Advanced data analysis and machine learning were used to optimize operations and the supply chain, resulting in significant savings for a client who manages and services over 45,000 ATMs in the US.
Client Needs
ATM Machines were stocked at fixed frequency (daily, weekly or monthly) using third-party Armored carriers
In case of ATM out-of-cash, emergency cash-order is placed, client pays daily interest on cash borrowed from banks
Total spend was over $50M per year in interest
Armored carrier costs (for scheduled and emergency fills) were on the rise
The challenge was to reducing stock-outs and costs at the same time
Solution
Reduced outstanding cash (and interest) through safety stock optimization - used forecast error in addition to historical data to compute safety stock
Reduced safety stock by reducing overall forecast error using machine learning algorithms to capture demand patterns across groups of machines
Optimized ATM Fill Frequency using linear programming model comprising interest vs fill cost using exponential smoothing and probabilistic demand modeling
Reduced Emergency cash-orders using probabilistic demand model and Executive Management Control System
Results
Reduced outstanding cash by nearly $270M while reducing stock outs by 18%
Reduced interest spend and armored carrier costs by approx. $5.8M annualized in less than six months.
Our team used a blend of traditional operational improvement techniques, supply chain optimization, advanced statistics and machine learning to rapidly drive operational improvements.
These improvements generated substantial financial savings in less than six months.