Case Study: a leading US bank reduces loss forecast variance with Scienaptic AI

Forecast better, faster, at scale with advanced ML techniques

The leading US bank faced significant challenges with its loss forecasting, experiencing over 16% variance from actuals due to its reliance on traditional methods. This inaccuracy led to P&L volatility. To address this, the bank partnered with Scienaptic AI and leveraged its advanced ML platform.

Scienaptic AI implemented a solution utilizing a suite of 15 challenger models, including RNN, Cox PH, and SARIMAX techniques, which incorporated a wide array of internal and macroeconomic data. This deployment dramatically improved forecast accuracy, reducing the variance to just 2%. The solution also provided confidence intervals for better risk assessment and reduced the model refresh time from months to mere days, creating a more robust and efficient forecasting process for the bank.