LEAN MANUFACTURING 12/04

Lean principles

In a traditional approach, how profitability can be increased?

- Increasing sales
- Decreasing costs
In which way?
o Reduce the features that our product is offering to the customer, allowing
to sustain lower manufacturing costs.
o Reduce material used.
o Reduce processing time; if we reduce the time that you dedicate to the
product you save cost (less use of machine and workers)
This was the traditional approach.

The approach that the lean manufacturing has adopted is totally different: in the
production, we try to look at which are the value adding activities and which not, and
we try to reduce as much as possible the not value adding activity.
To understand which activity add value for the customer, we should ask ourselves
which is the value for the customer.

- Value adding activities: those the customer demands, due to the fact that are
functional to the satisfaction of his/her need.
We can split the no value adding activities in two categories:
- Not value adding activities, non-avoidable in the short term (we can try to
avoid them by taking actions like redesigning the product or the manufacturing
system, or investing in innovation; these actions need time to be implemented)
- Not value adding activities, and avoidable in the short term. They are waste,
easy to eliminate.

Examples:
 Insert a ball bearing on its shaft  value adding activity, otherwise you don’t
make the product.
 Test a product before sending it to the customer  no value adding activity, in
the lean manufacturing perspective; if we develop a process whit no defect, we
don’t need quality control.
 Fix a rough material to the fixture of a vertical workstation  for customer,
take the material out is a value adding activities; however, we can find another
way to do it.
 Bring the electric motors from the warehouse to the place where the machine
tool is assembled  every transportation of material, unless it is required by
the customers, is waste.
 Urge the supplier for an overdue order  it may be necessary, but it is not a
adding, because no customers will be willing to pay people that make phone
calls.
  Plan internal and external personnels’ activities for the next week a customer
is not willing to pay for planning, it’s not a matter of the customer.

The 7 wastes of the lean manufacturing:

It’s fundamental to understand ways of eliminating these activities.
- Transportation. This has no value for the customer, as a customer do not care
that a product is manufactured 1000 km far from his home).
- Inventory. Inventories are present because you are not able to forecast what is
going to be asked by the market, so you produce in advance. It’s a supplier
fault, not a fault of the customer.
Inventories are considered value adding activities only in some cases like for
the wine or whisky, because its aging matter for the customer.
- Movement. We can refer to all the movements done by an operator in executing
a particular activity)
- Waiting. Waiting times are not functional for the creation of value for the final
customers. An operator that waits in its workstation is an inefficiency, a waste.
Waiting is the easiest waste to be generate and the hardest to be corrected,
because you don’t see waiting, “time does not pile up on the floor like material
wastes” (Henry Ford).
- Overproduction. It’s a waste because if you produce more, the risk is that
you’ve got obsolete products, that customer is not willing to buy.
- Overprocessing. Process more than the customer requires.
For instance, you do something and then the quality control check it. You
process twice, if the production is efficient, quality control is not needed.
- Defects.

The philosophy that lies behind this approach can be summarized by the following
scheme:

“House of the lean manufacturing”

 - Standardize work and Kaizen
Goal: standardize works, to not force the workers to find a solution any time a
problem occurs.
Methodologies to realize these goals: mutual trust, stability, supplier involvement,
robust product and processes.
Kaizen means improvement. Not stop to improve, analyse and improve in an iterative
way. Never stop thinking on how to improve.

- Just in time
The aim of the JIT is to have “the right part, at the right time, in the right amount”.
It’s fundamental to guarantee the integration between the various phases of the
process. In fact, this represents the only way to guarantee that a certain component,
in a specific version, locations and quantity is available for a determined activity.
by delivering customer demand without excess, it removes the seven wastes.
In order to reach this goal, just-in-time model provides the following tools:
o Continuous flow: keeping the workstream flowing smoothly without
interruptions.
o Pull system: starting new processes only if there is customer demand (avoiding
wastes).
o Takt time: optimizing production time to meet customer needs.
Tack time is the requested pace at which we should produce in order to satisfy
the customer demand. If cycle time is lower than tack time, we are able to
satisfy the demand.

- Jidoka
“Build in quality”: it’s important to make things of good quality from the beginning.
This objective can be pursued with:
o Automatic machine stop
o Fixed position line stop
o Error proofing
o Visual control
o Labour-machine efficiency

- In the end, with Kaizen + Jidoka + Just in time, we can reach the final
organization’s goal  reach the best quality, at the lowest possible cost
and in the shortest time.

