The sales of a key brand, accounts for about 50% of total revenue for a Biotech company, was suddenly experiencing slow down after many years’ steady growth. Demand forecast for rest of fiscal year were revised multiple times during the first few months. Senior management would like to know what’s the main reason for the sales slow down and whether company needs to issue a warning prior to the coming earnings call.
Without reliable patient utilization data, existing financial forecasting models were built based on transaction data only. The purchase pattern can be influenced by so many non-demand related factors, which often limits the usefulness of the transaction data. We took a total different approach by reconstructing true patient level demand from transaction data and other information using an innovative data fusion process.
We built a predictive models on each individual drivers and calibrated them with transaction data to assess the reliability of our models. A new underlying demand based forecast/simulator was built to provide realistic projection for future product sales.
There was no slowdown in underlying patient demand for the brand based on the analysis. Many confounding factors were all turned to one direction at the same time, which caused the panic to the organization. 12 months after the engagement, we learned that the error of our new demand forecasting model was less than 1%.