In my last post, I covered Rule-in-Use 1 as outlined in Stephen Spear’s 1999 Ph.D. dissertation on the Toyota Production System.
Rule 1: guides the design and performance of all individual activities.
Rule 1 states: design and perform every activity so that it is structured and self-diagnostic.
Now let’s move on to his second rule.
Rule 2: guides the design and operation of connections between activities.
Rule 2 states: design and operate the connection between every person who or every machine that supplies a good, service or information and the customer who receives the specific item so that the connection is direct, ‘binary”, and self-diagnostic.
According to Spear:
“For the connection to be direct, the customer — the person that uses a good, service, or information — must send a request for that item to the supplier who will deliver that good, service, or information without the request going through a centralized intermediary. Likewise, the supplier must be able to send a response to the customer without the response going through a centralized intermediary”.
“For the connection to be binary, the request must be sent in a form that can be interpreted as DO the set of activities that will result in the delivery of a particular good, service, or information, in a pre-agreed (‘defect-free’) form, quantity, and response time. Likewise, a response must be interpretable as a signal that the activities leading to the delivery of the good, service, or information in the defect-free form, quantity, and response time have been DONE”.
“For the connection to be self-diagnostic, a binary (yes/no) signal must be generated immediately if a DO-request does not generate a DONE-response (i.e., generates a NOT DONE response), or if a DONE response occurs without a DO-request”. Further, “The signal must be interpretable as ‘a problem has occurred in a specific request-response connection that links a specific customer-supplier pair’. Likewise, delivery in anything but the ‘defect-free’ form, quantity, and response time must be interpretable as ‘NOT DONE’ “.
So, while Rule-1 guides the design and performance of an individual activity, Rule-2 guides the design and operation of the connection between those individual activities.
Remember, the thesis behind this thread series is that TPS, as developed and practiced by Toyota, reduces the complexity and simplifies the operation and management of complex adaptive systems. Does Rule-2 help us accomplish that goal? If so, how?
For an in-depth discussion of “complex adaptive systems”, please go back to my 5 part series on “Simplifying the Complex” (see Table of Contents). But for your easy reference, I will again layout the 4 key attributes and general system dynamics of a complex adaptive system:
Complex systems begin with individual components called autonomous agents, which make decisions and produce results in the system. To be complex, a system:
- Requires diversity in the types of agents. If they are diverse, they will respond differently to various inputs, producing more varied results.
- Requires connectedness. …The agents must have a way to contact one another.
- Requires interdependence, which means that the agents influence one another.
- Requires adaptation. In complex systems, adaptation means more than change; rather it refers specifically to learning.
To understand how a complex system operates, it is necessary to think about the strength of each of these four elements. …At a setting of one, the system is uninteresting. It may have the elements of complexity, but nothing much is going on. Diversity is low, connectedness and interdependence are weak and the result is almost no adaptation or learning taking place. At a setting of ten, the system is chaotic. Agents receive too much information from too many sources and are stymied in their decision making by conflicting and overwhelming signals. Where complexity is most intriguing is…. the “interesting in-between”. This means the dials are set somewhere between three and seven, with each dial different from the others. This allows a good flow of information, interaction and learning among diverse agents, but not so much that the system becomes chaotic. This is the heart of complexity – a system that continuously produces surprising results without breaking down.
Remember, complexity is not necessarily measured by the quantity of each of these attributes that exist in a given system, but by how much total energy must be expended to make the system work (or not work) with the actual attributes contained within the system. And what is this energy I am referring too? That would be the human effort, money and time required to operate and manage that system.
And keep that last section about the “interesting in-between” in the back of your mind as you read through this post. We will come back to it again.
In my previous post, we saw that Rule-1 helped us manage the first attribute of a complex system – diversity of agents (Attribute 1) and their actions/decisions that generate varying results. Rule-2 will help us manage the connections between these diverse agents (Attribute 2) so as to minimize system complexity and simplify the interactions that these agents must contend with.
As I stated in my initial discussion of Spear’s “Rules” here:
“Attribute 2 deals with the connectedness of individuals with each other. Spear’s Rule-2 designs these connections in such a manner that they are always predictable and simple.”
Just as “Standard Work” was the most obvious tool that met the requirements for Rule-1 (but not the only one), “Kanban” (request for material) and “Andon” (request for assistance) are the two most popular tools (but not the only ones) that meet the requirements for Rule-2. And “Kanban” is the one tool most commonly associated with TPS – sometimes with the result that “Kanban” is used as the preferred name for the TPS system. (If you hear someone refer to TPS as the “Kanban System”, you can rest assured that the person understands neither).
