How to Optimize and Error-Proof your Lean Process Design

Let’s get started by going over the goals of this session, and what you can expect to learn. First, you need to be convinced, if you’re not already, that your Lean process improvement toolkit needs to include simulation modeling. It’s simply not an option any more. This is especially true if you’re involved, as we are, with designing and managing Mixed Model processes, which inherently add a lot of variability challenges.

Second, you need to understand what data will be required. There’s data you need for a static line design, and the good news is that this same data transitions nicely to a simulation model. Third, you need to have a software application. This is not a software webinar, you have a lot of choices in the simulation software marketplace. But the issue of software has been, in my experience, a major hurdle, and I’ll spend a little time on this. Finally we’ll build a very simple model, so that I can demonstrate what can happen if you don’t add this modeling step to your line design process, and on the other side of the coin, the benefits of doing so. My mission will be accomplished if you leave the webinar thinking, “I really need to learn more about this… this is important.”

Let me tell you a story about my first real Line Design, my first project as a Lean consultant. In retrospect it could not have been simpler. This was a faucet manufacturer that you’ve probably heard of, and the job was to design a whole bunch of assembly cells, organized by product family. Showers, Kitchen, Bath, etc. It’s not that the products were so different that they all needed separate cells or lines, but the volume was so high that we needed to get the Takt Time up to something that a human being could deal with. We ended up with a Takt Time of about 20 seconds per station, which means that each cell would be assembling and boxing three faucet sets a minute. Which is pretty fast, too fast really, looking back on it. Our current rule of thumb is to keep Takt above one minute. Hard to do when the total work content is less than two minutes.

So let’s do some math in our heads. The shift was about 400 minutes, we are designing this line to produce 3 units a minute, so the production for a shift should be 400 x 3 or 1,200 units, right? Only four workstations per cell, so balancing shouldn’t be too hard. So here comes line live on the first day, and remember, we tried to do everything right. Ergonomics, material presentation, work instructions, training, yada yada. And on Day 1 we were averaging about 850 units. Chalk it up to “learning curve”, we said to ourselves. Day 2, 875 units. See, we’re improving! Day 3: 900, yes, more improvement. Day 4: 900. Day 5: 900. At this point management is looking very concerned. We want and need 1,200. What’s going on?

I will have to admit that we were somewhat stumped. Observing the line didn’t really tell us that much, and operators seemed to be focused on the job at hand. But here’s what was happening: with such a short Takt Time: it was very easy to introduce small delays in the flow. Dropping a part, getting distracted, going to the rest room, etc. Many of these delays were not obvious, but because there were so many opportunities for delay (1,200 times 4 operators), at the end of the day we were getting 900, not 1,200 faucets. What was the fix? We added more buffers, what we call In Process Kanbans, between the stations, and throughput jumped up to around 1,100 per shift. Could we get to 1,200? Never!

Knowing what I know now, what could have been done? Years later I created a very simple simulation model of this cell design and was easily able to confirm what we actually experienced. As soon as any variability was added to the model, the 1,200 units per shift became impossible, no matter how many buffers were added. If I had known this, we would have designed for a higher number, say 1,500 per shift.

The lesson here is that even if your design seems pretty straightforward and simple, it is essentially impossible to understand the impact of variability on the system performance in your head. Sure with a little more savvy we could have anticipated this effect and taken evasive action, but that’s not being very scientific about it.

Let’s spend a few minutes and go over the static Line Design process (i.e. without simulation), and give you a feel for the steps involved, and the data required. As I mentioned earlier, the good news is that the static process is very compatible with the software simulation requirements, although you will need to add some additional pieces of information later on. Let’s take it step by step:

1. LINE. The goal here is to design a Value Stream, or a manufacturing production line. Not only that, but we’d like to make this line as flexible as possible, and design it to build a family of products. Products that are similar, but not all the same. An example of this might be a Lawn Tractor that comes in different engine sizes, deck width, and options, but all of them are built in the same line. We don’t know yet the limits of how much variation we can handle without making the line very inefficient, but we’ll get there.

2. PRODUCTS. So we start with a list of products that we believe will be good candidates to mix on the same line, based on our product knowledge and existing data. We can refine this list as we go forward, but in general we’re looking for three characteristics:

a. The products have process commonality. They go through the same processes.

b. The products have similar work content times. They don’t’ have to be identical, but they should fall within a range of times.

c. The products have similar material content. If every product has completely different material, this is not a total show-stopper, but it will make things more challenging.

3. PROCESSES. We’re doing work to manufacture these products, and we organize this work into Processes, or work of the same type. Every product will need to be documented using a tool called a Process Flow Diagram or PFD. The PFD shows not only the processes required, but their relationship in time, in a flow-chart format. Think “routing” to understand what a PFD is.

4. PROCESS TIMES. Every Product/Process relationship has a work content time. The work in a process is make up of smaller work elements, but at this level we’re interested in the total time in a process for each product.

5. VOLUMES. Every product needs to have a Volume or Demand value. This Demand is not necessarily current demand… we’re designing a line that will meet our production requirements for a period of time, so you need to be looking into the future in order to set this volume for each product.

