Manufacturers have spent years or decades refining their forecasting capabilities. So how good are they at anticipating demand? Recent data published in the Journal of Accounting Research has some surprising answers.
Leaders at 24,000 factories were asked to predict what their shipments would be in 18 months. Only 22% of the forecasts were anywhere close to correct, with 23% falling short of the lowest estimate, and 20% coming in above the highest anticipated increase.
These numbers show that the vast majority of forecasts are incorrect. What’s more, they miss the mark by wide margins.
Inaccurate forecasts can have expensive consequences, including stockouts, excess inventory, and missed service levels. All these costs grow even larger in times of market volatility, when forecasts are harder to get right and tend to be wrong to greater degrees.
Recent history has been defined by unpredictable events, making the forecasting trap painfully clear to many in manufacturing. Forecasters want and need a better approach that incorporates uncertainty rather than ignoring it. Fortunately, modern ERP solutions offer a viable solution for the first time.
Where Traditional Forecasting Goes Wrong
The research cited above highlights where so many traditional forecasts go wrong. They make a prediction about a point in time somewhere in the future. Then conditions change while the forecasts stay the same.
For example, a manufacturer believes that demand will grow by 10% over the next two years. Then new tariffs are announced that increase costs and cause demand to crater. Or a manufacturer intends to increase production by 20% over the next six quarters, until the outbreak of a global pandemic throws supply chains into chaos and makes materials scarce.
Both of these scenarios are real and fresh in the minds of manufacturers. Furthermore, forecasting disasters are not hypothetical. Nike famously lost $100 million and saw its stock plummet by 22% due to forecasting errors in the early 2000s.
Effective forecasting must be able to predict the future and then evolve along the way. The first part is difficult enough, but the second part has always seemed impossible. Now both are undergoing massive improvements.
Introducing Probabilistic Forecasting
Traditional forecasting is deterministic, meaning it frames predictions in concrete terms—e.g. demand will increase by 7%, or parts will take 90 days to acquire. It works well when there’s enough historical data to make reliable predictions, combined with stable markets. It breaks down when it encounters the unknown or unexpected.
Probabilistic forecasting generates a range of possible outcomes along with the probability of each one—e.g. there’s a 60% chance demand will grow by 10%, a 20% chance it will be higher than 12%, and a 20% chance it will fall below 8%.
Probabilistic forecasting has many benefits, primarily that it promotes thinking about multiple possibilities while framing outcomes in relative terms. Rather than expecting something definitive to happen, leadership looks at different possibilities and develops contingencies automatically. It treats uncertainty as likely.
The disadvantage, however, is that probabilistic forecasts are harder to develop, both because they’re new and because they’re more complicated. That fact has kept some manufacturers from embracing this methodology or trusting the forecasts. Modern ERPs eliminate those obstacles.
How Digital Twins Enable Probabilistic Forecasting
One of the most advanced and advantageous capabilities of modern ERP systems like Infor is the ability to create digital twins: detailed copies of a manufacturing environment or supply chain that operate exactly like the real thing. One application of digital twins is the ability to model how different scenarios would affect production.
Essentially, digital twins are a way to bring probabilistic forecasts to life. They make it easy to see the dynamic outcomes of various scenarios, regardless of whether those scenarios are likely or unlikely. Just as importantly, digital twins can be fed real-time data to better understand how it impacts existing forecasts. For example, if demand is projected to increase by 10%, but then key materials are delayed by 30 days, digital twins reveal the resulting effect on production, thus allowing forecasts to continually adapt.
Many ERPs that have digital twin capabilities also have the ability to automatically or easily generate probabilistic forecasts. And when both those capabilities work together, manufacturers can confidently forecast the future. Not because they know exactly what will happen. Because they can make educated guesses, then adjust their assumptions in response to real-time data. That way, no matter how volatile markets become, manufacturers are prepared for anything.
There’s only one word for probabilistic forecasting and digital twins working together. Game-changer.
Switch to Something Better
As manufacturers increasingly move away from deterministic forecasting, and with market volatility showing no signs of stopping, manufacturers that don’t adapt are at a serious disadvantage. Fortunately, enabling probabilistic forecasting and implementing digital twins have never seemed simpler.
Guide Technologies has helped a growing list of manufacturers upgrade their ERP to improve forecasting. We have also helped those clients update their business practices to take full advantage of smarter forecasting. The result is companies that are stronger in every way, all thanks to having a modern ERP at their center.
Take advantage of our experience and expertise to get more from your ERP. Contact Guide Technologies.
