On Leveling and Fluctuations—Introduction

Leveling (or Japanese heijunka) is a common topic in lean manufacturing. Leveling combats one of the three evils in manufacturing: unevenness or fluctuations (in Japanese mura). While reducing fluctuations is an excellent goal, it is unfortunately not free, and there is significant effort not only for implementing and maintaining the leveling system, but also on the way the fluctuations are reduced. While everybody talks a lot about the benefits of leveling, very few talk about the cost. This may be one reason why so many leveling systems fail. Let me show you the hidden not-so-nice side of leveling.

Introduction to Fluctuations

Every production system, or any organized system for that matter, has to combat three evils to be effective. These are waste (muda), unevenness (mura), and overburden (muri). While the West usually starts with waste (since it is the easiest one), Toyota considers overburden to be the worst one since it causes a lot of unevenness and waste. In between are fluctuations, which subsequently also cause a lot of waste. (For more, see my post Muda, Mura, Muri: The Three Evils of Manufacturing.)

In general, fluctuations are changes of any kind, as for example employee turnover, varying product quality, supplier timeliness, machine breakdowns, and so on. There are three fundamental ways to combat fluctuations: inventory, capacity, and time.

Inventory

You can handle fluctuations by having additional inventory. Any differences between demand and supply will increase or decrease your inventory. This is especially suitable for short-term fluctuations and works until you run out of space or out of inventory. On the downside, there is the risk of having large and costly inventories, especially for longer-term fluctuations or if a trend is believed to be merely a fluctuation. As you don’t know the future, you can never be sure if an uptick or drop in demand is a long-term trend or a short-term fluctuation until afterwards.

The advantage of decoupling with inventory is that the fluctuation is not passed on upstream or downstream. If your customer orders erratically but your outbound inventory buffer holds, you can produce (in theory) quite steadily. If your supplier delivers erratically but your inbound inventory buffer holds, you can also produce (in theory) quite steadily. Hence, inventory has the power to actually eliminate fluctuations, instead of simply passing them on upstream (capacity) or downstream (time). The downside is the cost of inventory.

Capacity

Especially longer-term fluctuations can be handled by a change in capacity. If the demand is high during the peak season, then you can plan additional capacity, limit vacations, etc. If the demand drops later during the off-season, you reduce working hours, let your people go on vacation, and generally slow down your production. As this requires planning, it is difficult to do on short notice.

One major drawback of using capacity to level fluctuation is that this will pass fluctuations on downstream. If you produce as fluctuatingly as your customer orders, your supplier will have to deliver equally fluctuating. Using capacity to handle fluctuations will pass fluctuations through to upstream. If your order from your suppliers with fluctuations, your fluctuations will now also be the problem of your supplier.

Time

The automatic fallback option that always works is time. This is the easiest to do. If your demand goes up, you let your customers wait. If your demand goes down, you let your operators and machines wait. Unfortunately, it is in many cases not the best approach. If you make your automotive OEM customers wait for parts and turn off their assembly line, they will charge you about USD 1 million per day for doing so. Similarly, having your workers sitting around twiddling their thumbs is also expensive. Some make-to-order manufacturers can let their customers wait until they have capacity (e.g., aircraft manufacturing), but this does not work for everybody.

As capacity adjustments pass fluctuations upstream, time delays pass fluctuations downstream. If your supplier simply lets you wait for his products if there are fluctuations on his side, then his fluctuations are now also your problem.

Introduction to Leveling

Leveling is an approach for decoupling fluctuations in demand that come from the customer (or generally downstream). It does this by reducing or eliminating fluctuations at one point in the value stream, so that upstream processes do not have to deal with these fluctuations. While inventory, capacity, and time can be used to decouple fluctuations, leveling aims to keep the production constant. Normally if there is a temporary increase in demand, you could produce more. If there is a temporary decrease in demand, you produce less. If you need more of product A and less of product B today, you adjust production. If tomorrow you need more of product B and less of A, you adjust again. Leveling does NOT do that, but tries to distribute production evenly across the day, week, or month.

There are three related but slightly different ways to do leveling. The first is to try to produce the same overall quantity every day even if the quantity for one part type may differ. Within a day you can also try to mix your daily quantities in a regular pattern with the smallest possible lot size. The images below show you an example of a production sequence for one day, where in the first row the parts are batched, but in the second row the parts are mixed in a pattern trying to give equal time between one part and the next part of the same type.

Finally, there is the option to make not only the same overall quantity every day across all parts, but to make the same quantity every day for every part. The images below show you three production plans.  The first one is not leveled, but batched. The second one is leveled, with the weekly demand being spread evenly across the days in a weekly pattern. All high runners have the same quantity produced every day. The third pattern combines this same quantity every day with a daily mix to get even more leveling out of your effort (read in rows from left to right and top to bottom).

If it would be that easy…

I regularly come across plants that see diagrams like the one above, like it, and want to implement it. However, it is not that easy. In fact, especially a longer-term pattern is quite demanding and difficult. Toyota can pull this off, but I have seen no company outside of Toyota (yet) where I was convinced, despite many companies in industry trying. There are two main caveats with leveling that are rarely talked about: other fluctuations, and the cost of leveling. In my next post I will look in more detail at how other fluctuations and the cost of leveling can cause problems. Now, go out, understand how fluctuations affect your production system, and organize your industry!