Case study card:
Detect fraudulent resale of pharmaceutical drugs
The challenge was to reduce fraud in internet resale, which included the sale of products to 10,000 pharmacies.
This is a slow, manual and inaccurate analysis (30 min per pharmacy).
Any analysis also had to account for seasonal variance, varying by location – and many other factors.
They built a system that automatically collated external data streams.
Their system classifified pharmacies automatically, according to the probability of fraud.
Their solution was a modular, extensible and interpretable algorithm.
The system caught more fraud and saved the time of a human alternative.