You are on page 1of 11

1

DRUM-BUFFER-ROPE
Drum-Buffer-Rope (DBR) is an operations scheduling and controlling
methodology based on the Theory of Constraints (TOC).
TOC sees any company as a system, that is, a set of elements in an interdependent
relationship. Each element depends on the others in some way, and the global
performance of the system depends on the joint efforts of all the elements of the system.
One of the most fundamental concepts is the recognition of the important role played by
the systems constraint.
The first step is to recognize that every system was built for a purpose, we did not create
our organizations just for the sake of their existence. Thus, every action taken by any
organ-any part of the organization-should be judged by its impact on the overall purpose.
This immediately implies that, before we can deal with the improvement of any section
of a system, we must first define the systems global goal; and the measurements that will
enable us to judge the impact of any subsystem and any local decision, on this global
goal...A systems constraint is nothing more than what we feel to be expressed by these
words: anything that limits a system from achieving higher performance versus its
goal...In our reality any system has very few constraints (this is what is proven in The
Goal, by the boy-scout analogy) and at the same time any system in reality must have at
least one constraint.1 If a company did not have a constraint, its profit would be infinite.
This process is the basis for TOCs methodologies, including DBR. This process has five
steps:
1. Identify the Systems Constraint(s).
2. Decide how to Exploit the Systems Constraint(s).
3. Subordinate everything else to the above decision.
4. Elevate the Systems Constraint(s).
5. If in the Previous Steps a Constraint has been Broken, go back to Step 1. But Do Not
Allow Inertia to Cause a System Constraint. We cannot overemphasize this warning.
What usually happens is that within our organization, we derive from the existence of the
current constraints, many rules. Sometimes formally, many times just intuitively. When a
constraint is broken, it appears that we dont bother to go back and review those rules. As
a result, our systems today are limited mainly by policy constraints."2
One of the main assumptions behind TOC is that every system, like a for-profit company,
has to have at least one constraint. Therefore, if we want to better the systems
performance we need to manage its constraint(s). "There really is no choice in this

GOLDRATT, Eliyahu. What is this thing called the Theory of Constraints, and how should it be
implemented. Croton-on-Hudson: North River Press, 1990. p. 4, my emphasis
2
GOLDRATT, E. What is... p. 6
2003 Prof. Thomas Corbett

www.corbett.pro.br

matter. Either you manage constraints or they manage you. The constraints will
determine the output of the system whether they are acknowledged and managed or not."3
The constraints are not intrinsically good or bad, they simply exist. If you choose to
ignore them they can become bad. If you choose to recognize and manage them they
become a great opportunity, a real leverage for your business.

PRODUCTION SCHEDULING
To have a good schedule we need to do at least two things:
1. Solve the capacity problems and;
2. Protect the schedule from murphy.
We have to make sure that our plan is feasible. In order to have a feasible schedule
we have to make sure that we have enough capacity available to fulfil it and that
we have protected the schedule from the statistical variances in the process
(uncertainties). We can achieve this by following TOCs five step process.
In the first step we have to identify the systems constraint. The definition of a
constraint is anything that blocks the system in reaching a higher performance in
relation to its goal. Therefore, the constraint can be, among other things, the
capacity of a resource as well as a marketing policy.
Whatever the production process, there will always be a resource that limits the
capacity of the process, as in a chain there always is a weakest link. TOC calls this
resource a Capacity Constraint Resource (CCR). In the situation where the market
demand is higher than the capacity of the companys CCR, this resource will be
the systems constraint.
The second step is: Exploit the systems constraint. Supposing that the system
constraint is a resource, the CCR, what we want is to get the most out of it. We
should not waste one minute of its time, as any waste will mean less capacity for
the whole system, and this will lead to fewer sales. If we waste time, we will never
be able to recover the lost capacity (as the demand is higher than the available
capacity). In other words, any minute lost on a constraint is a minute lost on the
whole system, forever.
However, we do not only want the constraint to be busy all the time. We also want
to use its time wisely. We have to make sure that we are producing products for