SO, the key points of the lean philosophy are:

 Reduce all the activities that do not add value;
 Eliminate the wastes;
 Bring out weaknesses/defects/inefficiencies, in order to solve them. Eliminate
the cause of the problem, rather than avoiding that it gives troubles.
A way to visualize how to bring out the weaknesses is with the metaphor of the small
boat and the cliffs:
The boat represents the firm, rocks are problems and inefficiencies, while water
represents the inventories or queues.

The inventories don’t allow to see the inefficiencies. According to the traditional
approach, we should minimize the inventory in order to reach a level that allow the
boat to navigate over the rocks; in other words, the company rely on safety stocks,
which guarantee a certain service level.
On the contrary, the lean approach is different: there isn’t a minimum level of
inventories, since stocks represent a waste, so it is necessary to reduce gradually the
level of the water in order to let rock to emerge.

As a consequence, the boat will go through these problems and, in order to continue
on its path, it will be obliged to overcome them and try to remove them.
Once all problems are levelled, it will be possible to think to a further reduction of the
inventory level; the boat will certainly bump into other issues, which should be
resolved, as a sort of never-ending process. In this way we pursue a continuous
improvement.
Lean principles:
There are 4 principle that we need to implement in the lean approach:
1. Flow  We should design processes that flow in a continuous way, avoiding
interruption.
2. Pace  if you synchronize the whole system with the tack time required form
the market, you will not create problems (you not create inventories  no
waste).
3. Pull  it is necessary to work only as a reaction to customers’ requirements.
4. Zero defects we have to operate in order to obtain an internal quality that is
every time higher, without adding defeats.

How to implement these principles? With 5 main actions:

1. Identify value
It is necessary to understand where and to what customers attribute value.
2. Map the value stream
We need to look at those activities that make the flow valuable, so identify and
follow the flows of transformation, information and innovation, dividing for each
flow VA activities from NVA ones.
3. Create flow
It is required to find out all those aspects that potentially can stop the flow,
their causes, and then try to remove them.
4. Establish PULL
Create pull, create production management system which create PULL.
5. Seek perfection
Kaizen philosophy, we have to improve continuously, until the limit of
perfection.
13/04
 1. Identify value
The value is defined by the customer and not by the firm: the value is what the
customer is going to pay for, the reason for the customer to buy the product. We need
to use this as criteria to decide what kind of activities we should keep inside the
manufacturing process and what not.
Understanding which is the value for the customer and which criteria he use to decide
are actions that derive from a shift from an internal focus, on the company, to an
external focus, on the what the market want.
So, the way we calculate the price is changing.
(Internal perspective: the starting point are the costs
to realize a product, from which is defined the price)

(External perspective: who is driving is how much

the customer is willing to pay; the output is the cost,
that becomes a constraint, because we need to
design a manufacturing system in order that the cost
is lower than the value that the customer will pay)
 2. Map the value stream
In order to keep the costs aligned with what are the expectations from the market
point of view, we need to understand where the value for the customers is created,
mapping the value stream.
The value stream map is the result of the value stream mapping (VSM).
To improve the process, we start looking at the as is situation,
There are 3 main flows that create value:
- Product/service development.
- Managing information from customer’s order to delivery.
- Manufacturing process; manage product service realization from the inputs
elements to the final output.
For each of them, it is necessary to distinguish between VA and NVA activities, trying
to remove as much as possible the ones of the second category.

 3. Create flow
It’s necessary to analyse the flow of material through the entire company/factory,
analyse the production flow, and analyse any event, process, equipment, which is
stopping the flow to remove it. The ideal situation is to have a continuous flow.

 4. Establish PULL system

In a push system, producing in advance, there is the necessity to forecast the
demand; but there is always a variance in this forecast and the risk of errors.
Moreover, many of the production system that we develop are too complex to be
controlled. If you have a complex system, and you want to adopt a push system, you
need to have everything under control, and in a complex system this is too difficult.
The lean management say that is better a pull system, because is a reactive system.
You don’t need to manage how to forecast the future, because every single part is
reacting based on stimulus that it receives from the downstream parts.
Instead of trying to make a better forecast, we should focus on having a better
response capacity.
With a lean manufacturing approach, instead of investing in inventories, to respond to
a variable market, invest in create more production capacity.

- 5. Seek perfection
Continuous improvement (kaizen). Quality should be continuously improved. Cost and
wastes may be continuously reduced.
Value stream mapping

A value stream map represents a map that is

able to provide a simplified vision of reality,
inside which it is possible to include all those
aspects that are strictly functional to the
objectives of the analysis we want to carry out.
In particular, it is required to map the structure
of the firm at single plant level, although it is
also possible to specify the map at a macro
level, from more plants of the same company,
towards plants of different firms.

How to create the value stream map?