Let’s keep our discussion simple and focus on Kanban for now. We get a hint of the simplification resulting from Kanban via these quotes from Taiichi Ohno (The Birth of Lean):
- “A big reason for adopting kanban was our desire to reduce the administrative burden of running a factory.”
- “Kanban automated production control. With kanban, people in the workplace issue production instructions automatically. They don’t need to think about writing up any special directions or about finding ways to convey those directions. With kanban, you don’t need a computer.”
- “Kanban highlighted problems as soon as the problems occurred and promoted kaizen activity.”
OK! Sounds good. But show me the money! How does Kanban do all that!
Let’s take a look at each of the qualifiers for Rule-2; direct, binary and self-diagnostic – one at a time. Maybe that will clarify things.
Note that Spear states that the customer/supplier connection with respect to goods, services or information is direct and does not go through a “centralized intermediary”. And who might that “centralized intermediary” be? Having worked in a few manufacturing plants, I’ll take a wild guess and say, “Production Planning” or “Production Control”. (I think the word “Control” sort of gives it away).
In a typical mass production plant, Production Control typically sends out daily production schedules to each manufacturing center in the production plant. These centers may or may not be connected to each other. And the schedules are based on customer orders (preferred) or forecasts generated by some body or some thing (less preferred). And at the end of the day, these production centers report back to Production Control what they actually produced versus the original plan. And production Control then takes this input and generates a new daily production schedule for each work center for the next day.
Pretty straightforward! Works great! Right?
Not so much.
Spear tells a true story about a non-TPS plant that he worked in that operated in this manner (unfortunately most plants still do). He described the daily routine that the plant went through to cope with these daily schedules. I’ll use his own words:
“…the production control department generated a production schedule each day at 7 AM. At 9 AM, the plant manager, the production manager, and area foremen would meet, and at 10:30 AM they would distribute a ‘hot-list,’ those items for which a failure to ship that day would cause the plant to miss an important delivery deadline. Perhaps at the same time, the shipping clerk, in the midst of trying to assemble one of the days shipments would discover a short-fall. Invariably, he would then try to get someone to make those parts so that the shipment would be complete. Later in the day, after doing an inspection of the existing inventory, an area foreman might give a shop floor operator another set of instructions. It was not uncommon in this plant, and it was representative of what I observed in other, non-TPS plants, for one person to receive multiple conflicting instructions as to what to produce. The inevitable result was confusion, not knowing what particular items to send, to whom, in what quantity, and in what sequence.”
Sound familiar? Of course it does! In fact Spear includes this illustration just below the above description:
An inordinate amount of time is spent trying to decide “what to do” rather than just “doing” and, more importantly, rather than thinking about “how to do it better”.
It is also illustrative of the high energy levels required when the “connectedness” of the production system is not well thought out and information and material flow are not aligned. The various work centers are usually somewhat connected when it comes to the flow of materials. A work center may know (but not always) where they will source their incoming materials and where they will send their outgoing product. But the flow of information is centralized and does not flow across work centers. Joe doesn’t know what Alice and Bob made yesterday and are making today. Alice and Bob are equally uninformed as to what Joe made or is making. They are disconnected with respect to information. And when information doesn’t flow, it can become very difficult to get the material to flow.
And, remember, we need flow to achieve the highest quality, lowest cost and shortest lead time.
But with the Kanban system, material and information flow concurrently along the same aligned path. Production Control only supplies information to a single downstream process center (close to the customer) and information flows upstream from there, pulling material downstream in the quantity needed, when it is needed.
Joe, as a supplier, knows exactly what to produce because his downstream customer, Bob, has told him what he needs. And Joe, as a customer, has told Alice, his supplier, exactly what he needs next based on what he is making for Bob. Bob, Joe and Alice are directly connected in real time. There is no confusion about “what to do”. All they have to think about is the “doing” (or “not doing” if that is what the Kanban system is telling them).
In actuality, no one needs to tell anyone anything if the Kanban system is set up properly. Visual signals, whether they be traveling Kanban cards, empty/full boxes, empty/full shelves, an empty/full space, rolling balls, on/off light, colored lights, etc., etc., are all that are required. And these signals specify the product/service requested (defect-free) as well as the quantity and expected response time.