NOREEN, Eric and SMITH, Debra and MACKEY, James T. The Theory of Constraints and its
Implications for Management Accounting. Great Barrington: North River Press, 1995. p. xix
2003 Prof. Thomas Corbett

www.corbett.pro.br

which we have demand and that are profitable for the company4. To ensure that
this will happen, we have to do the constraints schedule, called Drum.
The Drum is a detailed schedule, with the items to be produced, their quantities,
and start and end times. When building the Drum the starting point is the market
demand. We should first try to deliver what the market wants, in the quantities and
dates desired. When we do not have enough capacity to sell everything the market
wants, the first thing we have to do is determine what we are going to produce and
sell. Afterwards, we have to schedule the delivery of the orders.
To better understand these concepts let us use an example. Let us suppose we have
a customer order for product A, 200 quantities for February 10th. To be able to
deliver this order on the desired date we have to process it on the constraint some
time before the delivery date, so that it has enough time to be processed by all the
resources after the constraint. If this necessary time is 5 days all we need is to
process this order on the constraint 5 days before the scheduled delivery.
This means that product A has to be processed on the constraint on February 5th.
Below we have an example of a Drum:
Drum
Product

Quantity

Start
Date

A
Z
L
...

200
350
120

02/05
02/05
02/06

Hour

End
Date

Hour

5:03
21:43
15:13

02/05
02/06
02/06

21:43
15:13
19:13

When protecting the constraint it is not enough to develop the Drum. When
creating the drum we are dealing with the necessary condition of solving the
capacity problems. That is, we are making sure that the production plan is feasible.
However, we also have to protect the plan from the uncertainties in the process.
An example of an uncertainty is when a resource that feeds the constraint breaks.
If this happens the constraint might stop because of lack of materials to process.
What we have to do here is to create a stock of parts right before the constraint so
that, even if something happens with the resources that feed it, the constraint will
be protected for some time.

TOC does not use product costs to determine product profitability. To understand TOCs performance
measurements read: Throughput Accounting, Corbett, 1998, North River Press.
2003 Prof. Thomas Corbett

www.corbett.pro.br

This protection is created releasing the material to be processed on the constraint


some time before it is needed. This some time before is what TOC calls Buffer.
As an example, let us use the Drum we just created. Product A, 200 units, has to
start being processed on the constraint on 02/05 at 5:03. Let us suppose that before
reaching the constraint this part has to be processed by other 9 resources. If we
want these parts to be available to the constraint at the scheduled time we,
naturally, have to release it before 5:03 on the 02/05. The question is: how much
time before? In TOCs terminology this is phrased as: what should the size of the
buffer be?
The buffer is supposed to be there to protect the constraint from almost all the
statistical fluctuation in the process. We want it to guarantee the supply of
materials to the constraint. We cannot have a 100% protection because we would
need an infinite buffer for that. So we should aim for 99%. The buffer has to be
dimensioned so that there are enough parts in front of the constraint to guarantee
its production 99% of the time. A smaller buffer would put in risk our Drum and
therefore, the companys performance. A bigger buffer would not increase the
protection but would increase the inventory.
So, one of the factors influencing the buffer size is the statistical variance in the
process. The bigger the variance the bigger the buffer has to be.
The size of the buffer is also influenced by the capacity of the other resources in
the process, the non-constraint resources. The bigger the capacity of the nonconstraint resources in relation to the constraints capacity the smaller the buffer
can be.
To better understand this point, let us look at the process below, where we have
two resources, A and B, both with an average capacity of 10 parts/hour.
A
10 parts/hour

B
10 parts/hour

What we want to do is generate a protective buffer in front of resource B.


