How Do I Reduce Buffers To Zero?
Dear Gemba Coach
I run an assembly line for service station equipment. Our product is customized for each customer so we have a huge variety of options. My line is organized in segments (hydraulic assembly, wiring, etc.) with a buffer between each section of the line. I’ve long tried to reduce these buffers but haven’t found any good way to do it. Our lean consultant wants me to reduce the buffers to zero so that we have true one piece flow – should I do it?
Is he crazy? Or is it a case of lean misunderstanding? Is the consultant arguing for zero buffer tomorrow or is he - I’m assuming the consultant is a he - saying you should have a plan to progressively reduce the buffers? These are two very different statements, and as most conflict grows from misunderstanding each other’s intentions, it’s important to check. To be blunt, if the consultant tells you to take out the buffer right away, don’t even think about it. If, on the other hand, he wants you to have a plan to reduce the buffers say, by half, then by half again according to a monthly (or quarterly) schedule, then he’s got the right of it and he just wants to keep you thinking more deeply about the underlying causes of the buffers.
There is no debate that “zero buffer” is the lean ideal. We constantly seek true one-piece flow because this is the least wasteful way to produce and seeking to do so will reveal all the mura (un-levelness) and muri (overburden) generating muda (waste). On the other hand a buffer-less line tends to stop and start constantly because there is nothing in the system to absorb variation of any sort. I remember visiting the Smart car assembly plant many years ago – it was beautiful and kind of crazy. It was like looking at Ford’s Rouge dream come alive. The facility had been designed as a huge X with suppliers integrated on the production line into one huge moving conveyor. I don’t know where the plant is at now and whether they’ve changed the model, but at the time, one of the suppliers had two major gripes about the arrangement.
The first complaint was that every time there was a problem somewhere down the line at assembly, the supplier upstream would have to stop since there was no real buffer. mura down the line imposed high waiting waste on the upstream process – which went straight on to the supplier’s books. The supplier was paying an extra cost that led to tough price negotiations with the customer, and no real way to solve the problem at the supplier level.
The second complaint was unexpected but just as sensible. Since all personnel, suppliers and customers, worked and socialized together on the same location, information about wages got around and it turned out, not too surprisingly that the OEM paid better – which created a lot of discontent within the suppliers’ personnel. Now, that one, has to be a new form of waste, right? It’s obvious with hindsight, but who would have thought of it?
The Normal Abnormal
Buffers are the reflection of real variation in the process. If you take out the buffers without taking out the causes of variation, you simply end up with a process that works in fits and starts. If the process works in fits and starts, you can’t work with your people to develop standardized work, and so even kaizen becomes difficult. You lose productivity both from the non-repeatability of the job - operators are productive when they get into their “zone,” the effortless repetition of the right movement – and from the improvement ideas that originate from seeing abnormal situations hindering standardized work. If every day has “normal” abnormal situation, it’s hard to whip up the motivation to tackle them, particularly if it’s all causes by the overall system.
Back at the Gemba, we therefore need to study the causes of variation underlying the buffers. To do so, we don’t need much more than spending lots of time on the shop floor and look out for the 4Ms: Manpower, Materials, Machine, and Method.
In manual assembly lines, operators are likely to create two broad types of variations. Firstly, particularly when they’re new on the job, operators might work slower than the required cycle time – this slows the line and starves the next process. The buffer helps to compensate, because the person in the next process can take WIP from the buffer and not have to wait. The second variation will typically be a mistake or a quality issue, which needs to be corrected. Same thing, if it’s spotted at the next process, the operators can put the bad part aside and help themselves to a new sub-assembly. In lean, the way to work with operators to reduce these variations is andon. The operator calls the team leader to help when she’s late on the cycle or when she has a doubt about the assembly. The team leader intervenes to stay within the allotted time, or if there is a real problem, stops the line to get managerial attention and tackle the deeper issue. In general, the response to the andon call will be training of the operator by the team leader.
