FP&A's New Frontier: Algorithmic Forecasting

Oct 19, 2016

Nilly Essaides, Director & Practice Lead, Financial Planning & Analysis, AFP

Having clearer insight into what lies ahead is what sets top performing organizations apart from the also-rans. The greater the volatility in the business environment, the fiercer the competition, the more important it is for financial planning and analysis (FP&A) to come up with better ways to see around the corner. The good news is that there are emerging tools that rely on machine learning and artificial intelligence (AI) to enhance the view of the future. It sounds a little bit like science fiction but these tools are available today and more are being developed.

According to Gauthier Vasseur, vice president at Trufa Inc., a predictive analytics vendor, up to now, finance pulled data and fit it, at best, into a business data mart or most often into an Excel spreadsheet or a cube, and then used a reporting tool to throw a graphic on top of it. That approach is simply not feasible when it comes to millions of data points without going through aggregation that unfortunately erases important details. “The human brain cannot comprehend millions of data points. You need the inclusion of AI algorithms,” he explained. “New approaches can identify statistical patterns, trends and driver correlations, enabling an FP&A analyst to handle the results.”

“You don’t want to wake up in the morning and keep being surprised,” said Mario Manna, co-founder of Nightberg, a geopolitical strategy firm which produces forward-looking country-specific and market reports. “There are ways finance executives can anticipate what’s around the corner.”

Nightberg collaborates with another predictive analytics firm, Predata, which monitors huge amounts of data from online conversations across all languages and countries. The analytics platform then gauges the level of digital discussion around particular people or subjects. It uses machine learning to pick up the metadata, i.e., the number of people engaged in the conversation and the level of argumentation to assess an intensity level and construct a volatility index. It then regressed the index against over 200 different types of past events to come up with a statistical probability of the events taking place in the future. “The technology gives FP&A the ability look forward and assess the probability of geopolitical risk and its impact on the business,” said John Alfieri, vice president. “It’s a valuable insight for financial planners looking to prepare their organizations for what lays ahead.”

Entering endless dimensions

AI steps in when humans lose their footing.

When humans look at a spreadsheet, they can validate data relationships that are known to them already or see some that really stand out, but only to a certain level of detail, according to Henry Wong, a consultant with Global Treasury Partners. Technologies, like IBM’s Watson, use machine learning and AI to simultaneously investigate the relationships between endless sets of data, structured and unstructured. Even the more advanced predictive analytics tools today rely on cubes, which have a limited number of dimensions (typically three). “But tools using AI take a fundamentally different approach that expands the possible number of perspectives and can identify hidden gems,” Wong said. “Instead of looking at just the 50 fields of data you’ve uploaded, you can combine your data with other (external, unstructured) data and find macro observations you didn't think of,” Wong said. “You don’t develop a single algorithm. That would be a specific use case. Using AI to predict based on a goal can produce many viable scenarios or probabilities in a business context.”

Looking deeper, inward

Algorithmic forecasting is not just about having a better look at outside factors; it’s also about having a more nuanced look at what’s going on within the organization.

According to Vasseur, AI can identify the drivers to performance at the transaction level where the human eye would just see an ocean of information. “We can measure everything at the detail level. And with statistical models, we can predict what would happen if you do that transaction again,” he explained. “With pattern recognition and clustering of these operations, we can understand how business drivers correlate to financial performance. These are the same algorithms that are used to predict the weather or one’s next purchase at an online store. “That’s the next step for FP&A after business intelligence. It’s about preemptively taking action to affect financial performance.”

Vasseur insists this is not a fad or a fluke. Algorithmic forecasting is an answer to a real business need and a natural evolution in the trend toward using data more effectively. “The new era of finance analytics is about telling the future using pattern recognition and bringing together operational and financial data and correlating the outcomes and how they’ll drive performance. It’s about asking more why questions and getting at the core of things. BI stops at two to three whys. AI asked the next two to get to the famous five whys,” he said.

The lesson for FP&A is this: many existing financial planning applications and even BI tools tout their predictive analytics capabilities, but almost all of them rely on old technologies that work only by first predetermining the structure of the data and thus leaving out a lot of the information. New AI-powered technologies are opening a new window into the future. They can help refine the forecast and offer a clearer picture of how events and business drivers will affect the organization’s future financial performance. That’s going to allow FP&A to give the organization its biggest advantage: intelligence on how best prepare for what’s lies ahead.