However in the current situation it is not possible. Resource A does not have
enough capacity to regenerate the buffer. That is, the protective stock in front of
resource B will reach zero constantly.
This is the reason why we should not try to balance the capacity of resources. This
would make it impossible to determine the location of the constraint, as the
statistical variances of the resources would make the constraint change places all
the time.
2003 Prof. Thomas Corbett

www.corbett.pro.br

Still using the above example, for us to be able to protect resource B we have to be
sure that resource A has more capacity than resource B. In other words, all the
non-constraint resources have to have more capacity than the constraint to
guarantee the flow of work to the constraint and from the constraint.
Here is another TOC concept. It classifies the capacity of resources into 3 types:
Productive Capacity - the part of the resources capacity that will be used to
process materials;
Protective Capacity - that part of a resources capacity necessary to regenerate
the buffers and;
Excess Capacity - that part of a resources capacity that can be sold or
eliminated.
When the CCR is the constraint it will only have productive capacity, as we will
want to exploit it, get the maximum out of it. All the other resources, that are nonconstraint, should have productive and protective capacity.
So, the size of the buffer depends on the statistical fluctuation of the process and
the protective capacity of the non-constraint resources. The bigger the statistical
fluctuation the bigger the buffer has to be. And the greater the protective capacity
of the non-constraint resources the smaller the buffer can be.
Once the constraint is explored we have to go to the third step: Subordinate
everything else to the above decision. The above decision is the Drum. After we
have decided how to use the constraints time, we have to determine what we will
do with the other resources of the company.
Let us use the process below to help visualize what we should do with the other
resources.
X
15 parts/hour

Y
10 parts/hour

Let us suppose that resource Y is the constraint of this system. It can only produce
10 parts/hour. This means that the system as a whole can only sell 10 parts/hour.
Let us suppose that we have already done the Drum for resource Y. The next step
is to determine what to do with resource X.
Resource X can process 15 parts/hour. How many parts/hour should it process? Its
seems reasonable to say that it should only produce what can be sold, that is, 10
2003 Prof. Thomas Corbett

www.corbett.pro.br

parts/hour. If it produces more than that it will only increase the work-in-process
inventory. If we follow this logic, resource Xs efficiency will be of 66.6%, and
this does not constitute any problem in the TOC world, but it is a huge problem in
our traditional business world.
As a matter of fact TOC advocates the complete annihilation of these local optima
measurements. TOC says that the quest for high local efficiencies jeopardizes the
companys performance. If we look again at the above example we can see why.
To increase resource Xs efficiency from 66.6% to 100% we would have to feed it
with 15 parts/hour. As the system as a whole can only sell 10 parts/hour this
increase in Xs efficiency will only increase the WIP stock. If sales does not
increase but WIP does, for sure profitability will go down. One of the reasons Step
3 (Subordination) is necessary is to avoid unnecessary increases in costs and
inventories.
The non-constraint resources have to follow the constraints rhythm. That is why
the constraints schedule is called a Drum, because it determines the rhythm of the
entire troop. The non-constraint resources cannot let the constraint run out of
materials or it would stop and the systems performance would be jeopardized. On
the other hand, the non-constraint resources should not, in average, work faster
than the constraint because they would not be increasing the production but would
be increasing the level of WIP.
One of the conclusions of TOC is that the majority of the resources in a company
are non-constraint resources and therefore should stay idle part of the time. This is
one of the most controversial points in TOC and because of that it is one of the
hardest ones to be implemented. In our current business paradigm we believe that
high local efficiencies lead us to a good overall performance in the system as a
whole. But, as we have just seen, this is not true. However, even though we might
agree with TOC it is very hard to overcome this way of thinking because it is so
ingrained in everything we do.
As the non-constraint resources have more capacity than the constraint, we do not
need to schedule them. What DBR does is release the material necessary to fulfill
the Drum and determines that the other resources should work as fast as they can
when there is material waiting to be processed by them, otherwise the machine
should be idle and the operator should do something else (preventive maintenance,
quality circles, or other similar activities).
Taking the Drum as the point of departure and then subtracting the buffer
calculates this release time. Let us go back to the example of our Drum. Product
A, 200 units, is scheduled to be processed on the constraint on 02/05. Let us
2003 Prof. Thomas Corbett

www.corbett.pro.br

suppose the buffer size is 7 days, this means that we should release the
corresponding material on 01/29. This schedule of material release is called Rope,
because it is holding back the non-constraint resources so that they will follow the
constraints rhythm. Below we have copied the Drum and built the Rope.
Drum
Product

Quantity

Start
Date

A
Z
L
...

