over $5 Billion of its annual sales. Less than 25 percent of physicians accounted for a loss of prescription sales worth approximately $350 Million. Looking at the situation, the company wanted to predict which physicians were likely to show a continued decline in drug prescription behavior to effectively develop an engagement strategy for them well in advance for to sustain their market share by improving sales.
To address these revenue challenges, D Cube identified the potential drivers of prescription sales, which were impacting physician’s sales behavior and helped in identifying which physicians were likely to decline in their drug prescription activities.
Physicians who were highly probable to show a decline in sales were analyzed further with their patient volumes, market shares, etc. and were profiled into similar segments. Priority list and next best engagement recommendations were provided for each segment group to drive actionable decisions. Using this, physicians were targeted through marketing and sales interactions to improve prescription behaviors.
In order to predict prescription behavior decline, multiple machine learning models were developed and tested, which used physician’s data who were present in the client’s sales force system as targets.
To identify the relevant features/data which would impact prescription behavior, numerous hypotheses were generated on prescription characteristics. Using this, we identified more than 2000 features from various data sources(like sales, calls, marketing, claims, and physician attributes) that capture physician activities. These features included ratio variables, change variables, direct variables, etc. like change in DTC TV impression on Physician, the ratio of calls given this month to last month, etc.
Various traditional and advanced data science techniques, such as generalized linear model(logistic regression), ensemble models, machine learning models, were used to benchmark the model and develop the best robust model. The finalized model identified different physician drivers for decline behavior, which validated the various hypotheses and revealed the impact of all the features on physician prescribing behavior.
Algorithms identified over 1,100 physicians who had a high probability of a decline in prescription behavior. This revealed ~$45 Million annual revenue opportunity, which could be potentially achieved with the appropriate targeting interventions on sales and marketing for these physicians.
To make the most of this opportunity, different personas were created based on the drivers for each physician segment overlaying overall prescription volumes to prioritize the selection of the physicians for targeting. The biopharma company was recommended with specific targeting interventions to achieve the desired impact, including but not limited to increase in penetration during a call by connecting with multiple stakeholders, customizing the detailed message to focus on solving reimbursement issues and ensuring that call detailing also right set of messages on Safety/ Efficacy/ Support.