If Forecasting Is Never 100% Accurate, What’s the Point?
The point of demand forecasting is not perfect accuracy but the process of understanding the underlying assumptions behind the forecast, monitoring drivers of demand, and making updates quickly.
Forecasting is an integral part of supply chain management. It is the process of predicting future demand for products or services, allowing organizations to allocate resources, optimize inventory levels, and plan production schedules. However, there is a widespread misconception that forecasts must be 100% accurate to be useful. This article features insights from an interview with Chakri Gottemukkala, Co-Founder and CEO of o9 Solutions, Inc., on why forecasting is still important, even if it’s not 100% accurate.
The Importance of the Forecasting Process
"To your point of whether forecasting is required, it absolutely is required because of supply chain lead times. The process matters because it creates the mechanism through which the entire organization is acting as one around the forecast and the demand."
According to Chakri, demand forecasts are necessary because of the lead times in the supply chain. If you could magically make all of your lead times zero, then you wouldn't need forecasting. But that's not reality. You have to make decisions based on an estimate of demand, which is a forecast. Said Chakri in an interview with Simon Joiner, Director of Product Management at o9 Solutions, Inc.:
However, Chakri also emphasizes that the process of forecasting is more important than the accuracy of the forecast itself. Instead, it's about understanding the underlying drivers of demand, as well as the risks and assumptions, behind the forecast. He encourages supply chain organizations to monitor the risks, opportunities, and demand drivers of the forecast continuously. By doing so, they can update the forecast and capture opportunities and mitigate risks more quickly.
"It's really the process, the focus on the leading indicators, the assumptions, and measuring and monitoring them that is a critical aspect of the forecasting process, rather than just getting the [accuracy] right."
Understanding the Range of the Forecast
Chakri acknowledges that forecasts are rarely 100% accurate. However, he argues that it's essential to understand the range of the forecast. The planning process should take into account the risks of the supply chain and the ability of the supply chain to support that forecast.
For example, risks are higher if you're trying to fly the supply chain at 120 or even higher if it's at 140 versus a demand of 100. Understanding the range of the forecast and creating a planning process that's able to estimate the risk in that forecast is a very important part of the demand planning process. According to Chakri, this creates a dialogue about how much risk you're willing to accept around the forecast versus trying to get the number to be perfectly accurate.
"The planning process is about understanding the risks of the supply chain and the ability of the supply chain to support that forecast."
Combining Algorithmic and Human Intelligence
While forecasting accuracy is of relative importance, there is still much room for improvement. Chakri believes that it's possible to improve forecast accuracy using algorithms, especially for complex product portfolios and multiple locations. He recommends segmenting product portfolios and markets and locations properly, so that forecasts can be improved using algorithms. However, there are certain portions of the product portfolio that are better forecasted using a combination of algorithmic intelligence and applying human intelligence on top.
"We have to combine these [AI and human] aspects in creating the demand plan."
The process of forecasting is more important than the accuracy of the forecast itself. By embracing the process of forecasting, rather than fixating on accuracy, supply chain organizations are better positioned to update the forecast quickly and capture opportunities and mitigate risks as events unfold and new information becomes available.
1.The process of forecasting is more important than the accuracy of the forecast itself.
2.By understanding the underlying drivers of demand and monitoring risks, organizations can update their forecasts and capture opportunities and mitigate risks more quickly.
3.It's essential to understand the range of the forecast—from the more pessimistic potential outcomes to the more optimistic—and estimate the risks associated with supporting those scenarios.
4.While forecast accuracy isn’t everything, it can and should be improved. Combining algorithmic and human intelligence is key to improving forecast accuracy, especially for complex product portfolios and multiple locations.
So can I add an element of skepticism into our conversation? What is the point of forecasting? Because it seems a question that's asked many times by people is the forecast is wrong? What's the point?
When I hear that, I'm reminded of a couple of quotes. One is Mike Tyson, who said everyone has got a plan until they get punched in the mouth. That's pretty funny. Building on that, actually, Dwight Eisenhower, I think it was, who said, you know, plans are useless in battle, but the process of planning is extremely important.
If you peel that, of course, forecasts are required. You have lead times in the supply chain. If you magically can shrink the lead times in the supply chain down to zero Then of course, you don't need forecasting. Yeah, that's not reality.
You have lead times in the supply chain and you have to drive decisions based on an estimate of demand. That's a forecast. So the question really is what is the approach to forecasting? A lot of people look at forecasting as the magical button to increase the accuracy of the forecast to three decimal points or get the number Precise, exaclty.
Yes, there's an element of how to improve the accuracy of the forecast, but really the process of forecasting and the process of planning is about the most important aspect of that. Borrowing from Dwight Eisenhower statement is really understanding the underlying drivers and the risks and assumptions behind the forecast. What are the risks and opportunities and the assumptions behind the forecast? Because once you put that into process where you're monitoring the risks, what you're monitoring the opportunities, you're monitoring leading indicators of the forecast, then the process will start capturing potential changes to the forecast upside or downside earlier, that gives you a better way to manage more lead time to respond, but also a better learning capability of what is going wrong and what is going right.
So it is really the process, the focus on the leading indicators, focus on the assumptions, focus on measuring and monitoring them is a critical aspect of the forecasting process, then just getting the precision right. The second aspect of improving the forecast accuracy, it's not about the mindset of having a magic button to automatically generate a forecast that is extremely accurate. We have to recognize that the forecasts are going to be inaccurate. The question is what is the range of that forecast?
That's possible. The forecast could be 100, it could be 120, it could be 140. The planning process is about really understanding the risks of the supply chain, the ability of the supply chain to support that forecast, right? So obviously my risks are higher if I am trying to plan the supply chain at 120, even higher if it's at 140 versus a demand of 100.
So understanding the range of the forecast and understanding and creating a planning process that is able to estimate the risk in that forecast is actually a very, very important part of the demand planning process, because then you're actually having a dialogue about how much risk you are willing to take around the forecast versus trying to get the number to be exactly accurate. That said, there is a lot of scope for improvement in using algorithms to improve the forecast accuracy, the complexity of the product portfolio, the complexity of the number of locations at which you are planning.
All of this dictate that if you segment your product portfolio and your markets and locations properly, there are portions of that portfolio of products that can be forecasted better using algorithms. And there are certain portions of the product portfolio that are better forecasted using a combination of algorithmic intelligence and applying human intelligence on top. There is no doubt about that, right? So we have to combine these aspects in creating the demand plan.
But to your point of is forecasting required? It absolutely is required because you have supply chain lead times is the process that matters. It's a process that creates the mechanism through which the entire organization is acting as one around the forecast and the Demand Planning.
The Forrester Wave™ for Collaborative Supply Networks 2022
o9 received the highest possible scores in ten criteria across the Strategy and Current Offering categories.