200
350
120

02/05
02/05
02/06

Hour

End
Date

Hour

5:03
21:43
15:13

02/05
02/06
02/06

21:43
15:13
19:13

Rope
Buffer = 7 days
Product

Quantity

Release
Date

Hour

A
Z
L
...

200
350
120

01/29
01/29
01/30

5:03
21:43
15:13

With the Rope we release only what the constraint can produce and by doing this
we guarantee that all the other resources will work at the same rhythm as the
constraint and therefore we will not have unnecessary increase in WIP.

PRODUCTION CONTROL
In the Drum-Buffer-Rope methodology the production control is done by using
Buffer Management. To understand how this control system works let us see in
more detail how the buffers operate.

Buffers
The buffer is created to protect a schedule. It is an anticipation in the release time
of materials so that we can certify ourselves that the schedule will be achieved.
Constraint Buffer - In the example above, the Drum is being protected by the 7
days anticipated release of the materials to the constraint. This buffer is called
Constraint Buffer.

2003 Prof. Thomas Corbett

www.corbett.pro.br

Shipping Buffer - The constraint schedule is not the only schedule that has to be
protected. The schedule of the delivery of orders also has to be protected, so that
the company can be reliable to its clients.
The constraint has to process the parts with enough time left so that the resources
after the constraint can process them. Here we are also using the Buffer concept.
This buffer is called Shipping buffer.
In the case given above the shipping buffer is of 5 days, that is, the constraint has
to finish processing the orders 5 days before the delivery date.
Assembly Buffer - When the parts that have been processed by the constraint have
to be assembled with parts that did not go through the constraint, it is necessary to
create another protection. In this case we have to use every part that went through
the constraint to form a final product; therefore we cannot allow any nonconstraint component to be missing (in which case we would miss a sale). For this
not to happen TOC created an Assembly Buffer.
This buffer makes us release with a pre-determined antecedence in time, the parts
that will be assembled with the parts that went through the constraint.

Work-In-Process Inventory
These 3 buffers are enough for us to protect our production schedule against the
uncertainties in the process. Not every company needs all 3 buffers; it depends on
the kind of process and on the location of the constraint.
If the constraint is in the market demand, the only needed buffer is the shipping
buffer, as there is no physical constraint in the process. In this case, all parts will
go through only one buffer, the shipping buffer.
If there is a physical constraint, there will be at least 2 buffers, the constraint
buffer and the shipping buffer. The assembly buffer will only be necessary if there
is an assembly that involves parts that went through the constraints and parts that
did not.
When there is a physical constraint, all parts go through one of two alternatives:
1. Those parts that go through the constraint will have in their flow two buffers,
constraint and shipping buffers
2. Those parts that are assembled with other parts that went through the
constraint will have two buffers in their flow, the assembly buffer and the
shipping buffer.
2003 Prof. Thomas Corbett

www.corbett.pro.br

When using DBR it is easy to know what is the size of the stock in progress. This
is dictated by the size of the buffers. For example: if there is a shipping buffer of 5
days and a constraint buffer of 7 days, the stock in progress will be of 12 days.
This because the constraint will be scheduled to process a certain order 5 days
before its delivery date and the raw material for this order will be released 7 days
before its schedule on the constraint. Therefore, the raw material will be released
12 days before the delivery date. This also shows that in order to reduce the WIP it
is necessary to reduce the buffers. It also shows that the buffer size determines the
production lead-time.