The first step consists in defining the concept of family: we need to put together those
products that require the same resources, as shown in the table below:

Equipment, technologies and competences needed are located on the columns, while
products represent the rows. In this sense, X indicates that a product requires this
specific equipment.
We can observe how A, B and C can be grouped in the same family. Obviously, the
overlapping in not total, because B and C require the addition of one resource.
However, the more products are similar, the more they will require the same
productive processes.
We can also observe how at the beginning there may be a non-perfect match between
products of the same family, but when we redesign the product/service or the
process, then the matching can be more precise.
Then we will map the main activities in the process, the key is to simplify the process
without losing so much information, to obtain a map which shows only the things
which matter.

We must keep in mind that is not required to ma all the possible processes, but focus
on main stream only: in fact, a map should have as few elements as possible; all the
rest must not be there in order to simplify.
The main KPI is the time, which can be declined in:

- Cycle time: time which goes from one item which leaves the assembly line and
the next one. Pace at which we produce products.
- Value added time: time that we spend working on the product and really add
value to the product (without waiting time or without quality control activity for
instance).
- Lead time: time between the moment when I sent the order and the moment
when I receive the output.

HOW TO BUILD THE VALUE STREAM MAP:

In order to better understand these concepts, we can build the present state matrix,
with reference to a manufacturing company that produces two types of stirrups (one
single family): the left and the right. In particular, the company wants to recur to the
VSM in order to evaluate the structure of its production system, identifying the
potential critical points of the system and implementing further developments.
Firstly, it is necessary to start from the identification and characterization of the type
of customer, from which we can derive the demand of the two products and where the
client attributes value to.
In order to do that, it is possible to recur to the basic symbology of the VSM, the data
box, which is able to put in evidence the most important information about the
reference entity:
Input:
In this case, to the client “State Street” is associated a
demand of 18400 pieces per month, with a mix of 12.000 left
products and 6.400 right products (therefore with a ration
equal to 2.1) Generally, order is delivered through trays of 20
pieces each (minimum batch size) and the firm works on two
shifts.
Quantity of pieces for month: 18400 pcs/mo. Tray = 20
pieces. 2 shifts per day.
Knowing that, it is necessary to identify the stage of the working process: you design
all the stations of the whole production process, also specifying the number of
operators.
In the process are present:
- A finished products warehouse, “shipping” (due to the fact that there are
inventories, it is not a dedicated resource);
- A final assembly stage, inside which 1 operator works;
- A pre-assembly phase, with 1 resources too;
- The final weld, with one dedicated operator, the C/O is not 0 because I need to
change the piece to weld, cleaning, adjusting, checking…..;
- The pre-weld, still with one resource;
- A phase of stamping, where in addition to the operator, includes a press of 200
tons. It is important to underline that, unlike the other stages where the
resource is dedicated,
In this case the press has to realize more than two products of the products
mix. In particular, for the reference family, the machine produces products
every two weeks (EPE – every part every), which means that every time it
produces a batch, it can fulfil the requirements of two weeks. We can also
understand it from the low value of the cycle time (1 second).

Then we add some other information: the data boxes (with information about CT cycle
time, C/O change over time, uptime or availability, number of shifts, etc.) and
inventories (the triangles; they identify the quantity of inventories, that we have
observed with the analysis of the system). The value of inventories is obtained
through observations of the reality.
Note: inventories between operations are justified for:
o FP, to react to unexpected order from customer;
o The batch production;
o Breakdown for maintenance;
o Setup time;
o Variability in the demand of the different processes;

Then we add information on the material flow, in input and output.

In this case, deliveries to costumer are performed once a day, instead on Tuesday and
Wednesday (two ships per week) we got from the supplier 500 feet of coils.
Then we add the information flow: how we control the overall production planning of
the system.
As regards the information flow, we can notice that:
- A customer sends forests about his requirements at 90, 60 and 30 days, but
also daily demand, and this data are received as input of the production
control;
- Then, production control processes forecast from customers to send suppliers
the weekly order and the weekly schedule to each stage. We can use this
information to plan the production for next week, and we print for each stage
the schedule.
- In addition, daily, the production control elaborates the ship schedule, on the
basis of the orders received during the day form the clients: this schedule, sent
to the shipping department, indicates what need to be shipped the next day.
It is important to underline how straight lines put in evidence a paper information
flow, while arrow lines an electronic information flow.

Now we have to understand the lead time of the process.

In order to transform pieces into time, we need to recur to the Little’s law.

