Affecto Oyj : Can a machine learn financial planning?

Finance functions have traditionally operated in monthly cycles while operative activities happen daily. As pretty much anything in operations can be measured, analysed and acted upon, finance is at risk of losing its relevancy if it only produces financial reports once a month. In other words, the gap between the finance view and the real operative world is becoming chasmic.

With that reality today, can a CFO still be an active business partner and advise the CEO in financial, operative and strategic terms?

Making sense of data

Clearly, CFOs and Financial Planning & Analyses (FP&A) leaders should find a way to sense the operative world and make financial sense accordingly. There are many ways to do so, but by adopting Machine Learning to make sense of increasing amounts of operative data, they can regain relevancy as active business partners.

Machine Learning (ML) is a subset of Artificial intelligence (AI). It is, in short, a technique of data science that helps systems learn from existing data to forecast future behaviours and explain trends and outcomes. Machines, such as IBM Watson, can use data to explain complex phenomena and learn to forecast emerging developments and trends. This will improve financial decision-making as well as the ability to react.

Machine Learning and Artificial Intelligence are already widely used outside of finance. Kone (elevator and escalator company) gathers data from elevators using IoT (Internet of Things) sensors to improve the user experience and optimise maintenance. The Hospital District of Helsinki is collecting new data from patients to forecast infections and other dangerous situations that a doctor would not be able to observe in time. Why should we not have similar examples from finance domain?

Business benefits and outcomes

Advanced and predictive analytics are already used in consumer business (e.g. hotel and airline ticket pricing), but Machine Learning is only just being introduced more widely in financial planning. There are major benefits to be gained, since companies have huge amounts of internal and external information which they are still are unable - or lack the time - to exploit.

To begin an ML journey, one will need a machine such as IBM Watson, and a machine user, such as a controller or business user. Human intuition is still needed for context and in implementation, but a machine can handle data analyses and ensure that decision-making is based on all the data available. Machine learning can become a reliable assistant, ensuring good fact-based decision-making while avoiding human error and false assumptions.

Can a FP&A trust his or her own forecasts? Sales forecasts can be based on information from either a sales team, or they can be challenged and improved using machine learning. A statistical (machine learning based) model can be built to explain order book dynamics in B2B business activities. This leads to more reliable sales forecasts than relying on the sales team alone. It also reveals how external factors affect sales, and which customers and activities are key to achieving sales targets.

Consumer goods producers can quickly evaluate the success of a new product based on social media conversations and formulate reliable sales forecasts. In such cases, information produced by machines is unbiased and real-time, and thus it helps to make better decisions faster. Credit controllers can use machine learning to ensure sufficient liquidity and make cash flow forecasts. Machine learning models can provide a clear, intuitive and transparent regression and decision tree models. Thus, cash flow forecasts become actual action plans to improve cash flow, since customers needing special attention can be identified in time.

A CFO which uses machine learning is free to focus on the most important topics, such as strategy-based resource allocation and business partnering, with the CEO.

Culture and legacy systems as on obstacle

According to a report by KPMG in 2016, every second senior executive is losing trust in data. A real concern is whether, as data volumes continue to grow with digitalization, finance will be able tell where figures are coming from. Financial planning is still too often done manually using Excel spreadsheets, and much-needed history data is fragmented across multiple systems. This makes it difficult for an analyst to compile a forecast, and the overall process of doing so is time consuming. As an outcome, a plan, budget or forecast may be already outdated by the time it's finally ready. With such fragmented architecture and outdated forecasts, the reason for executives' mistrust of data may not be much of a mystery.

The way forward

Machine learning brings more intelligence and efficiency to financial planning, and it can be achieved with cost-efficient cloud-based services. Financial teams can use advanced tools and algorithms to make internal and external phenomena visible - and, most of all, to forecast the impact of such phenomena on the company's bottom line. I believe executives' trust can be regained by starting the Machine Learning journey with visual and transparent models such as regression and decision trees.

Modern CFOs can marshal increasing amounts of information by using machine learning to sense their operative business and make financial sense.