In my experience, missing parts tend to be the most frequent reason of putting a sub-assembly aside and pulling another one from the buffer. Along the line you usually find another space – not a buffer – with machines waiting for specific components. Most production lines now have their main components in some sort of kanban system with double bins, and so on. But this doesn’t solve the problem of specific parts needed for customized items. Keeping every kind of component in stock can be very costly as specific parts tend to run in thousands, and as suppliers of these parts are unlikely to be happy to ship them one by one. There are many debates there. The first one is that purchasing must understand that ordering one component might be more expensive both in terms of part price and transport cost, but still cheaper than purchasing a thousand at a lower cost per part. The thousand components will need to be stored somewhere and will drain cash until they’re used, which can be months. The second issue is that not keeping special components in inventory means relying on the MRP driven process to deliver exactly in time. This is okay, but in many cases, the supplier’s lead-time is longer than the customer’s expected lead-time, which leads to all sorts of problems. In general, between high runners in fixed quantity stock replenishment kanban and true exotics which will be ordered, well, on order, it’s useful to create a category of components where a small stock is replenished in variable quantity every time it is used. It might be more costly per part, but the overall cost is much lower.
Look Out for Tricks
Machine mistakes, slowdowns or breakdowns are also a main cause of variation. The andon concept originated from Sakichi Toyoda’s automated looms, and the principle applies to equipment as well as to operators. The key here is to have a local support structure that reacts quickly to every machine abnormality rather than waiting for serious breakdowns. This is by no means easy as most plant managers I know tend to be stingy in technical resources, and as many maintenance specialists I know are only interested in saving the day, but not keeping machines running sweetly. Oddly, although “standardized work” has spread quite widely in the industrial world, the Production Capacity Sheet which is one of the three core documents of standardized work can hardly be seen anywhere. This sheet records the maximum capacity for parts processing, with the time spent on manual work, the machine’s automatic operation time the time spent in changing tools, etc. Each supervisor can use this sheet to make sure that their equipment functions normally every hour of every day.
Finally, we come to Method – probably the trickiest source of variation when you’ve got many options. On a multi-model line we usually find that not all products are assembled according to a standard process – which means that if product one has the assembly sequence ABCD, product two might be assembled according to ACBD – which would mean to move the product ahead on the line to C station and then back to B. You probably deal with this every day, and chances are this leads you to bunch similar sequence products together. But then, you’re batching.
The issue is working with engineering to create a set sequence of assembly. But unfortunately, that’s not enough. When you look into the process further, you will realize that even though you have a set sequence for all products, there is a balancing problem: not all options require the same amount of work at each station. This is hard and can be attacked by three angles: balancing the sequence of products on the line, creating areas to externalize variation from the line when there is an extra load of work, and using team leaders to step in and help with some products. Neither of these activities is easy, and all rely on the supervisor’s superlative understanding of the detail of work on the line, product option by product option.
At the end of the day, you’re facing a difficult teamwork issue. Teamwork between engineering and production in order to design the product so that different options will create the least possible mura. Teamwork between your supervisors and operators so that they work every day at solving all small assembly issues in order to aim for standardized work. Teamwork with technical departments to solve quickly all machine issues.
By all means, have an ambitious target in terms of reducing the buffers, but don’t take them out on a whim. In this case, going to flow first requires building the organization’s capability to do so, which, in turn, rests on daily problem solving to build teamwork between the various stakeholders. In other words, this is an excellent lean project – no qualms about it. But it is also a very challenging one that will surface many deeper technical issues, and it’s worth going there with open eyes, step by step in true kaizen fashion.
How can there be standards -- or kaizen -- in a service job when no two instances are the same?
Dear Gemba Coach,
I’m in a service job and I struggle with the idea of standards. I read that there can be no kaizen without standards but how can you have standards when no two instances are the same?
I don’t get kanban -- I don’t work in production so how would it apply to one-off work?
Dear Gemba Coach,
I feel that I still don’t get kanban. I don’t work in production, and I fail to see how stock replenishment would apply to one-off work.
Can lean be used to turn around a company?
Dear Gemba Coach,
Can lean be used in a turnaround situation? Does a burning platform make it easier or harder?