Using technology to improve forecasting

A CGMA Future of Finance report highlights the role of machine learning in streamlining processes.
Using technology to improve forecasting

Editor's note: Thousands of financial professionals and hundreds of organisations across the world participated in the Future of Finance research led by the Association of International Certified Professional Accountants. This excerpt, gleaned from research interviews and published in the new Future of Finance initiative report Re-inventing Finance for a Digital World, looks at how one company rolled out machine-learning technology in the revenue forecasting area and regained finance function time. The full report is available at

In the course of our research, we talked to a multinational company whose finance department was trying to improve its revenue forecasting. Their goals were to improve the accuracy and frequency of finance forecasts. They hoped to provide a stronger, unbiased baseline forecast, on a more frequent basis, that would enable the finance function to respond more swiftly to business issues.

As it was a technology company, the CFO was able to connect with the vice-president of machine learning. Together they came up with a technological concept to help improve the accuracy of finance function forecasts.

Massive reductions in time and effort

Before the concept of machine-learning revenue forecasting was introduced and tested, revenue forecasting was a complex affair using spreadsheets involving 800 analysts across many business channels. It could take up to three weeks to process and generate a quarterly forecast.

Over seven revenue-forecasting quarters, the new machine-learning system was run in parallel with the traditional, human-compiled CFO forecast. The new trial system reduced the process from three weeks and 800 analysts to just two days involving the input of just two people.

By the end of the trial, the machine-learning system had also significantly improved forecasting accuracy. In addition, it had reduced the human input into a quarterly forecast from 16,000 to just four workdays.

The machine-learning system now provides global analysts with an accurate forecasting benchmark that is comparable to internal human-generated forecasts. It has given the company more confidence in the forward-looking revenue ranges it provides to external stakeholders.

Freeing time to add value

Of course, the introduction of machine-learning revenue forecasting led to the displacement of tasks — but it did not lead to the displacement of people from the company. The time saved is now being used to enable the finance function to respond to the key issues that add more value to the company.



"Watchlist for Future Finance Innovators",FM magazine, 28 January 2019

CGMA Store products

As part of 2019 member benefits, CIMA is offering the Digital Mindset Pack, which is made up of more than six hours of free CPD focused on digital competencies such as blockchain, data analytics, and robotic process automation. Find out more at

Data analytics resources,

Robotic process automation resources,