Considering our example, we can ask ourselves: once the piece has left one stage of
production, how much does it have to wait, on average, before being sent?
For instance: after the assembly, the piece is inserted at the end of an average queue
equals to 2.700 + 1.440 = 4.140 pieces; since every day, 960 pieces are sent to the
customer. This means that, on average a piece has to wait 4.140/960 = 4,5 days.
Recurring to the Little’s law and, it is now possible to draw a timeline.
In the upper part of the line we put waiting time that do not add value, while in the
bottom part the processing time. It is important to underline how processing time and
value adding time are the same.
If we subtract from the total lead time the processing time, we will obtain the waiting
time. Change overtime is a waste so it is not appropriate to include it in the
processing time; as a consequence, we consider only the cycle time.
Inventories include already set-up because it is a measure of the problems.

The system performances will be measured by the ratio between the processing time
and the total lead time (the so called “flow index”).

The future state 20/04

We have seen how to design the value stream of the current situation (AS IS).

Now we have to design the future state, defining how we want to redesign our
factory.

The lean manufacturing

theory is using a
methodology that goes
through a number of steps,
that are a kind of check
list. We should ask our self
a series of question and
find the answer to all the
questions.
 - 1. What is the takt time that we have in our system?
How to calculate the takt time? We have to start from the available time. Example:

The takt time is the available time that we have to produce one unit. If we are able to
keep a pace where a unit is produced in a time which is below 60 sec, we will be able
to fulfil the demand. If we have a cycle time that is higher than the takt time isn’t
good. It’s the “maximum cycle time” allowed to fulfil the demand.

If we go to the value stream, we can see that we have one station (the 4th station)
with working time of 62 sec. This is not good, we have to stay below 60 sec.

What can we do?

We can try to rebalance the tasks, distributing the workload between the station in
order to assign the same quantity of workload. If we reallocate tasks, we go below the
takt time (we do not consider the stamping, the first station, because stamping is
done by a machine).

But if we do this, we can see that there are “opportunities”, in each station there is
empty amount of time. Can we improve? Can we distribute the workload between 3
stations instead of 4 stations?
If we are able to increase the efficiency of the first three stations, we can think about
cutting the last 7 seconds of operation, to use only 3 stations.

With this procedure we are able to reach the minimum number of people, and to
quantify the waste. In this case we are doing with 4 people a work that should be
done with 3 people.

- 2. Will you build to a finished goods supermarket, or directly to shipping?

Directly to shipping:
We produce and we deliver the order activating the process after the acquisition
of the order itself.
We receive a customer order, the goes to the assembly (the last point of our
system), the assembly produces and the quantity which is on the order is
 moved to the shipping area, where the product is directly shipped to the
customer. There are no warehouse or supermarket in the middle.

Supermarket (make to stock):

In supermarket you have the product available on the shelves.
You receive the customer requirement, that issue an order to the shipping area.
The shipping receives the order, turns back to the supermarket (warehouse)
where there are all the possible products available, there you pick the products
according to the quantity which is in the supermarket and any time we remove
a product from a supermarket, this supermarket issue an order back to the
assembly to replenish the consume quantity of the product.
Supermarket is where a certain level of inventory is stored. This could be raw
material, work in progress or finished goods inventory. The process will
continue to withdraw items from the supermarket for uninterrupted production.
Supermarket will get replenished once the inventory is used by the process.

- 3. Where can you use continuous flow processing?

The third fundamental topic about lean regards the flow processing: when does it
sense to produce with separate workstations or with a continuous flow?
A system with disconnected stations can be represented as follow:

Phases are not coupled: A produces following its schedule,

retrieving materials from the upstream buffer and putting
the final piece in the downstream one; so do the B and C,
referring to their relative upstream warehouses.
The flow in interrupted, is not continuous, since there are
inventories in between the stations.
On the contrary, in a continuous flow, the system is not decoupled, there are not
stocks or intermediate buffers or batches: inventories are located only at the
beginning of the flow.

In this case, the input box of the first station became the input of the whole line, the
output of the third station is the output of the whole line and between the stations
there are no inventories. This is how to create a continuous flow.

Looking at the welding + assembly station, the current situation is this:

We can note how the structure of the stage favors the
continuous flow, the teamwork, the interaction between
resources and the mutual support. This structure is not casual
for two reasons:
- Coupling, we have the possibility to reduce the processing
time, putting resources closer to each other, and spare an
operator;
- This configuration is the one that allows resources to relate
to each other, enabling the continuous improvement process.
To allow the continuous flow in the station, we should remove the set ups: we have to
redesigning the process in order to remove them.

- 4. Where will you need to use supermarket pull systems to control production
of upstream processes?
If we have a pull system, the upstream processes are activated by station which is
sending upstream the signal to start to do something. The activation comes from a
downstream station, since is a PULL system (not from the production control, as in the
as is situation).
Where we position the ‘external
force’ that activate the system?
This is the meaning of the
question 4.

