Managing bank and credit union reserve cash is a complex exercise: manage it too tightly and your institution may be subject to high-interest Federal Reserve borrowing fees. Manage it too loosely and your firm may lose out on substantial interest revenue from parked cash.
A wholesale financial services provider served hundreds of credit unions nationwide and typically maintained a large volume of cash in reserve to account for member credit union activity. Because the credit unions conducted business autonomously, the organization was constantly challenged to predict members’ cash reserves without any direct control or visibility. However, technology could provide them with the visibility they needed. More specifically, they wanted to apply advanced analytics to predict member activity and drive better returns on reserve cash. To meet this goal, they partnered with New Era Technology Alliance to find a solution.
While the financial services provider could not directly influence credit union spending and borrowing, they possessed one critical element — decades of financial transaction data to support the cash reserve engagement. Company leaders understood there were patterns in the member credit union data based on calendar milestones (payroll activity, mortgage pay activity, etc.) but needed help identifying these regularities in the noise across hundreds of credit unions and billions in cash.
We proposed using machine learning to "read" 18 years of historical cash data to predict the next 60 business days of member activity in aggregate and by cash account. This initiative would provide a discrete view for the investment desk to simulate cash and borrowing needs to effectively partner with finance.
Our team developed a machine learning algorithm that would account for the entire body of transactions while still favoring more recent data. To develop this algorithm, we landed and cleaned data in our client's Azure Cloud platform and gathered success metrics on a variety of algorithms to achieve the desired liquidity aims for the organization. From there, we selected a long short-term memory (LSTM) recurrent neural network.
After achieving the desired metrics for cash management, we moved to the next phase, where our team developed an analytical website solution that: