Demand forecast is never 100-percent accurate, but it can get better. Brian Dooley, expert and demand-driven thought leader with AIMMS, relates the progress of technology and business-process transformation toward an improved planning environment.

Q: How is the supply-chain planning landscape changing?

Dooley: If we at how our supply chains are changing, that also describes how the planning landscape is changing. Supply chains are becoming ever more complex, lead times are extending, and variability is increasing all of the time, exponentially.

That makes the planning landscape, and our ability to forecast demand accurately, significantly more challenging. We've got to do that over much longer lead times, spread across multiple enterprises. It just much more complicated.

Q: What are the planning horizons that we're looking at these days?

Dooley: Typically, just for the execution piece, it can be 12 weeks-plus. For strategic planning, you're looking at six months, nine months, twelve months and beyond. With those extended lead times, the ability to forecast accurately at a SKU level is challenging to say the least, if not impossible.

Q: When planning gets down to the execution level, it shades into execution, doesn’t it? It becomes more of a reactive than predictive kind of thing. I imagine it's hard to know when that happens.

Dooley: Yes, that's one of the real challenges today — how do you move from the forecast horizon to the execution piece, and how do you decouple the execution part of the planning activity from that forecast? Think about how we've done supply-chain planning for the last 20, 30, 40 years. It's all forecast-driven. We generate a master schedule. We then push that to all our planning engines. Essentially, we are propagating all the way to execution.

The one thing we all know about forecasts is that they're always wrong. We’re not necessarily buying, making, storing and transporting the right items. The challenge is how we decouple our execution piece at the front end from the forecast further up the supply-chain. Which is where demand-driven comes into it.

Q: What are some of the strategies that being deployed in order to deal with this challenge?

Dooley: There's a technique called DDMRP [demand-driven material requirements planning] that's gaining more and more momentum now. It's moved on from being what some people referred to as a science experiment, maybe two or three years ago, to where some big organizations are rolling it out.

Q: Is it a technology or a business-process change?

Dooley: It's a business process, a methodology that allows us to decouple the execution side of the supply chain from the forecast side. Forecasts are still important. They don't disappear, and we use those to size the buffers. But the ultimate replenishment signals — be it purchase orders or shop-floor orders to the factory — are generated based on real demand hitting those buffers and depleting them.

Q: How are we assessing demand? Are we getting accurate point-of-sale information? Are we getting unstructured data? How are we getting our arms around the whole question of what demand actually is?

Dooley: With advances in technology, the closer we get to the point of consumption, to as pure a demand signal as we possibly can, the better. Typically, that could be point-of-sale data. The technology landscape has moved on to allow us to process that volume of data, and use it as our demand signal to hit the buffers and trigger the replenishment requirements.

Q: What's happening to physical inventory during this process?

Dooley: I don't think there's a simple answer for that. It depends upon the particular supply chain. It comes down to the customer proposition, and the degree of visibility we can get of demand. If our customer-promise horizon is much shorter than our actual resupply horizon, then we need to put more decoupling points into the supply chain. We’ll need more strategic inventory positions, which may be at several points within the supply chain. If our customer tolerance time can be extended, then we can remove some of those and push inventory further back up the supply chain. It's a balance between the customer’s tolerance time and where we need to place strategic inventory buffers to cope with that.

Q: How does technology need to change? Which aspects are most important in arriving at this concept of the truly demand-driven supply chain?

Dooley: There are two areas. The first is our historic approach to supply-chain planning. A lot of supply-chain technology that's out there right now still supports those same processes. Software vendors need to start embracing the new approaches.

Another key area concerns simplicity. We hear a lot about digital transformation, all of the buzzwords. But the complexity in the software itself is becoming mind-boggling. It needs super-expert users with months, if not years, of training and expertise to be able to use it.

Quite often, they become a bottleneck in the process, because the one expert is the only person who truly knows how to use the solution. There's a real need to create not necessarily simpler solutions, but a more straightforward user interface, presented more in supply-chain language, so that the entire supply-chain team can utilize the software, and not just a single expert.

Q: How good a job are we doing now in terms of achieving visibility of inventory across the supply chain?

Dooley: The data is all out there. The challenge comes in our ability to actually do something with it. How do we bring it into our tools? How do we run scenarios, "what-if" simulations, in a timely manner?

Q: Can data take the place of physical inventory?

Dooley: I think it can. There will always be a need for physical inventory. There’s natural variability in the supply chain, which we need to buffer for. But the richer the data, the better our ability to understand it. If we can understand the variability, then we can control the exact amount of inventory we need.

Q: How optimistic are you that artificial intelligence can play an important role in analyzing data and making for a better forecast?

Dooley: It's going to play a huge role. Something that particularly excites me is how A.I. and machine learning can do some of the heavy lifting for us. Rather than relying on the end-user to do a lot of the manipulation and analysis, let the A.I. do that. You can almost envision a scenario where A.I. is creating "what-if" scenarios on the fly, and providing us with prescriptive analytics: "Here are the suggestions I'm making, all the points to look at, and the investigations to carry out."