How a pull system with kanban

works?
Let’s take two stations, A and B
(the flow of product is from A to
B). One is called supplying
process of B, the other is called customer process of A.
The withdrawal kanban works in this way: anytime B needs a component to do the
process, the withdrawal kanban is used to withdraw from the upstream supermarket
the quantity of components that B need in order to do the products.
The production kanban goes from the supermarket upstream to the supply process A:
the supermarket asks to station A for the production of the quantity that must be
replenished.
Where is the best position for a supermarket? The best position of supermarket is
here:

Because we have a huge setup in the first station, and then the best position is to put
the supermarket as much as possible upstream, just downstream the station that has
a higher changeover.
This is what is going to look like the actual future state:
Sum up of the changes:
We decide to have a supermarket for the shipping, and a supermarket just below the
stamping, to deal with the high changeover that we have in the stamping.
The two welding and two assembly station were merged into one continuous flow line,
where we have welding and assembly all together and the 3 operators. The tack time
is 60 sec, with CT 56 sec.
The daily order come to the production control, then go to the shipping. The shipping
ships the products; as soon you consume the product in the supermarket, you issue a
kanban to the welding assembly station to replenish. The welding and assembly needs
components, so issue a withdrawal kanban to the supermarket, that then create
production in the stamping station.

5. At what single point in the production chain (the “pacemaker process”) will you
schedule production?
General situation with two examples:

Focus on the upper part: if you have a pull between every process, you put the
pacemaker in the last one process before the pull system; the other processes will be
regulated by the process 4.
Second situation: we have pull system only between process 1 and 2, in between of
the other processes there is a FIFO system. Here you position the scheduling point at
the latest process before the last supermarket.
So, pacemaker is put as much as possible upstream close to the last supermarket, at
the last process that is followed by the pull system.

NOTE: What kind of options we have to synchronize two stations, one close to the
other?
4 options:
- Supermarket (B activate the A, system with withdrawal kanban and production
kanban)
 - FIFO: push system, where A activate B; A
push to B and B do something, one batch at
the time. B waits for A.
- OPF, one piece flow: push system, you push
down from A to B one piece at the time.
- PUSH system, the stations are activated
independently one from the other, by the
production control.

After putting the pacemaker, actual situation:

After these 5 steps, we already see some improvements:

 6. How will you level the production mix at the pacemaker process?
The pacemaker become the heart of the whole system, where we control the system.
We need to have a levelled production, because if we want to implement the
continuous flow, we need to level as much as possible in a uniform way the production
in the different phases.
The tool used to obtain a levelled production is called Load-levelling box (or Heijunka
box). Heijunka try to minimize the impact of crashes and peaks of the customer
demand, regulating the correct and constant production flow in order to react to
fluctuation of the demand. The concept of Heijunka strategy is to concentrate
productivity on small quantities of materials, that are more manageable in terms of
planning of the different phases.
In the Heijunka box, the rows represent the different types of product and the column
represent the different time intervals, that are all equal (20 minutes for instance).
Looking to the box, we can see that we have to produce a certain amount of product
A, a certain amount of product B, and so on, till the composition of a mix batch.
In this way, we make sure that we produce a sufficient quantity to satisfy the mean
customer demand for the products portfolio.
Example:

Let’s assume that we produce 3 products (A, B and C). the Heijunka box is containing
all the kanbas which are showing the production quantity; assume that the minimum
quantity is a quantity that requires 20 min to be produced (because the minimum
shipping quantity is 20 pieces per box  60 sec per piece (takt time) multiplied per 20
pieces = 20 minutes). This time will be the pitch: every 20 minutes there will be a
release of the kanban to the production and a withdraw of the material produced.
If we have 3 products, a way to level the production is to produce a pitch of A, a pitch
of B and a pitch of C, and then come back. In this way we are creating the most
possible uniform levelled production.
The Heijunka box is used by the pacemaker to manage the situation.
Every 20 minutes, the pacemaker bring to the production (pacemaker process) the
next kanban and withdraw the previous kanban, with the product produced in the last
production phase.
We can have many alternatives:
 Option A (without Heijunka box)
The customer sends the order to the
warehouse, which picks the required
boxes and prepares them for the
shipping. Picking a tray, one or more
transport kanban are detached and
sent to the cell (as an input for the
replacement) without levelling (the
cell sees the arrival of the kanban on
the basis of the same logic of picking
in the warehouse).

