Cash optimization for ATMs and branches

Stock enough cash, not more than you need

Every ATM and branch has its own rhythm. Demand jumps at the start of the month, before holidays and around local events, and it shifts as customers move to mobile payments.

Fixed stocking rules cover the peaks by overfilling, so cash sits idle and carries a funding cost. When a machine still runs dry, your team arranges an unplanned delivery and a customer walks away. Interest or funding costs on idle cash vary by market, but the pattern is the same everywhere: money sitting in a machine is money not doing anything else.

Your cash operations team plans routes, orders and branch staffing by hand from spreadsheets. A forecast is only useful once it reaches that plan in time.

How cash optimization works

Your team trains forecasting models on the cash history you already hold, reviews the forecast for each location, and turns it into a plan that operations approves.

Step 1: gather cash history and calendars

Start with daily withdrawals and deposits per ATM and branch, plus calendars for paydays, public holidays and local events. Add location attributes such as footfall and nearby merchants where you have them.

Step 2: forecast demand for every location

H2O Driverless AI builds time series models with lag, rolling and calendar features for each machine or group of machines, and checks forecasting accuracy on held-out weeks before any plan depends on it.

Step 3: turn the forecast into a plan

An agent reads the forecasts alongside branch traffic, transaction volumes and wait times, then drafts replenishment orders and staffing suggestions for your operations lead to review, edit and approve.

Step 4: deploy and monitor

H2O MLOps deploys the forecasting pipeline and tracks model drift as payment habits change.

Connect the forecast to the daily order

A cash forecast saves money only when it drives the daily order, so connect it to the replenishment plan from the first pilot rather than treating it as a standalone report.