By connecting activities directly, we have a much more effective, efficient and simplified production system. Energy levels are just right – that “interesting in-between” for complex adaptive systems.
Which brings us to……
Within a properly connected (direct) set of activities, there are only two acceptable binary request/response scenarios: DO/DONE and DON’T DO/NOT DONE. If the customer needs a material or service, they will send a request for that material or service to the designated supplier (DO). The supplier will then send the requested material or service to the designated customer in the specified quantity and time-frame (DONE). If the customer does not need a material or service, no request is sent (DON’T DO) and the supplier does not produce or deliver anything (NOT DONE).
Sounds simple! That’s because it is simple. That’s the whole point.
But simple doesn’t necessarily make it easy.
As Taiichi Ohno states in “Workplace Management”:
“Producing ‘the things we can sell, in the amount we can sell’ is a very simple idea, but there is nothing so difficult as actually doing it.”
But I will get into that basket of worms in the “self-diagnostic” section below.
In my previous post covering Rule-1, I described the differences Spear experienced between a seat installation operation at a Big-3 auto maker and Toyota. But the differences did not stop there. At the Big-3 facility, the individual installing seats has no control over when and if the next auto body is delivered to his work area. Nor does that individual have any control over when and if his current auto body is sent to the next work station. He will get another auto body whether he is ready or not, and his auto body will be sent to the next workstation whether that product is “defect free” or not. The control of connections between activities is minimal at best.
But at Toyota, Rule-2 states that the seat installer has some control over both of those options. The seat installer will not receive another auto body unless he specifically requests it. This DO request or “signal” can take many forms – an empty space, no Andon light, etc. – but the supplier will not send the auto body unless he has been requested to do so by the customer (seat installer). But if he has received a DO request, the supplier will send the next body to the seat installer, fulfilling his DONE response.
And that same seat installer will not send his completed auto body to the next station (he is now the supplier) unless he is requested to do so by a DO request from that customer. Also Rule-2 states that he will not respond to a DO request from the customer unless his completed auto body is “defect free”.
What a difference!
As long as we are getting DO/DONE request/responses between the connected activities, we know everything is going as planned. Material and information are flowing. When we get a DON’T DO/NOT DONE request/response, we know the connection is working properly but we may have a problem if DO/DONE is expected to be the norm. But that scenario is usually indicative of a problem within an activity and not the connection itself.
The connection may be problematic, however, if we get DO/NOT DONE or DON’T DO/DONE request/responses. In the first case the supplier may be having problems with delivery and in the second case we are supplying product we don’t need. Both imply problems with the connection that needs fixing. In the first case, information may be flowing but material is not. In the second case material is flowing but correct information is not.
Note that this information is always visible to the connected parties/agents in real-time. And the first law of the Jidoka Pillar is “Stop and notify for abnormalities”. And according to Spear:
“Rule-2, implies two hypotheses. Within a given period:
a: the customer will need a specific mix and volume of outputs from the supplier;
b: the supplier is capable of providing outputs in that specific mix and volume to the customer.
These hypotheses are tested with each request/response cycle.”
So if these two hypotheses are not met, and we will know that from the request/response diagnostics, the responsible activity must stop and proper notification be given.
But some will say: “But won’t this TPS method cause complete chaos with the flow of product to the customer? They are stopping the line! You can’t do that on the whim of a lowly line operator! You gotta keep that line moving! That’s why we have a rework department, after all! Those guys will take care of it!”.
Yes, some will say that. I sure said that in my earlier days.
But a little later in his thesis Spear gets to the heart of Rule-2:
“Rule-2 guides the design and operation of the connections through which adjacent customers and suppliers transfer material, energy, and information. These trigger the start and stop of supplier activities. Rule-2 plays a key role in creating opportunities for problem-solving and learning. Rule-2 reflects the theme that by pre-specifying expected performance, observations about actual performance can be used for problem-solving and learning.”
So that’s what that “Andon” cord is for! It’s not there to tell the world that somebody did something “bad” or that a “bad” thing just happened. It is there to tell the world that we now have an “opportunity” for some good old-fashioned problem solving and “learning”.
As Taiichi Ohno eloquently stated in his book Toyota Production System:
“A production line that does not stop is either a perfect line or a line with big problems. When many people are assigned to a line and the flow does not stop, it means that problems are not surfacing. This is bad. …There is no reason to fear a line stop.”
