InComm Gains Ability to Accurately Uncover Fraudulent Charges While Converting Inquiries to Sales.
I think this was awesome. I think that the POV has benefits that we can already improve, whether from a transactional perspective or from a model perspective. I think this was exciting, incredible and met all of my expectations and should be celebrated.
– Anup Thomas | Senior Information Technology Executive
InComm’s former fraud detection system was returning frequent false positives due to a manual rule-based process. Too many abandoned carts and failed transactions showcased a lack of a view of the customer’s journey and further harmed financial forecasting.
Pandrea set out to create a machine learning model that could outperform the current fraud identification system that Incomm was currently using. Pandera utilized AutoML for this POV which allows for scalability, complexity, and provides a higher degree of accuracy. Additionally, Pandera created a Looker dashboard with data based in BigQuery to give InComm the ability to see the full picture of the customer journey and better understand the root causes of abandoned carts on their website
Pandera’s newly created fraud model outperformed the previous model, resulting in decreased false positives by nearly 50%. Furthermore, InComm gained the ability to view the end-to-end customer and transaction journey. The implementation of new dashboards also reduced research time for InComm developers by 66% with the assistance of a unified research zone.