 Option B (with Heijunka box OXOX):

The second option requires to level

the kanban before they are
transferred to the cell: the system
does not receive the entire amount of
batches, ma one at a time, on the
basis of the levelling logics.

 Option C (put the hejiunjka box between the daily order and the preparation of
the shipping):

Example of how Hejunka box works:

The pacemaker (the guy) goes to the Heijunka box, takes the kanban and drop the
kanban in the process. At the same time the pacemaker picks up the finished pitch
quantity and take it to the supermarket. How often the guy must do the pitch in our
example? Every 20 minutes. The problem can be that the whole tour may takes
longer than 20 minutes.
  7. What increment of work will you consistently release and take away at the
pacemaker process?
How often we should do the pitch. It can be every
pitch, every hour, every shift…
Frequent pitches means more quality and
efficiency, but implies a higher waste in terms of
pacemaker’s movement.

 8. What process improvements will be necessary for the value stream to flow as
your future-state design specifies?
Final future state:

Actual performances of the system:

Lean tools 26/04

What kind of methodology we can use to design the future state of the value stream
mapping?

1. IDENTIFY A PRODUCT/SERVICE FAMILY: PROUCTION FLOW ANALYSIS

One of the principles at the basis of lean redesign is the possibility to dedicate
resources. However, especially in those realities characterized by a high variability,
the single types of products have not enough volumes to guarantee the dedication of
a resource. Therefore, it is necessary to group them together and form a family.
We can understand that the first step is to identify a family in order to devote
resources to a specific family.

Thought the production flow analysis we identify the families through the analysis of
the production flow of each product. If the products have more or less the same
production flow, they belong to the same family (use the same machines for
instance).
In the production flow analysis, we start from the product/process matrix, that have
on one dimension all the different products that we have and on the other dimension
all the machines. The cross on one cell means that this product has to visit this
machine.

Could happen that a product look very similar in terms of process to another. The left
matrix is a revised version of the matrix on the right, that allow to see that there are
two groups of products, in which the products have in common very similar flows
because they are using the same machines. This means that there are two families,
and we have to make two different value stream maps.
Product A requires also machine n2 and product G require machine n3, they are out
line, that not belong exactly to the characteristic of the family.
The machine 3 is called bottleneck because all the products are going thought the
machine.
What kind of tool we can use to go from the left matrix to the matrix on the right??
- RANK ORDER CLUSTERING
It’s an algorithm to redesign the matrix.
Example:
Looking at this matrix we are not able to identify any family. How to identify them?
We reshuffle the sequence of the rows and of the columns with an alternative
approach, in an iterative way. The algorithm stops when, reshuffling rows and
columns, we get the same results.
We interpret the row as a binary digit number, and we calculate the value of each
row.

Now we rank them, putting the first one with the highest value and the last one with
the lowest value.

Now we do the same changing the dimension: we calculate the values over the
columns and reorder them.
Now we go back to calculate the values of the different rows. We should reshuffle the
rows, ranking them again.
After this procedure of “computing & ranking”, we see that we have 3 families.
 2. CREATE FLOW
5S approach:
In the creation of the flow, so in the fixing process of
the possible problems, we can adopt the 5S approach:
- sort: find the problem.
- set in order: solve the problem.
- shine: clean it (remove all the disturbances)
- standardize: create the standardisation; we
should create processes with the purpose not to
give possibility to workers to do mistakes.
- sustain: make it to work in the long term
Due to the fact that everything that stops the flow can
be classified as a waste, we have to understand which could be the causes of these
stops.
What does stop the flow?
- Production change (set up)
- Wastes/lack of necessary information/re-entrant loops
- Breakdowns
- Batch (quantity)
- Batch (time), you need to collect more activities to be done at the same time in
order to deliver them
For each of these elements, we will put in evidence the possible solutions, so as to
avoid the insurgency.

 Production change: SMED

To create continuous flow, we have to try to reduce the setup time as much as
possible. To do this, we can adopt the SMED (single minute exchange of die)
approach.
According to this approach, there are two types of set up activities:
- internal setup activities: carried out when the machine is not working (like
change the battery).
- external setup activities: carried out while the machine is working.
If we identify all the setup activities done (internal and external) and we put them in
sequence, on a row, we will obtain this result:

