Building a Cash Forecasting Solution? It’s All About the Data.
You can’t read the news without seeing daily reports of the economic climate with rising rates, snarled supply chains, increasing volatility, and the shadow of a looming slowdown. I’m not taking bets on the likelihood of a recession, but I do know that this uncertainty is top of mind for corporate treasurers.
Recent Celent research confirmed a trend over several years that better visibility into cash and cash forecasting remains a top priority for treasuries of all sizes. Across the market, 37% of corporates cited real time forecasts as a service that would bring value to their operations. This includes 42% of those with annual revenues greater than $10bn. If anything, the current market conditions may have increased the importance. Add in the plethora of new real time payments (RTP) systems, and treasuries have more methods than ever for sending and receiving funds, all impacting cashflow.
From a bank perspective, the development of cash forecasting solutions is a logical extension of traditional information reporting capabilities. Essentially, this takes the basics of a day-to-day cash position and forecasts the balances 7, 30, 60 or more days into the future.
Celent recently released a report covering the solutions of several leading banks, Breakthrough Innovation in Cash Management: Bank Pacesetters and Point Solutions Partners in Cash Visibility and Forecasting. If banks can deliver it, cash forecasting can provide a significant value-add to clients, and a competitive edge in the progression toward more sophisticated treasury solutions.
Of course, to be of value to a treasury client, the forecasts also need to be accurate! Rather than straight-line projections that many treasuries perform in Microsoft Excel (still a dominant cash positioning tool), leading banks apply machine learning models to create the forecasts. Unlike a cash position, a cash forecast is a model-based prediction with a probability indicating the level of confidence. If the predictions are not accurate or reliable enough, clients will see limited value and the bank will gain limited adoption on a significant investment. So, what is the key to the most accurate forecasts possible?
Developing or acquiring advanced analytics and ML technologies to deliver the forecasts into the future isn’t by itself a silver bullet. Simply put – the primary factor is data and data capabilities at scale. Models need data, perhaps billions of rows, depending on the number of clients and the sources of data. This adds multiple layers of complexity when launching analytics-based solutions to clients. For example:
- What data sources and data elements are available to the forecast models?
- How much data is required to calculate forecasts?
- What historical data is available to train models?
Beyond the actual data, the following data-use considerations will also impact a bank’s ability to successfully deliver cash forecasting solutions:
- Ensuring legal use of data for each geographic jurisdiction.
- Identifying and managing the risks of poorly performing ML models.
As banks embark on a journey toward forecasting and new intelligent treasury solutions, they must evaluate options based on existing and planned data platform capabilities.
Is your treasury solution strategy integrated with your data strategy?