Ways to optimize cash forecasting

Does my company have enough liquidity to survive in times of crisis? In the current market environment, marked by high volatility and uncertainty, a growing number of treasurers are concerned about this significant question. Anticipating future cash flows is vital if companies are to manage their finances effectively, make informed business decisions and achieve sustainable profitability in the long term. And yet, many treasurers are already encountering problems with the data basis, as treasury management systems or even subsidiaries are not fully integrated, resulting in a high degree of manual activity and inaccuracies in the preparation of cash flow forecasts. To accurately forecast future cash flows, complex and extensive data sets are required. This is where machine learning (ML) applications, including predictive analytics models, can help automate and accelerate forecasting processes, minimize risks and make decisions based on optimized data and analytics. Likewise, interesting use cases can be applied to simulate various scenarios with variable assumptions in a system-supported manner and thus provide support for financing decisions.

When implementing such solutions, however, there are some pitfalls that need to be avoided ahead of time.

Among the most challenging issues facing companies is providing complete and error-free data as well as the necessary historical information in order to obtain accurate cash flow forecasts. Keep in mind that if the data is incorrect, the forecast may be distorted, which has a negative impact on the prediction quality. Indeed, the quality of the data AI works with is critical to the accuracy of cash flow forecasts. For this reason, companies need to ensure that the underlying data is clean, consistent and complete.

To develop the right model for the company, a thorough understanding of complex mathematical algorithms and machine learning methods is required, combined with an in-depth familiarity with the business model. In Data Science, various modeling options are available, such as ARIMA(X), additive regression, neural networks and others. Apart from selecting the appropriate model, it is also important to strike the right balance between pragmatism and complexity so as to develop and implement models that are easy to understand for end users. This is why it is vital to implement transparent and comprehensible processes.

Optimal and efficient planning takes into account that the tools need continuous training to optimize the predictions, as they work on the basis of algorithms that have been trained on consistent data and model parameters. This means that to maintain accurate forecasts, the data must be validated on a regular basis; for this, training and validation processes will need to be put in place.

Once all the prerequisites for an optimal ML tool are met, the accuracy of estimates for cash forecasts can be improved. ML tools can detect complex relationships that treasurers might miss. That said, it is still imperative that companies use a combination of human expertise and modern tools to optimize forecasts and be prepared for unforeseen events such as macroeconomic conditions.

Artificial intelligence is already being used in corporate treasury for optimizing cash management. Many companies are already leveraging the benefits of artificial intelligence to improve cash flow forecasts and support decisions on liquidity positions. With the aim of automating processes and increasing efficiency, companies should integrate AI-based models into existing treasury management systems and minimize manual efforts, so that 

• Anomalies can be detected using an analysis of patterns and variances
• Automated dashboards and reports can be created
• Complex data from various sources are processed independently
• Data is being analyzed in real time and used as a basis for making decisions

Besides liquidity planning, ML tools can also be used for other purposes in corporate treasury such as: 

• Monitoring treasury risks in an automated way to prevent any unwanted activity
• Capturing and integrating data from multiple sources, such as the treasury management system, ERP as well as the banking tool
• Fine-tuning investment decisions to maximize returns and minimize risks
• Automating payment processes and minimizing manual errors

For such an endeavor to succeed, a number of factors need to be considered. The following are a few important ones:

• The completeness and accuracy of the underlying data: an automated cash flow forecast is based on historical data. It must be complete, accurate and sufficiently detailed.
• Selection of the right drivers: economic and business conditions can change at any given moment. Successfully forecasting liquidity relies on identifying and modeling the relevant business drivers.
• IT architecture and interfaces: the high-performance running of the cash forecast model requires a powerful architecture. The prompt implementation of the required interfaces is critical to the success of the schedule.
• Pooling of business and technical expertise: just as with many digitalization projects, the meaningful combination of specialist expertise (incl. correlations of payment flows in the respective business model) and technical know-how is imperative. This entails in particular straightforward communication and a clear division of responsibilities and mutual expectations within the framework of project management.

Overall, using an ML application can optimize cash flow forecasting by generating accurate forecasts through the use of selected algorithms. Treasurers are in the position to generate better forecasts by using historical and forward-looking data and continuously validating the models with new data and characteristics, thus obtaining a solid basis on which to make decisions. ML tools can also streamline manual interventions and automate the planning process, which saves time and increases accuracy. When all is said and done, adopting ML tools can help minimize the risk of bad decisions and ensure a company's financial stability and liquidity. That is, of course, provided that the data basis is correct. For only when this is a prerequisite can the use of ML tools be a useful addition.

Source: KPMG Corporate Treasury News, Edition 131, April 2023
Authors:
Börries Többens, Partner, Finance and Treasury Management, Corporate Treasury Advisory, KPMG AG
Nico Krämer, Manager, Finance and Treasury Management, Corporate Treasury Advisory, KPMG AG