Buffer Size
The buffer is measured in time. In the above example we used a constraint buffer
of 7 days. This means that the raw material is released 7 days before its schedule
on the constraint. During these 7 days the parts should arrive at the constraint with
a certain slack. As a matter of fact, a buffer is well sized when it generates an
average stock in front of the schedule it is trying to protect equal to half the size of
the buffer.
In the above example this means that raw material is being released 7 days before
its schedule on the constraint and if the buffer is well sized, we should see an
average stock in front of the constraint of 3.5 days. If the average stock in front of
the constraint is less than 3.5 days (less than half the buffer) this means that the
constraint is not being well protected, therefore there is a considerable risk of
having interruptions in the schedule. On the other hand, if the average stock in
front of the constraint is greater than 3.5 days, this means there is more buffer than
necessary, which means more WIP than necessary. This increase in the buffer does
not increase protection, it only increases inventory.
That is why in buffer management we have always to try to maintain the average
stock in front of the schedule we are trying to protect equal to half the buffer.
When one is starting to implement DBR there is always the question of how to
determine the size of the buffers. There is no mathematical formula for this.
However, the process is quite simple. We have to keep in mind that the buffers are
needed to protect our schedule, that is, protect our sales. So, if you have to
determine the initial size of a buffer, should you pick a bigger buffer or a smaller
one?
If you pick a bigger buffer size than what you need, what will happen with the
stock in front of the constraint? It will be bigger than half the buffer. And,
according to what we have just seen, that means that we have more protection than

2003 Prof. Thomas Corbett

www.corbett.pro.br

10

we need. This is a clear signal that we can reduce the buffer size without loss of
sales.
Now, if you pick a buffer size that is too small, what will happen to the stock in
front of the constraint? It will be smaller that half the buffer, and might even
interrupt the constraints production.
Therefore, whenever we have to determine a buffer size we should start with a
conservative estimate, because by doing this we do not run the risk of losing sales.
Then, as we do our buffer management, we can start reducing the buffer size until
we reach the adequate level of protection.

Buffer Management
To really be able to guarantee the constraint and the order delivery schedules we
have to control production. We have to certify that the parts will reach their
destiny as planned. DBR uses Buffer Management to do this control.
As was already stated, every part goes through one or two buffers (depending on
the existence or not of a physical constraint). Therefore, to control production all
we have to do is control the buffers; that is, we have to verify if the parts are
arriving at the buffers as was planned.
All we have to do is certify that the parts are arriving as planned. If there is any
missing part, a hole in the buffer will be created.
Keeping a control on the buffer holes we can predict when a problem will interrupt
the schedule. Therefore, we can solve the problem before it jeopardizes our
schedule. Buffer management is a sonar that alerts us of any potential problems.
To control production all we have to do is keep an eye on the buffers. By doing
this we are able to control all the production by monitoring very few points of the
flow.

PROCESS OF ONGOING IMPROVEMENT


Buffer Management does not only allow for production control. It also makes
possible the implementation of a process of ongoing improvement.
We have seen that buffer management is done so that we can anticipate the
problems that can interrupt the smooth flow of production and by doing this we
can solve these problems before they jeopardize the companys performance.

2003 Prof. Thomas Corbett

www.corbett.pro.br

11

The holes in the buffers are a warning that some problem is coming up. These
holes are caused by the statistical fluctuations in the process. Therefore every time
a significant hole occurs we should register the occurrence and identify what
resource is causing this hole and why. Thus, we will form a history of the buffer
holes and their causes.
In order to improve the process, decreasing its statistical fluctuations, it is
necessary to attack the causes responsible for most of the buffer holes.
The history of the causes of the buffer holes will show which non-constraint
resources are responsible for the size of the buffers. By doing this, it will also be
indicating what resources the company should focus its improvement efforts.
In DBR, instead of trying to improve the performance of all resources, only the
constraint and the non-constraint resources that cause many holes in the buffer are
the ones that have to be improved.
When the performance of a non-constraint resource that is causing holes in the
buffer is improved, its statistical fluctuation will go down, attaining the so much
desired reduction of variability. As we have seen, the greater the statistical
fluctuation the greater the buffer size has to be. Therefore, if the statistical
fluctuation has decreased, the buffer size can be reduced without any loss in
protection. If the size of the buffer is decreased, the inventory in the system
decreases, and the company becomes more competitive.
This is how buffer management not only controls production but also enables a
process of ongoing improvement.

2003 Prof. Thomas Corbett

www.corbett.pro.br

You might also like