6. RESOURCES. There are a number of calculations you can now do, with the goal of calculating the number of machines and people you will need to be able to produce the Volume numbers you have set as a goal. We’re talking here about concepts like Takt Time, Effective Work Minutes, and then a Resource Calculation by Process.

7. LINE BALANCE or WORKSTATION DEFINITION. If you’re doing work sequentially, like on an assembly line, then some processes will need to be divided into individual work stations. This is a not a minor step in the process, and one which can contribute to a lot of process variability. Remember too that each workstation will have its own workstation time.

8. SIGNALING. In a Lean production line we will also consider the use of what are called In Process Kanbans, or buffers. Although they can’t always be used, they can be very helpful in overcoming imbalances and improving the process flow.

9. CONCEPTUAL LAYOUT. You don’t need a CAD-level drawing yet, but you do need to have a conceptual or block layout of your proposed design. This shows sequential versus parallel workstations, the use of buffers or IPKs, and how the various processes are connected.

That’s a very quick review of the minimum data needed for a static, i.e. non-simulation, line design. This is a part of what we cover in our Mixed Model Line Design workshop, but in a lot more detail. The bottom line is that if you don’t have this data, then you can’t move on to the simulation model step.

Let’s say you’ve taken it this stage of maturity (which can take months of work, by the way). In a purely static design the next step would be to audit your conceptual layout, and then create a detailed CAD layout that you would physically create. The line would then be brought live, and any performance issues that you had not been able to anticipate will need two B a dressed in real time and fixed through a series of kaizen activities. This is the most expensive way to debug your production line, and the most time consuming. If you can avoid it you don’t really want to do this.

So let’s talk for a moment of the software. If you want to build a simulation model, you will need to use a software application. There are literally dozens of systems available, but I have found that software related issues our one of the biggest hurdles. The problem is that there is a learning curve in using any of these systems, and in fact to get them to do exactly what you want often involves something called coding. Every software vendor would like to convince you about the Kings of use of their system, but in reality as soon as you get beyond a very basic requirement it starts getting complicated. It’s a little bit like computer programming in general. It’s hard to get really good at programming if you don’t do it all the time in fact we could say definitively but you’ll never checked good at computer programming if you don’t do it on a regular basis. The same thing applies to simulation software.

ne solution to this dilemma use to build software simulation expertise within your company, as a separate group or department. These people would focus on simulation modeling build their expertise, and provide that expertise to you as a user when the need arises. This can create something of a bottleneck within an organization as the demand for simulation services grows. You will be relying on a relatively small group of people to help you, and it may not always be available.

Another option, which I will show you in this webinar, is to use a simple simulation tool, and yes they do exist, and build the model yourself. Of course because of the complex cities that can be involved there are limits to this approach, but for sequential and a lean year types of models, this can be a good approach.

A third option, which is probably the best, is a hybrid approach where you as the designer would do most of the simulation modeling work, but where you can also rely on the IE expert within the organization to oversee your efforts. This approach can result in high quality results and also reduce the total time required to build a model.

We’re going to be looking at a simulation modeling tool, but we don’t sell software or even recommend one Software System over another. The good news however, is that they are all fundamentally very similar in the way that they work, and our discussions on model building and line design optimization should apply to just about any software package. We will be using and in the house excel based tool, but this is a tool that we use for internal use and training online. So at least for the moment don’t ask me about being able to get it. As I said there are a lot of other options.

So what are the Lessons Learned?

A static Mixed Model line design does not test for changes in mix and volume, or for process variability. You probably have both of these things, and both will degrade process performance.
A line performance is too complex to understand in your head. Sure, you can add IPKs or buffers and overdesign without simulation, but that’s pretty imprecise and could be expensive. Or you could fix it later, which is really expensive.
Simulation modeling tools help to test and improve the static line design. Fortunately we have some technology at our disposal, so we need to learn to use it.
Consider simulation modeling as a requirement! If you don’t have it already, build the expertise in yourself and your organization. Don’t wait.
Foundation: the Lean Process Design methodology. If you don’t have this, then you won’t have a good baseline design, before simulation. You need to learn this as well.
Here are some opportunities to move these requirements forward.

1. We are planning a 3-day workshop on the subject of Simulation Model, in partnership with some of the best simulation experts in the world. Our vision is to share with you the software that you just saw, do a line design following the Lean Design Process, building a fairly complex model, testing line performance, and making improvements. This will be mostly a hands-on experience, with some explanation required along the way. We’re targeting this Fall for this workshop, and we’ll let you know as we get things finalized.

2. We teach a Mixed Model Line Design workshop, 3-days, in detail on the Line Design Process. The next one is hosted at Toyota Material Handling on June 16-18, and includes plant tours, interaction with Toyota experts, and a deep dive into the line design methodology.

3. Back at Toyota in September we’ll be covering the topic of Lean Material Management, also 3 –days.

You can find out more about these at, under the Workshops menu option.

Leave a Reply