All-Seeing AI: The Facts About Real-Time Cash Forecasting

Published: August 26, 2021

Download this articles as a PDF
All-Seeing AI: The Facts About Real-Time Cash Forecasting
Sarah Eckstein picture
Sarah Eckstein
Product Manager, Coupa
Tom Alford picture
Tom Alford
Deputy Editor, Treasury Management International

A real-time view on corporate cash and liquidity positions has become crucial to optimal decision-making. Yet for many companies still struggling to aggregate their siloed data sources, this might as well be a notion from a work of science fiction. Prepare to open your eyes, says Sarah Eckstein, Product Manager, Coupa. Here’s why.

There’s a problem with cash forecasting. And the more it is relied upon to give companies the best chance of surviving, let alone thriving, the worse the problem gets. If that sounds like an end-of-days scenario, be assured that there is a solution, and what’s more, it’s one which most companies could be moving towards now.

The real challenge concerns the data. In stressed times, such as during the current pandemic, the importance of knowing where the company stands now, and where it will be in the next hours, days, weeks and months, increases significantly. Having the right data at the right time enables reaction in a timely manner to a broad sweep of mounting risks, not least with foreign exchange (FX) and counterparties, says Eckstein.

Removing silos

The trouble with financial data is there are many commercial scenarios that push and pull it in the wrong direction, often by creating silos, inconsistent formats and multiple versions of ‘the truth’. This styles data management, and ultimately data value, as a matter of serious concern, especially when it relates to cash forecasting.

“Businesses grow and change over time, and the simple dynamics of being in business are making it almost impossible to rely on historical data to manually derive a forecast,” explains Eckstein. At the highest level, she believes that the key challenge arising from data mismanagement is simply gaining insight into overall cash positions, without which decisions become mere guesswork.

The exact cause of the issue here could stem from a lack of cash visibility across bank accounts, or poor cash forecasting automation where few insights for liquidity management are delivered. It could even flow from bias in forecasting data, where business units providing target rather than actual figures are inadvertently degrading overall accuracy.

Regardless of cause, maintaining manual forecasting across a disparate and disconnected data landscape means that treasury will continue to struggle to prepare timely cash flow forecasts. In a stressed trading environment, time may not be available to request and assimilate data from other functions, departments, or entities notes Eckstein. “Decision-making can fail as a consequence.”

Of course, the nature of forecasting is that it is only ever a ‘best guess’, but logic then dictates that the best available data sources and tools should be deployed in order to achieve the desired efficacy. As the pressure to improve forecasting accuracy ratchets up with events such as the pandemic destabilising the economic environment, so issues created by ineffective data management grow increasingly obvious, and their effects ever more damaging.

Indeed, says Eckstein, because a cash flow forecast is a fundamental requirement of being able to make fast and effective liquidity decisions, any obstructions delaying the treasurer’s precise understanding of the organisation’s cash positions could have serious consequences upstream.

Under-funding, for example, could result in the business being forced to go to market, exposing it to higher costs. “To prevent an unexpected shortfall coming from inaccurate cash flow reporting, a company may find itself regularly borrowing at higher rates,” she explains. “The cost of carry of those rates alone is an unwarranted expense.”

In the end, she argues, the more accurate a forecast is, the more a business can reduce those operational costs. “For that to happen, it requires all the right financial information to be on hand, so that decisions are based on transparent data points instead of partial assumptions or out-of-date data.”

Real-time progress

The solution is real-time visibility and transparency. “For organisations with multiple subsidiaries, currencies, and accounts, instantly knowing the cash positions, being able to move funds with a few clicks, and run scenarios to see how long funding will last if income or expense categories are adjusted, is a powerful proposition,” says Eckstein. It is also very much part of a fully realised Coupa platform.

Every business knows that cash use and demand are always changing, especially in stressed times; this is why they are difficult to predict. Technology can provide answers. Indeed, while Eckstein notes that automated cash forecasting tools typically rely on comparative statistics, one of the most significant and promising lines of thought in the cash forecasting space is how artificial intelligence (AI) can assist.

