Last Updated: November 9, 2021
Finding New Value from External Data and Machine Learning Algorithms
CFOs and FP&A professionals are moving beyond the early adopter phase of machine learning solutions, and incorporating predictive AI into their planning processes. New classes of intelligence and algorithms are enabling financial management to ‘future proof’ their strategic plans and forecasts.
Future proofing is essentially the ability to predict, plan and then protect the outcomes of business decisions. All organizations formulate strategies and plans based on forecasts and predictions of what will happen to their business – it is a fundamental and critical aspect of operational success.
Organizations predict and forecast what is likely to happen to their various lines of business over the short, medium and longer terms. For example over the next 12 months an organization might:
- Forecast demand for 650,000 AC units in North America thru their partner channels
- Project and plan for a market growth increase of 7% in 2022 in its frozen goods category
- Lower its manufacturing output and requirement by 3% next year based on softening demand projections
As we all now realize, COVID has been the root cause of a tremendous amount of market and economic volatility, much of it beyond the control of individual organizations. From inflation, to rising material costs, to supply chain disruption, to changing consumer mobility, to staffing shortages, all of these are likely have a negative or positive affect on business forecasts and plans. Or not. Some businesses are less impacted than others, some positively, some adversely.
If you don’t credibly identify which economic and industry drivers impact each of your lines of business, and then quantify and incorporate them into your predictions and plans, you are basically trying to run a business with a blindfold.
How to Manage Uncertainty
There is now a new way to manage uncertainty, leveraging AI and a 3 step process. This might initially appear daunting, but not with the right tools.
- Step 1 – Identify Indicators
It is now clear that forecasting based on historical performance and internal projections will ultimately result in missed forecasts and market blind spots. There is a realization that strategic planning and ongoing forecasting needs to incorporate external market dynamics. Planning needs to quantify the unique set of market drivers for each line of business, whether that is a brand, or product, or division, or region, or category, or channel. So out of the thousands, maybe millions of external market signals, identify those leading indicators that contribute the most impact, and incorporate them into your planning models.
- Step 2 – Build Models
Take those 5, 10 or 15 leading indicators, especially those with leading or lagging prediction qualities, and build predictive models. By leveraging econometric forecasting methods, cause and effect relationships between what is known as the dependent variable (for example product demand) and economic data results in the creation of a predictive model. With machine learning and AI algorithms, thousands of models based on permutations of those predictive indicators can be created, simulated, and refined at massive scale, to find the one with the highest predictive accuracy, or score. That model will then provide market validated forecasts for the short, medium and long term, anything from 1 month to 5 years. In our experience, predictive models are frequently over 94% accurate, and they can be back tested to determine model accuracy.
- Step 3 – Monitor for Change
Step 3 is critical, because the forecasts are only as good as the model’s predictive accuracy, at that moment in time. If any of the underlying data that contributes to that model changes, then the predictive score will likely change too, possibly resulting in a loss of prediction. Underlying projections for inflation, or government relief policy, or the minimum wage may well shift. If any of these impact leading indicators that are incorporated into your predictive model, then the predictive score will also change. Organizations need to constantly monitor what is known as model health, to ensure there is confidence that market validated forecasts are still valid, and if not then refresh the model, and refine the business plan based on the very latest contributing data.
Predictive AI practices and algorithms provide a scalable and systematic way to apply external and global data within economic based models to provide accurate forecasts for either an organization’s business performance, or even the total projected business for a market segment.
At Prevedere, we help businesses move from the early adopter phase of predictive AI to operationalizing market validated forecasts. Creating this intelligence internally with a data science team can take months, even years. We provide a model-ready economic data repository, a predictive AI modeling platform, and a team of data scientists and economists who can advise and interpret the impact of economic volatility on your business.
A CFO at Kraft Heinz recently stated: “All retailers and CPGs should consider macroeconomic factors to improve their business planning. Not only will forecasting accuracy improve, but go-to-market planners will be fully educated as to what external factors are important to their markets.”
Read more about Kraft-Heinz’s use of AI and macroeconomic data HERE.
Thanks for your time today.