So, stopping the line is not “bad” – not stopping the line is “bad”!
By stopping the line we have isolated a point in time where a real problem exists, where the maximum visibility of underlying causes also exists and where effective countermeasures can most easily be discovered and tested. “Learning” is maximized. That’s a “good” thing.
(Incidentally, the Andon is probably the second most famous “tool”, after Kanban, that was originally derived from the use of “Rule-in-Use 2” in the development of TPS. The Andon is a direct, binary signal for the need for assistance, by the line worker, operator, technician, etc., to the Team Leader, Maintenance, etc. Pulling the Andon cord is a DO signal and the immediate arrival of assistance is a DONE response. But remember what Spear stressed, Kanban and Andon are not fundamental to TPS, they are merely off-shoots from the use of Rule-2)
Some will still say: “But rework always fixes the problem. Besides it takes special skill to find problems and fix them. That’s what we pay extra to those rework guys for! Those guys on the line aren’t qualified to do that!”.
Yes, some will say that, too. How much do we pay them?
Kanban is, of course, the most well-known example of the utilization of Rule-2 (but by no means the only example). But, like Standard Work, most discussions around this “tool” miss the underlying principle behind its use. Let’s re-read that last sentence from the previous Spear quote: “Rule-2 reflects the theme that by pre-specifying expected performance, observations about actual performance can be used for problem-solving and learning”.
If you remember the discussion about Rule-1 in my previous post, Spear also made a similar statement about that rule: “Rule-1 plays a critical role in creating opportunities for problem-solving and learning, and it reflects the theme that each use of an activity is a chance to verify the assumptions implicit in the design of the activity”.
While Standard Work provides a platform to evaluate the effectiveness of a given activity, Kanban provides a platform to evaluate the effectiveness of the connections between activities. Rule-1 deals with the flow within an activity, Rule-2 deals with the flow between those activities. And when either flow is interrupted, Jidoka kicks in and Kaizen begins. Problem solving (i.e., learning) takes place when the problem is most visible and the underlying causes are equally most visible.
So, that rework department is not solving problems, it is just hiding those problems from the customer. And in doing so, it is hiding the problems from the system. The problems occurred somewhere upstream, but the cause is now invisible. Rule-1 and Rule-2 insure that the problem is visible in real time at the source and root causes can be seen in real time and resolved in real time. And we are not just talking about “defects”, we are talking about anything that inhibits the flow through the system. Problem solving is simplified!
And since the problem solving has been simplified we do not usually need “specialists” to find and fix the problem. The people working where the problem is occurring are the most qualified to fix that problem. The system structure has also been simplified.
And since we are talking about the system, let’s go back to the beginning of this too long post and take another look at the last two Attributes of a complex adaptive system (e.g., TPS). We covered the first two Attributes with the first two Rule-in-Use.
Attribute 3 “requires interdependence, which means that the agents influence one another”. But we have significantly streamlined the number of agents that are connected to each other. And we have codified and simplified that connectivity. We use much less energy to make the system work than when it was all controlled by a “centralized intermediary” (I.e., Production Control).
And take another look at Attribute 4 which “requires adaptation. In complex systems, adaptation means more than change; rather it refers specifically to learning”. Learning? I think we just talked about that. Not only are we now doing more problem solving and learning than before, we have simplified the learning process. It is now done in real-time at the source. And although we have more brainpower working on the learning, we are using less energy to accomplish the learning since the learning process has been simplified.
So, with the adoption of just two of Spear’s Rules-in-Use, we have simplified all four key attributes of a complex adaptive system. And as we learned from my series on “Simplifying the Complex”, the best way to keep a complex adaptive system out of chaos and eventual collapse is to constantly pursue simplification. That is how TPS, properly carried out, keeps us in that sweet nirvana of:
…. the “interesting in-between”. This means the dials are set somewhere between three and seven, with each dial different from the others. This allows a good flow of information, interaction and learning among diverse agents, but not so much that the system becomes chaotic. This is the heart of complexity – a system that continuously produces surprising results without breaking down.”
I think that will do it for now on Rule-in-Use 2. I covered much more than I originally intended. But if you want more (a lot more) read Steven Spear’s dissertation. I encourage it. And if you want more (but not a lot more) see my previous post where I contrasted MRP vs. Kanban in a complex adaptive system.
Next, I will cover Rule-in-Use 3. This is where we will simplify the System Flow.
Hope to see you there.