How to turbo-charge your forecasts and decisions with machine learning models

Clive Humby, a British data scientist and innovator, said “Data is the new oil” of the 21st century. But in the same breath, he made an even more important observation: “It’s valuable, but if unrefined it cannot really be used.”

In my article, “The future of finance is data driven,” I outline how it is crucial for finance teams to leverage all data available to them—financial, operational, and external—to drive the right decisions for the business. If data is oil, and finance is in the driver’s seat of decision making, the oil needs to be refined and turned into energy. According to an article in The Wall Street Journal, about 64 zettabytes of data was created last year alone. That’s 64 followed by 21 zeros. But with the explosion of data, how can finance get their arms around it be “refined” into something usable, to help with planning and decision-making?

Enter machine learning (ML), which is ideally suited for dealing with large volumes of data— in fact, the more data, the better, since more data helps to improve the accuracy of the learning models. In finance, ML has the potential to improve forecast accuracy and uncover new insights.

Imagine being able to refine sales forecasts using learning models based on multiple internal and external drivers, such as historical sales, price, promotions, GDP, climate, and industry volumes. Using drivers that are more business-relevant can generate more accurate sales forecasts.

But all these ML models don’t have to be built from scratch. Many organizations have ongoing data science projects where they build machine learning models. The opportunity is for finance to incorporate an appropriate subset of these models within the context of their planning activities, rather than reinvent the wheel.

Oracle Cloud EPM now supports this capability, which we refer to as “Bring Your Own Machine Learning Model” (BYOML).

Oracle Cloud EPM lets you import ML models built in popular languages such as Python. These customized ML models can work automatically behind the scenes to run predictions without manual intervention. Planning professionals simply gain the ability to use existing ML models in their planning and forecasting process.

Generate more accurate forecasts

BYOML helps EPM planning users take advantage of advanced machine learning techniques, generating more accurate forecasts based on multiple drivers without skilling up on data science.

Leverage data science investments

Often the biggest challenge in data science projects is to operationalize the ML models built by the experts. With BYOML, you can deploy these specialized models in Oracle Cloud EPM. This gives business users the ability to run the models and consume the insights within their everyday business activities, helping to complete the last mile of data science initiatives.

Ease of setup and on-going maintenance

With BYOML, you can map ML models to the Cloud EPM Planning process through a wizard. Therefore, ML models can now be easily operationalized using the existing skill set of EPM administrators.

BYOML is another important capability delivered through Oracle’s Intelligent Performance Management initiative. By leveraging all relevant drivers and advanced ML models from data scientists, it has the potential to deliver substantial business value. Imagine HR being able to accurately predict attrition; or marketing teams optimizing marketing mix modeling to drive better return on their investments; or finance performing predictive cash forecasting for optimal working capital management. These improvements would have a material impact on your business.

I recently had a conversation with a large telco customer who used machine learning to plan the roll out of their 5G masts. To ensure a successful rollout from a cost, coverage, and customer satisfaction standpoint, they used geographic, topographic and population data to predict the optimal locations for their 5G towers. This use case illustrates the potential value of using ML models in critical business decisions.

Business planning has evolved beyond simple extrapolation of historical numbers to incorporate predictive techniques that consider all available data—as well as ML models developed within and beyond the finance function. If you want to be a data-driven finance organization powered by the “oil” of the 21st century, now is the right time to consider how you can turbo-charge your forecasts and decisions with machine learning.

Learn more about ML-driven forecasting and BYOML in this webcast.