IDC research makes the case for AI-driven supply chain forecasting

Recently-issued research by Framingham, Mass.-based market research and consulting firm IDC highlighted the firm’s top 10 predictions and underlying drivers that the firm expects to have the biggest impact of manufacturers’ IT investments in 2022 and future years to come as well.

A top-level look at the predictions sees that they address remote operations, supply chain management, product and service innovation, security, data, and application sharing, B2B commerce, low code/no code, and sustainability.

Perhaps the most germane prediction, relative to our industry, was prediction number two, which was the following: “By 2023, 50% of All Supply Chain Forecasts Will Be Automated Using Artificial Intelligence, Improving Accuracy by 5 Percentage Points.”

That one really caught my eye, given everything that the supply chain has been through going back to the onset of the pandemic in March 2020. And while things have been uneven, to be fair, the pandemic really highlighted the need for better supply chain forecasting on myriad fronts, for things like supply chain resiliency, demand planning, inventory management, equipment and labor availability, among many others.

IDC’s analysis of the need for improved forecasting pulls no punches, in explaining that manufacturers have been running supply chains for centuries, and, for nearly that entire period, it is something that they have agonized over, with this zinger to complete the working thesis: “The only thing that has been universally true about any forecast is that it will be wrong.”

Maybe that is a little bit harsh (and humorous), but it is not incorrect either.

IDC goes on to observe that supply chain planning falls into three categories: short-term, focused on operational planning; medium-term, focused on tactical planning; and long-term, focused on strategic planning

The underlying theme of each of these categories, it observed, is that they each play a part in determining drivers such as capacity, suppliers, and inventory, with the caveat that as the supply chain gets closer to a particular event, it results in an improvement in precision, but not to the point of 100% accuracy either.

And that leads to this interesting observation made by IDC: “The ability for a supply chain to meet an unexpected demand ends up being about the decisions made earlier in the planning cycles. A supply chain tied too rigidly to a forecast will find itself unable to respond to material variations outside that forecast — because of a lack of either capacity, materials, or broad flexibility. We've talked in the past about supply chains ‘leaning’ themselves to the point of being ‘brittle.’

What’s more, this is helping to move the needle forward for manufacturers to “look outside the box” for what IDC described as innovative, technology-enabled new ways of approaching the challenge of replenishment, like AI, that are not limited to looking for “that next best algorithm.”

That was made clear in an interview with Simon Ellis, Program Vice President, Supply Chain Strategies, IDC.

When asked what needs to happen to get manufacturer buy-in to get supply chain forecasts automated using AI by 2023, as per IDC’s prediction, Ellis noted it is a broad recognition that the current ways of doing forecasting, particularly short-term, aren’t working.  

“For many companies, their forecast accuracy is worse today than it’s been for a while,” he said. “Mostly this is due to massive swings in demand that are hard to predict. So, let’s stop doing it. Not all companies have the point of sale (or comparable data) that they need, but it’s mostly out there, so it requires more extensive data collection efforts. Also there needs to be some recognition from companies that their responsiveness capabilities may be insufficient.”

As a follow-up to that, Ellis said data management is standardized and measured across the organization and decision automation tools are real-time and comprehensively deployed for tactical task replacement and process enhancement.

As for how manufacturers can counter the notion of forecasts becoming “brittle” in the event they are unable to respond to material variations outside that forecast, Ellis was direct, simply saying that if they are unable to respond, then they’re unable to respond.

The reason for that, he said, is that calibrating the supply chain rigidly to a forecast is not a best practice anymore.

“Companies need to balance forecasting with responsiveness (or agility) if they are going to be resilient to demand fluctuations,” he said. “Whether that means agile inventory, or manufacturing capacity flexibility.”

He also offered up a few thoughts related to that below:

• Application of digital procurement tools allows for better balance of cost and capacity across most items, with increased supply flexibility; 
• Broader prequalification of backup supply; 
• Flexible production plans based on forecast variations; 
• Modern postponement capabilities; and 
• Full integration with all enterprise systems, and new and legacy systems/equipment are connected providing a real-time view into supply chain performance and needs

Like a whole host of other things, it is fair to say supply chain forecasting is an inexact science. That said, there are things that can be done to narrow the gap on the road to improved accuracy and forecasting in the future. Will it happen by 2023? It is likely too early to say, but IDC’s research certainly helps to make the case for why it should.