Forecasting with higher accuracy and Greater versatility in a competitive market
A large pharmaceutical/biotechnology company wanted to better understand the product potential of an on-market, non-retail drug distributed through a limited network of specialty pharmacies and distributors in an active, competitive environment.
Viscadia was asked to build a custom forecast model that could integrate seamlessly with the client’s existing Latest Estimate (LE) process and that would also be versatile enough to support multiple stakeholders and the long-term brand planning process.
Increase in accuracy compared to previous forecast model
CHALLENGES
Data complexity
- Product and distribution model involving multiple and disparate data sources
- IntegriChain (unit demand, inventory)
- Patient data (enrollments, restarts)
- Internal (finance, ex-factory)
Market event modeling in absence of competitor data
- Specialty (non-retail) distribution model across all competitors hindered the availability of market data, and complex market events were impacting product performance — increasing the complexity of the required modeling
Cross-functional stakeholder needs
- Sales, marketing and finance each had distinct objectives for the model, including harmonization of the unit-demand forecast with patient enrollments
APPROACH
Improved user experience
- Built an organic model with a streamlined flow, incorporating VBA-driven automation, with complete transparency and user-friendly design
Mapped cross-functional stakeholder needs
- Identified discrete needs across sales, marketing and finance stakeholders and created customized outputs to meet requirements
Modeled complex market events
- Assimilated market research (AMUs) to understand and model for multiple market events using diffusion curves
Designed multiple forecasting pathways
- Created capabilities for users to choose from multiple forecasting pathways (unit demand vs. patient enrollments) that were harmonized across the platform
Embedded in-house forecasting palette
- Integrated a robust and transparent statistical palette (Exponential Smoothing, Holt-Winters, Box-Jenkins, etc.) resulting in improved forecast accuracy and model flexibility
Developed forecast platform
- Customized platform for the product allowed for dynamic, comprehensive evaluation of key performance drivers, and served cross-functional stakeholder objectives
OUTCOMES
End deliverables
- Forecast platform with a comprehensive set of output views and summaries addressing cross-functional stakeholders
- User manual providing extensive documentation of model design and operation to facilitate easy knowledge transfer
- Executive slides to facilitate quarterly LE discussions, generated by an automated module within the model
Key highlights
- Reduced the MAPE (mean absolute percentage error) by nearly 50% in actual vs. forecast for the new model compared to the existing model
- Dynamically modeled for five major market events occurring at different times in the forecast horizon
- Successfully enabled the forecast team to tie the revenue forecast to weekly patient enrollment targets for the sales team
Impact
- Reduced workflow for quarterly LE from ~40 hours to ~8 hours
- New model served as the base platform for future Life Cycle Management (LCM) strategy scenario-planning exercises
- Client asked Viscadia to design and build a new model for another product within the same franchise
