On Leveling and Fluctuations—The Downsides of Leveling

In my last post I gave you a brief introduction to fluctuations and leveling. And, I already hinted that there are two downsides to leveling: properly set up it handles only fluctuations coming from upstream, and—what is rarely talked about—there is an increase in inventory and/or waiting time to decouple the fluctuations in leveling. The second one especially is widely ignored or even vehemently denied in industry, but it is definitely there. Let me show you!

Fluctuations Other Than Downstream

Leveling is an approach for decoupling fluctuations in demand that come from the customer (or generally downstream). This is important. Leveling only decouples fluctuations that come from downstream, so these fluctuations are no longer “fluctuating”—or at least fluctuating much less downstream of the point of leveling. Ideally, this would look like the image below. At one point (preferably toward the downstream end of the value stream) someone should level the fluctuating customer signal. From there on, your plant, your supplier, your supplier’s supplier, and so on all the way down to the mining should have a leveled signal, since you decoupled the value stream from the customer fluctuation. Everybody can just continue to use the same leveling pattern as at the point of leveling for the materials needed to make the products.

I can already hear you snickering. Yup. It usually is NOT like the image above. For the sake of the argument, let’s make it easy and assume that the leveling actually works, and it is not just an ill-designed system that pretends to level something (and yes, I can hear you snickering again…). Leveling can decouple the fluctuations coming from upstream. But customer fluctuations are by far not the only fluctuations you could have. Other fluctuations can also come from upstream. Your supplier could mess up and deliver late, the wrong parts, defective parts, or not at all. You probably have an inventory buffer to decouple fluctuations from downstream, but this can also sometimes fail. And, let’s not forget your own plant. You have breakdowns, absent operators, more urgent parts for another customer, a boss yelling for other parts, and so on. In reality, even assuming the leveling works, the picture may look more like the image below. And, you are probably snickering again that the fluctuations are not big enough.

However, if the value stream cannot provide a stable signal beyond the point of leveling, then what is the point of leveling in the first place? This still assumes that the leveling is well designed and can create a stable signal, which is a big “if.”

It can even be worse. You are probably not the only plant in your value stream that attempts leveling. There are probably multiple suppliers of yours (and possibly also some intermediate customers) that attempt leveling, without having a fundamentally stable system. Then the situation would look something like this.

For leveling to have any beneficial effects, you need a stable system! And by stable, I don’t mean that you usually manage to get the parts to the customer that they want, but that you can actually follow a production plan without changing it. If you have a two-week leveling pattern, then you actually need to manage to produce this pattern for two weeks, no ifs, ands, or buts. If you make a leveled production plan, you need to actually follow it. I have seen so many plants that can’t produce tomorrow what they plan today, yet they attempt a two-week pattern. That is bollocks! These plants had machine breakdowns, absent operators, lack of raw materials, and many other problems. The leveling also frequently broke down because they ran out of products of a certain type and had to rush production of this product to restock, neglecting other products that should have been made in a leveled signal. Overall, many different fluctuations of all types forced them to change the leveling pattern to avoid a stock-out. And since they often tried to stick to the pattern until the last minute, the rush and the urgency became even worse and the shop floor was repeatedly thrown into chaos.

Leveling brings the most benefit if the stable signal is not only for the plant doing the leveling, but also for its suppliers. Hence, your purchasing department would also have to order according to the leveled pattern (yes, I hear you snickering again). If you make X products of one type each day, then you should order the corresponding material for this quantity of X products each day.

Instead, your purchasing department orders their parts once per month based on the forecast rather than actual production. And if they come across the ocean from China, you kind of have to. The lead time is just too long for a good leveling pattern. If you adjust your pattern every two weeks but need three months to ship from overseas, then it is just not feasible to continue the leveling pattern with your orders. By the way, avoiding this bullwhip effect is one benefit of pull production.

The Cost of Leveling

When people think about the cost of leveling, they think about the effort of setting it up and the time needed to run and manage the leveling system. While this is also part of the cost, here I want to talk about the impact on your inventory.

Leveling is an approach to decouple upstream fluctuations. As mentioned before, the goal of leveling is to keep production as constant as possible. Hence, you can not use capacity to decouple the fluctuations in leveling! Yet, you still need to decouple the fluctuations. And this is rarely talked about if there is a discussion on leveling.

As mentioned before, there are three basic ways to decouple fluctuations: inventory, capacity, and time. Capacity is not an option for leveling, as leveling explicitly tries to eliminate capacity fluctuations that otherwise could have counteracted the incoming fluctuations in demand. Hence, you need to decouple through inventory or time. And decoupling through time and letting your customers wait is not an option for many companies. Hence, often the brunt of the decoupling falls on inventory. And this means more inventory!

Yes, the dirty secret of leveling is that you need more inventory!

If your leveling manages to stabilize the rest of the supply chain, this may still be worth it. If your leveling falls apart again on your own shop floor, it would be a waste.

Let’s make a numerical example. Let’s assume your average demand per day is 100 pieces of one product for the next 20 work days (approximately one month), or a total of 2000 pieces. However, this 100 pieces per day can fluctuate. It could be zero, or it could be 200 per day.

Let’s assume a worst-case scenario: The customer orders nothing for 10 days, and then 200 pieces per day for the next 10 days. The average is still 100 pieces per day. If your only way to decouple is inventory, you would build up an inventory of 1000 pieces in the first ten days, and then bring it down to zero again in the second ten days (ignoring any safety buffer). The diagram below shows you how the inventory will go up and down again, while the production is perfectly leveled.

Let’s take the same worst-case scenario, but now allow also decoupling by capacity. For simplicity’s sake assume you can change capacity between 50 and 150 parts per day. In this case, you would produce 50 per day for the first ten days, ramping up to a total inventory of 500 pieces. Then you would produce 150 pieces per day, bringing inventory down again to zero (also ignoring safety stock). As shown in the diagram below, the maximum inventory is now only half of the previous example. Of course, this is at the cost of not having a leveled production.

Admittedly, these examples are quite abstract. However, you DO have fluctuations from your customer, even though they may look different. Your fluctuations may be smaller, or larger than in this example, but definitely more random, and most likely unknown until they happen.

I sometimes hear the counterargument that “on average it will work,” but the whole bloody point of bloody leveling is to deal with the not-average fluctuations. The longer your leveling pattern, the more inventory you need additionally to decouple the fluctuations! Or alternatively, if you decide to decouple your fluctuations by time, the longer your leveling pattern, the longer your customers will have to wait! If your inventory is too small, then the fluctuations will break through the leveling and destroy your pattern (or you run out of products). If your own fluctuations are too large, it will also destroy your pattern again quickly. Overall, if your system is not stable enough or if it has not enough inventory to decouple the fluctuations, then the whole concept of longer pattern leveling is a total waste. Don’t do it!

Hopefully this post helps you to decide whether longer pattern leveling is suitable for you or not. In case of doubt, it is not. In my next post I will talk a bit on how Toyota does leveling. Now, go out, see what leveling makes sense for your plant, and organize your industry!