The survey revealed even more opportunities for improvement in cash forecasting processes—improvements that are necessary to reap the benefits of technology upgrades. Respondents reported that inefficiencies in several areas continue to hamper treasury teams' ability to generate cash forecasts in a timely manner, and will continue to do so unless companies improve and implement best practices across each stage of the forecasting process.

Treasurers must clearly define the goals of a cash forecast—e.g., understanding short-term liquidity, improving decision-making around interest and debt, managing liquidity risk, etc. Then they must capture the appropriate data, process that data to generate an accurate forecast, and ensure that the insights from that forecast support strategic decision-making within the organization.

To improve their cash forecasting, and optimize their use of sophisticated forecasting technologies, treasurers should apply best practices in these four areas:

1. Capturing Accurate Data

Forecasts are only as good as the data that fuels them, and in many organizations, cash forecasting accuracy suffers because some of the data inputs are low-quality. Inaccurate inputs to the forecast usually stem from communication disconnects among the parties providing and using crucial data. For example, treasurers might be collecting historical data for use in ML-generated cash forecasts, but the business unit providing that data may misunderstand how it will be used, meaning the forecast will be off-base from the start. If that business unit does not effectively communicate the likelihood of a particular event or uncertainty, it limits treasurers' ability to make accurate decisions based on that forecast.

One way to reduce the chance that a miscommunication can undercut the effectiveness of forecasts is to remove humans from the equation wherever possible. Treasurers can use automation, especially when integrating systems, to streamline communications and improve the chance that systems will quickly collect accurate data. For example, proactive notifications can alert those responsible for particular forecasts about exactly what types of data are needed, and when.

But technology is not a panacea. Treasurers also need to undertake an education campaign to ensure that everyone providing data to the forecast is invested in the treasury team's goals, understands the value of the data they've been asked to provide, and includes insights into factors that could impact the data. Subsidiaries and business units within a larger enterprise often have disparate data collection policies or nuances to their operations that affect how the data looks.

Only clear communication can resolve such variables into accurate forecasting data. For example, business units might base forecasts on different assumptions and scenarios than those that treasury teams would perhaps more naturally consider as part of their role. Therefore, it's important for treasury to give clear direction about goals and objectives to everyone who's providing forecasts at the subsidiary level.

The treasury group needs to have real conversations with every business unit and corporate function whose data feeds into cash forecasts, including finance teams, department heads, and operations managers. In many organizations, a lack of such conversations remains a problem—one that treasury can and should solve. Communication from treasurers should make clear that forecast data has a critical impact on strategic decision-making, and should make proactive efforts to get all parties on the same page about what management wants and what those generating forecast data should provide.

It's also crucial to continue evaluating data quality after completing a cash forecast. Treasurers should regularly compare forecast data against actual outcomes and investigate the causes of any discrepancies. These investigations may reveal further opportunities to improve communications and/or data collection processes. Addressing those issues will then result in better and more complete data inputs, directly leading to better forecasting.

2. The Processing of Forecast Data

Machine learning is growing in popularity as a forecasting tool because of its ability to handle massive amounts of historical data with very little user interaction. Today's data-driven organizations need to crunch vast volumes of information in order to forecast frequently enough to strategically navigate the uncertainties in their business environment. ML can help treasurers who are in a bind due to the volume of data they're required to process.

Done well, ML is valuable in efficiently identifying otherwise hidden trends. However, ML treasury implementations can quickly go sideways. Treasurers who over-rely on this technology learn the hard way that ML cannot solve all problems, and is only as good as the data it analyzes. That said, ML also becomes more accurate as the volume of data grows.

Multiple factors can undermine the effectiveness of an ML implementation. First of all, treasury teams may fear that the tools will replace them. However, this is not the case: Humans are, and will remain, essential to the decision-making process. It's the combination of company-specific human knowledge and ML insights that enables the most strategic workflows and drives the most accurate business decisions. As ML technology continues to evolve and provides an increasing value to organizations, humans will still be in charge—but making better and better decisions.

Second, machine learning processes are a black box. They require blind faith that the algorithms will behave as expected. But even when the tool works exactly as intended, it may not be sophisticated enough to fully consider all the factors at play in a complex situation.

Consider how a business acquisition or macroeconomic event might shift treasurers' interpretation of the results from an ML tool. Take the pandemic, for example, where treasurers needed to check and recheck forecast assumptions, increase the frequency of forecasts, tune forecasts to prepare for shifts in foreign exchange (FX) rates and other rapidly changing variables, etc. The pandemic—and subsequent global events—have shown how critical forecasting is, and how big of a challenge it can be to execute well.

Treasury leaders can build trust in black-box processes by comparing forecasts from ML tooling with traditional manual spreadsheet forecasts. More important, they need to treat machine learning as a tool in forecasters' arsenal, not as a replacement for human attention. ML-powered forecasting will be less accurate than treasury staff in scenarios that require a human perspective to understand. They provide the most value by giving treasurers the flexibility to adjust forecast parameters quickly.

3. Optimizing Cash Visibility

Cash forecasting that offers only limited viewpoints may lead to dangerously misguided decisions. Treasurers should view forecasts from multiple viewpoints, compare current period actuals with historical data, and use that variance analysis to improve forecasting going forward.

Analysis should be able to determine accuracy versus actual outcomes, as well as changes from one forecast version to the next. Automating these comparisons enables treasury staff to focus on more strategic activities, with confidence that they will receive an alert anytime results fall outside a defined benchmark or threshold—for example, if forecast results are more than 5 percent above or below expectations.

Treasurers also find great value in having the flexibility to visualize and present data in different formats, as appropriate, to meet the needs of specific audiences. Case in point: An executive team will respond more favorably to a high-level presentation of business insights than a spreadsheet data dump, while the treasury team may want the latter level of gritty detail.

4. Adapt to Meet Forecasting Challenges

Yogi Berra said, "It's tough to make predictions, especially about the future"—and that's cash forecasting in a nutshell. While forecasting accuracy will never reach 100 percent, continuous improvement is a worthy and practical goal.

As global treasury professionals like those in Strategic Treasurer's "Cash Forecasting & Visibility" survey pursue new technologies, it's important to remember that technology alone will not bring success, and that best practices play the essential role in capturing, processing, and visualizing data to drive accurate strategic decisions.

Jo Stevens is a senior product manager for GTreasury. She works with the GTreasury executive management team to help prioritize the features that will be on the company's product roadmap, defines scope and expected behavior of new functionality, and works closely with the company's development team. With more than 25 years of experience in treasury risk management, Stevens is passionate about building the best treasury products, responding to where the markets and technology are heading.