She confirms that Coupa is fully engaged in this field and is now piloting its own software with corporate clients. “Our early adopters show that AI is revolutionising the process of cash flow forecasting; it’s resulting in greater efficiencies, profitability and productivity.”

Based on analysis of historical transactions that have been automatically imported into the Coupa treasury management system (TMS), its AI system uses proprietary algorithms, developed in tandem with Coupa’s 2020 acquisition LLamasoft to generate forecasts. The advanced capabilities of the system are based on machine learning (ML) software that, to date, has managed to increase forecasting accuracy to at least 80%. And its accuracy is complemented by speed.

“The big advantage is that our system automatically generates a new rolling weekly forecast,” continues Eckstein. “Treasurers can immediately analyse output in the context of risk management, working capital and funding requirements, for example, and quickly begin their decision-making processes.”

In purely practical terms, the advantages of having rapid access to relevant data, and facilitating almost instant analysis and decisioning off the back of that, is of huge value in itself. However, more is derived from Coupa’s approach to treasury, which is based on its concept of ‘community intelligence’. This establishes an ecosystem in which clients can share information and data within and across geographies and sectors.

Within the full Coupa solution, where spend management is directly connected to the TMS within a single platform for example, not only is collaboration across different departments considerably easier, but also the information necessary for forecasting is made far more accessible, says Eckstein.

With a globally distributed customer base in excess of 2,000, Coupa is able to collect and process forecasting data from each. The power of this data when aggregated – and anonymised – is immense both in terms of helping to fine-tune the forecasting algorithms and also assisting its clients to construct useful benchmarks and key performance indicators (KPIs).

Using this level of immediate, automated data delivery, the platform enables users to work through numerous stress-testing scenarios, workflows, and ‘actual versus plan’ analyses to “boost confidence in their short-, mid- and long-term business health”.

It also means businesses can prepare for cash surplus or shortfalls, reducing borrowing costs and enabling more effective investments. Of course, in current conditions of market volatility, interactive forecasting also enables rapid reactions to business cash needs, which could be a vital differentiator.

Getting results

Preparing for a boost to forecasting results of this magnitude will require treasury to communicate to the rest of the business just how much time and effort can be saved by using modern software to automate, says Eckstein. “It should be explained that the time saved may be invested in more strategic projects, and also that the adoption of process automation reduces incidences of error which, in forecasting terms, means more stable and valid data output for decision-making.”

With the Coupa AI-based forecasting solution being rolled out to early-adopter customers, the benefits are real-world. Eckstein says user feedback is enabling an iterative approach to development that will see the whole customer base eventually benefit from solutions that have been cast in the heat of frontline activity. “We are already planning the next steps to build out the general availability of our forecasting solution, and we are beginning to implement more value-adding features.”

The next phase will leverage the synergy of the full Coupa solution, Eckstein explains. As an example, by incorporating into the AI solution invoice data that is already available on the platform, it will drive further advances in cash forecasting results.

With data volume, speed, accuracy, and timeliness all coalescing around a single platform, Coupa has helped AI find a natural home in the world of treasury. Treasurers should consider leveraging the opportunity, not just to help mitigate the risks of various crises, but also to continue optimising cash visibility and forecasting in the face of increasing competition and stakeholder expectations. By doing so, real-time forecasting moves even further from being a one-time work of science fiction, into the realm of science fact.  

Sarah Eckstein
Product Manager, Coupa

Eckstein joined Coupa, formerly BELLIN, in 2016. Her work in one of the development teams within Coupa Pay currently focuses on enhancing forecasting functionalities, especially integrating artificial intelligence forecasting into Coupa Treasury. She is also working on the reporting solution within Coupa Treasury. Eckstein studied mathematics at the University of Freiburg, where she also worked as a research assistant while completing her doctorate.

Sign up for free to read the full article

Download this articles as a PDF
Article Last Updated: May 03, 2024

Related Content