The SMED approach lead to reorganize the sequence of internal and external activities
in order to reduce the machine stop, grouping together all the external activities at
the beginning or at the end of the set up.
Step of the SMED:
- look at the activities required by the setup and classify them in external and
internal. In this phase internal and external operations are often mixed.
Example of criticality:
a. equipment are brought at the machine in internal setup (while machine
stops)
b. tools are placed far from the machine
c. broken tools
d. failure to verify the correct functioning of the equipment, before starting
the setup or before starting production.
- Re-organise activities in order to have the greatest number of external activities
grouped together at the beginning of the setup, or at the end of it. Switch the
largest number of activities (time) from internal to external.
- Simplify and shorten the duration of internal and external operations.
Therefore we have the possibility to prepare what is needed under a preventive logic,
avoiding the necessity of a too long machine’s stop and allowing a flexible interchange
of production.
Examples of how to reduce the time of the single activities (example of engineering to
make it simple and reduce the time)

 Eliminate wastes and reworks

Scrap and rework are additional causes of stops in the flow and should be eliminated.
In the lean approach, reduce scrape does not mean increasing the rigidity of the
quality control, but increasing the efficiency of the production process, fixing any
possible problem.
There are simple methodologies to do this.
We have the fishbowl diagram (top right), that
allow to visualize the possible problems.
We create a logic sequence of the problem. We
start form the head of the fish, identifying the
principal problem: scraps. Then we ask
ourselves “which is the first level of reasons
why we have scraps?”. We identify all the
possible reasons directly related to the scraps and we represent them with bones.
After this, we identify the “sub-reasons” of the “first level reasons”, and we represent
also them in the diagram.
In this way we create the hierarchy of the possible cause which create a problem,
which is creating another problem, which is creating the scrap (the principal problem).
Once we have doon the fishbone, we have to look at the occurrences of the different
problems, and we start to fix the problems from the one with the highest occurrence,
to the one with the lowest.

 Eliminate process stops (breakdown): TPM (total productive maintenance)

Try to remove any possible cause which stop the flow, by doing total productive
maintenance  try to increase the MTBF (mean time between failure)
The MTBF depends on the reliability of every single component that the machine is
made of. Higher is the reliability of every single component, higher will be the
probability that the system works for a longer time without stopping (i.e., longer will
be the MTBF).
The time in which we repair/maintain the machine is called MTTR (mean time to
repair); it depends on how easy is to repair the machine.
The actual capacity of a machine is computed as:

The actual capacity of a system is the theoretical capacity (how many pieces system is
producing if the system is working) multiplied by the availability (%), that reduce the
theoretical capacity due to the fact that sometimes the system doesn’t work.
The availability is the probability that the machine is working.
How to increase availably? We can try to increase as much as possible the MTBF or we
can try to decrease as much as possible the MTTR  we should improve the reliability
of the component and also reduce the time that you need to fix a failure.

 Eliminate mix batching

Till the 80s’, all the main assembly lines were essentially of two typologies:
- Single model: realizing one single product.
- Multi model: lines capable to realize more models, under the condition of
operating following batching logics:

The setups between a batch and another don’t allow the continuous flow.
If we are able to remove the setup, we can create a mixed model production: we
produce one model of the product after the other in a totally random sequence.
In this way we are allowing to reach the maximum flexibility possible and to follow
exactly what the customer wants, with a continuous flow production.
Example: assembly line of car, where every car is different from the other (colour,
interior, components, variants…).

 Level the demand

We should try to avoid that variation in the volume required by the market will be
propagated upstream in your system. How to do it: buffer, that decouple the market
variation from the internal variation.
In this way we release to the manufacturing system a levelled demand, with the
purpose to not propagate outside variation in the internal production system and to
allow the production system to work in a levelled way, creating continuous flow.
Example:

We have 4 products and the quantity required by the demand in a week for each
product. The overall weekly volume is 200 units. These products are made by
components (part 1, part2, …). For instance, to produce product 1 we need 2 units of
part 1, 1 of part 2, 3 of part 4, 2 of part 5, in total 8 components.
The average service factor represents the average number of units of part x to be
present in one unit of the finished product.
For instance, average service factor of part 1:
We produce 75+50+50+25 = 200 units in a week.
How many units of part 1 we need in a week?
2*75+0*50+0*50+1*25= 175
Average service factor = 175/200= 0,875
Every service factor could be larger than 1.
If we have to produce 200 units of finished products in a week, which is made by 5
working days, what will be the average production capacity?
200/5 = 40 units per day.
Producing 40 units per day we satisfy the demand of each product.
Part 1 is needed any time we produce product 1 and product 4; we need 2 piece for
product 1 and 1 for product 4.
So, on Monday we need 80 units of part 1 (40*2) to produce product 1, on Tuesday
we need 70 units of part 1 to produce product 1, on Friday we need 25 nits of part 1
(25*1) to produce product 4.
We can see that part 1 is only produced on Monday, Tuesday and Friday. This is not
an issue because if the production of part one is balanced with the production of other
components maybe at the component level, the production will be quite levelled.
But in order to produce product 1 on Monday, we also need part 2,4 and 5; so we
need to multiply 40 units of product 1 by 1, by 3, by 2. So we have that we need also
120 units of part 4, 80 units of part 5.
Any time we produce a finished product, must probably we will allocate some demand
to supplier one and supplier two. So the supplier 2 produce only part 2 on Monday.
What kind of demand on Monday the two suppliers will be generated by the customer
(the producer of the finished product)? The first supplier got 280 production volume
on Monday and the second supplier only 40 production volume.
In the next days the production volume for supplier 1 are 285,320…; for supplier 2
are 45, 80…

