Abhishek Dash

The Client (Pharmaceutical major in United States) needed to predict the potential oncologists for a drug that catered to secondary treatment of leukemia patients.

The leukemia drug market is primarily governed by a competitor drug. Only when this drug fails, the secondary treatment starts.
Due to lot of churn in the market (a prescriber can be in the highest decile in one year, and not in the following year).
The data set was for a period of 4 years till July, 2013. We took all… Show more

The Client (Pharmaceutical major in United States) needed to predict the potential oncologists for a drug that catered to secondary treatment of leukemia patients.

The leukemia drug market is primarily governed by a competitor drug. Only when this drug fails, the secondary treatment starts.
Due to lot of churn in the market (a prescriber can be in the highest decile in one year, and not in the following year).
The data set was for a period of 4 years till July, 2013. We took all factors ranging from prescription behavior, demographics, affiliations, external factors (cigarette smokers, average temperature).
A logistic model with a six month refresh was decided since churn percentage was significant)
After the exploratory data analysis, we proceeded with variable treatment (uni-variate, bi-variate, correlation)

We build the model on the training set and validated on the test, which was a 70-30 split.

Of the 300 variables, the model generates 18 significant variables.

The models accuracy turned out to be around 85% if we take the top 50% of oncologists in the test period.
Show less

The client wanted a tool to represent the SC metrics for all drugs produced by him across 3 continents.

The Tool was designed to produce a causation hierarchy channel for better understanding of the business problem. For e.g. North America performs poor on Actual vs Budget for July, 2012. The flow would lead us to Country -> States -> Brands -> Molecules path.

The tool was built to provide bottom-of-the-pyramid driver analysis of different performance metrics.

SC… Show more

The client wanted a tool to represent the SC metrics for all drugs produced by him across 3 continents.

The Tool was designed to produce a causation hierarchy channel for better understanding of the business problem. For e.g. North America performs poor on Actual vs Budget for July, 2012. The flow would lead us to Country -> States -> Brands -> Molecules path.

The tool was built to provide bottom-of-the-pyramid driver analysis of different performance metrics.

SC metrics like Actual vs. LE vs. Budget, Stock outs, Back orders and Turns were represented.

The Client appreciated a wholesome dashboard with the drill downs explaining the sufficiency and insufficiency while selling a drug. Show less