Traditionally, when it came to supply chain planning, demand and supply streams have been projected as a specific deterministic value. For instance, 22 pallets of a specific item will flow through a given location on Monday.

But that single number is just a guess. It could be more or less, depending on how a host of assumptions and projections actually play out.

Yet despite the fact that this number is so uncertain and it’s not real – it’s only a guess –the whole supply chain is driven off this one number. And if it’s wrong, or should we say when it’s wrong, we’ve got all kind of problems. Review processes are set in motion. Costly expediting and reaction (or worse, overreaction) may begin. Up and down the supply chain. All because the number wasn’t the exactly the one we planned for – which was really just a guess to begin with.

A probabilistic supply chain model, like we use at ToolsGroup, uses a different approach. It assumes that this value is not a single number, but a probability function. So instead of saying there will be 22 pallets, the system says that there is a 50% chance that there will be 22 pallets, a 25% chance that there will be 21 pallets, a 15% chance that there will be 20 pallets, and so on. Rather than a single number, you get a range of values with different probabilities, also known as a probability distribution.

With this approach, you model the entire supply chain through this probabilistic distribution. And because it so much more mirrors the way the world really works, it has an inherently better chance of correctly describing the expected outcome and getting the planning process correct.

A simple example would be filling a truckload. If you have fixed numbers, you don’t know how best to fill the truck. When you have probabilistic requirements, you can use the natural uncertainty of your supply chain and the flows of materials to load the truck, and you will fill it in a way that optimizes your margin (or whatever goal you have set for yourself). And because you understand which outcomes are likely and which are not and can cause a problem, you can focus your review and expediting on those situations that truly deserve a reaction. Because you weren’t expecting an exact number, but a range of numbers with probabilities.

Labels: probabilistic modeling, probability distribution, supply chain planning

Since we at ToolsGroup began making a big deal about lumpy demand, a question we hear from time to time is “Just what is lumpy demand?”

Lumpy demand is demand that is intermittent and hard to predict. A commonly accepted threshold for intermittent demand is the point where there is at least a 50% probability of having a time bucket with zero demand. If more than half the time you have no demand for an individual SKU-Location, you have intermittent or lumpy demand.

Some products, like slow moving spare parts, have naturally infrequent and lumpy demand. But many other products have demand that becomes lumpy due to changes in the business and supply chain environment, such as when the demand stream is split into smaller and smaller time buckets. As the demand stream gets split more ways, it usually changes from a relatively smooth flow to variable and erratic. This happens more often than most people think, because of three business trends that are driving the growth of lumpy demand.

The first is product proliferation. Customers have far more choices than ever before. This divides demand into smaller buckets.

The second issue for most companies is that they are replenishing more frequently. Shorter time buckets mean more demand variability. The same SKU may look like a “fast mover” (relatively stable demand) if observed in monthly buckets, but looks like a “slow mover” if observed in weekly buckets, and appears intermittent or lumpy at the daily level.

Third, many manufacturers who used to focus on stocking big regional distribution warehouses are now minimizing out-of-stocks further downstream, often at the end node of demand. As the replenishment planning focus shifts from the primary distribution centers to secondary distribution centers and retail shelves, demand is increasingly disaggregated into smaller demand streams.

Lumpy demand is much more challenging than high volume mainstream business. Demand variability is high and typically skewed. Supply chain noise increases. Demand signals are harder to read. As you would expect, forecasting and inventory management in this environment is more challenging. Supply chain planning gets tougher, and you’ve got a lumpy demand problem on your hands.

Labels: demand stream, intermittent demand, lumpy demand, product proliferation