Predictive analytics: Its role in treasury and finance

Predicting the future may not be that difficult with the help of ‘big data’.  Sophisticated system design now enables financiers to capture more and more data and it can be put to a myriad of beneficial end uses.

Predictive models provide a framework in which to design a window to look through, while ‘big data’ can provide the tools needed to build the window. This is predictive analytics and this article will examine the area and its possible end uses.

It is always advisable to start with a definition and an examination of its scope of operation, so what are the key tenets of predictive analytics? There are many different interpretations but my favourite is to essentially describe it as statistical modelling using historical data to look for patterns in data.  The patterns are then used to identify trends, and the trends are most commonly used to predict outcomes:

Figure1: The ‘modus operandi’ of big data.


gather data ⇒ identify data patterns ⇒ identify trends ⇒ predict outcomes


Source: Actualize Consulting.

It is important to remember that a model is only as good as its inputs.  Without meaningful data, you could have the most sophisticated model in the world but the results yielded would still be off, or even worse, false if you didn’t feed it with accurate data.  Before deploying predictive analytic models, therefore, it is very important to understand what data is available and if it is actually usable.

One key assumption is that the past will predict the future, but this assumption must be periodically tested and challenged, with the understanding it may not always be true.

The practical application of statistical models casts an extremely wide net, so I will focus specifically on how treasury and finance departments can and are using predictive analytics, as well as the software enabling them to spot important patterns in their data.

How are predictive analytics used in treasury?

There are three main uses for predictive analytics in treasury, which I have listed below and will examine in turn:

  • Cash collections
  • Cash forecasting
  • Analysing market data.

Cash collections: The predictive algorithms in use today are helping treasury and finance capture cash faster, thus improving cash collections, while reducing risk. Enterprise resource planning (ERP) systems can feed customer data not only to the credit/collection system but also separately to the predictive analytics model.  The model then returns a consolidated risk score, based on observable data and basic assumptions.  Users can then determine which customers are ‘high risk’ and who they may need to monitor more closely, or indeed contact pre-emptively to ensure invoices will be paid.

As an example of the value that can be captured by using a predictive model, theoretically an invoice that is not past due may be flagged as a potential payment risk. The model would analyse past payment behaviour from the customer.  If irregular payments began to trend to a worsening position, shifting from paying early, to on-time and a few that were paid late, the behaviour could be an early warning sign of cash flow risks that the analyst may not have been able to identify as quickly. In fact, a well-known treasury management software company already offers a predictive model that works with their payments and collections system, allowing accounts payable (AP) and receivable (AR) teams to analyse customer data based on observable data, flagging potential customers for payment risk. This predictive capability can help bring cash in and risk down.

Cash forecasting: Treasury and finance can also benefit from the speed, intelligence, and forward-looking capabilities of predictive analytics software to enhance their cash forecasting.  Cash forecasting can become more dynamic by incorporating recent and relevant events.  Examples include supplier contract pricing changes, financial market changes, and large customer base changes.  Software vendors provide in-memory engines to access large data stores and predictive models to aid in cash forecasting. These forecasts can be more dynamic and capture events faster than monthly, quarterly or yearly updates. Using archives of stored data and running algorithms or macros in excel does not produce the same result and is therefore a less robust alternative for building a forecast.

Analysing market data: Not only can predictive models alert companies to payment risk, and help provide up-to-date cash forecasts, they can also prepare treasury for adverse liquidity and market conditions.

If predictive modelling flags an impending volatile market shift, what is the responsibility of the treasurer?  Should they lock in funding now? Should they layer more hedges? This is why a popular market data company developed a predictive modelling tool specifically for analysing volatile market shifts.  Historically, banks and hedge funds have been the organisations with the most sophisticated algorithms guiding their decisions, but this company has now made it possible for treasury to navigate and mitigate large loss situations.  Credit and sovereign statistical risk models evaluate macroeconomic, political, and local market data to determine whether countries will default on debt, change interest rates, and whether sovereign currencies will appreciate or depreciate over multiple time horizons.

Why is big data a big deal?

Predictive analytics are becoming more popular among the treasury and finance communities, and software vendors are offering more sophisticated products, in order to help companies:

  • Make informed decisions they otherwise may not have been able to make
  • Forecast cash and attribute liquidity needs better
  • Mitigate impact of significant liquidity and financial markets events.

Treasury and finance departments that use predictive analytic models are gaining a competitive advantage in their respective communities.  They can capture and deploy cash with greater understanding of market climates, while garnering additional intelligence surrounding future cash liquidity needs from customer, supplier, and market events.

With predictive analysis of big data, corporate treasury can come much closer to actually predicting the future – through science, not magic.


Related reading