A member’s quest to improve cash visibility begins to bear fruit just as the crisis hits.
“When there is a crisis, treasury becomes very important,” a member of the European Treasury Peer Group said at a recent meeting where he and a colleague shared their progress on a project to improve cash forecasting, in part by viewing cash as inventory to be optimized.
- Cash forecasting, a critical concern for treasury teams at all times, is especially important amid the extreme uncertainty created by the pandemic.
- “We get taken out of the shadows and up to the front,” the member said of crises. “Can we quickly ascertain where the cash is? Can we see it in real time?”
Then and now. When this company’s multiyear journey to digitize treasury began, the team manually prepared custom spreadsheets for once-daily consumption.
- Pain points included balances in nonfunctional currencies and how cash stacked up against target balances.
- Today’s automated daily process generates mobile-friendly descriptive cash dashboards with current and historical data that users can drill into.
- This allows the cash team to be guided by hotspots that require action and to make better decisions on what cash to move where and how to use it.
- “The tool has to be able to tell us where we are today and tell us where we are going to be in three to six months” and whether there will be money for buybacks, dividends and M&A, the member said.
Cash as inventory. The team’s ultimate aim is to automate cash flow forecasting using analytics (AI models) to optimize cash levels, similar to how product inventory can be optimized.
- Next comes the optimization into the automated forecast, plus building in predictive analytics based on external economic indicators.
- By using data sets from the ERP and other sources of financial data, plus AI technology combined with a combination of traditional forecast algorithms, the company is building a hybrid model to forecast short- and long-term cash flows.
- It learns over time the way cash flows work and has demonstrated a high accuracy rate so far.
- Of course, a forecast cannot do everything, i.e., tax hits and unexpected events, but it can help with the amounts that are able to be predicted with pretty good accuracy, freeing up treasury staff to take action on the information rather than producing it, the member said.
Don’t expect overnight success. Not only was it trial and error to see which forecasting algorithms—and in what combination—produced the highest accuracy, it also took time for the machine learning aspect of the tool to work.
- “It was definitely a journey,” the presenter said, adding that it required “lots of exchanges with the data scientists.”
- Treasury went through multiple models, saw that different model mixes yielded different results, and tried “different combinations until we got to the right combo” for the different liquidity items that needed to be forecast.
- It took many iterations until the model produced accurate results and it took time. “But as we progressed the machine learning started to kick in,” the member said.
ERP advantage. In addition to having data scientists, a single-instance ERP certainly helps the company, the member said. “We’re on one instance of SAP so it makes it easy.”
- Companies with multiple ERPs, several instances and a TMS will soon notice that “it can get clunky.”
- Clear ownership and KPIs for the project are also key to success.
Optimizing cash as inventory = savings. By treating cash as an inventory to be optimized in each region and currency, the company will free up balances, earn yield on invested (vs. idle) cash, and optimize or reduce credit lines. In fact, one of the pilot regions managed to reduce idle operational balances by 46%!