So the example shows that even though we started form a finished product very
levelled demand, when we go to the supplier level, this level demand could be not
levelled anymore.
Also, it could be that this list of components 1-9 sometimes they could be made by
sub-components and the system can diverge.
How to avoid this issue?
It is true that we have levelled the overall production at the product level; what is not
levelled is the mix of the product. We are producing first product 1, then product 2,
then 3 and then 4. Maybe a good way could be to produce every day either product 1,
2, 3 and 4. So maybe we should re design the production schedule. How?
Let’s produce every day 1/5 of the weekly demand of very product.
75 of product 1 divided by 5 day = 15 units every day
Etc. for each product.
If we do this, the result will be a very levelled demand for supplier 1 and 2. But
Monday is made by two shifts (mooring and afternoon); we need to decide what to
produce in the morning and what in the afternoon. If we apply a “natural approach”,
20 should be produced in the morning and 20 in the afternoon. Since we have product
1 and 2 and 2, 3 and 4 in the afternoon, we create anyhow a difference in the demand
that suppliers are facing in the morning and in the afternoon. So even though from
the finished products every single day is the same one after the other, since the day is
made by the morning production and the afternoon production, what the suppliers are
facing is anyhow a not levelled production. (See w2 in table).
So, we should distribute the level also thought the morning and the afternoon (see w3
in table). The consequence is a levelled demand for the suppliers.

 Eliminate time batching

In many offices, and service activities, time batching is frequent; we can think for
example:
- To process invoices once a week;
- To process subscription requests every other day;
- To process change requests of the design of a product once every 15 days.
These cannot happen: the system must be structured and organized in order to be in
the condition of providing every activity “exactly when needed”, adapting perfectly to
customer’s requirements.

3. LINK EVERITHING TO THE CUSTOMER (PULL)

In the lean approach, the objective is to be able to link every activity to the
necessities and the requirements of the customer. A system can be considered lean if
it is structured in order to react, as much as possible, to what customer want,
You should not produce if the customer has not made a request, otherwise you
overproduce.
Forecasts are always wrong and plans are destined to change, so to be efficient we
have to work on reliable information. However, due to the fact that market is not
willing to wait, there is a need of an extremely fast to react system. So to have a pull
system, lead time should be reduced.

Pull system: every time the customer wants something, we serve it and this activate a
new production.
All of the customer requests don’t have similar priority. We
put a kind of system to create priority to be produced ant
therefore fulfil the demand.

Example of pull system:

The market activates the last process, that activates the previous one, that activate
the previous one.

Slightly different from the pull system is the sequential pull system; we have a
sequential pull when the customer order doesn’t go to the last point of the chain but
goes to the supermarket between the first two process, where supermarket is
positioned. The customer order goes to the last process before the supermarket.
The system is activated from process 2, that receive the input from the demand and
activate process 1 and, when process 2 finish to do the job, push to process 3.

Is still a pull system because the whole system is anyhow activated by the market,
even though some parts are working not as a pull system but as a FIFO.
PULL advantages:
- Simplicity: when something happens in the market, you start. (You don’t need
forecast)
- Control over stocks
- No overproduction from the supplier
- An easy way to push towards constant improvements
- A single planning point, everything else just react
- Limited amplification of production and inventory across the supply chain
(bullwhip effect reduced)
The bullwhip effect is the increase in the demand variability as we go far form
the final market, moving upstream thought the supply chain.
Creating continuous flow 10/05

After the future state we for sure got improvements

on the single plant side. Now, the question is, are
these improvements enough? Are there any other
activities in order to get more level of performance?
To answer this question, we could concentrate no
more on the single plant but on the process level.
We were before studying and implementing the lean
approach at the level of the single plant, now, we
need to focus on the process level and create
continuous flow there.

Example:
We are in a plant that produce components for trucks; in particular, we are creating 3
types of products: truck S, truck L and truck A, that belong to the same family, as
shown in the table:

The value stream of the plant looks like this:

First we have to extrude the tube, then we have the end forming, then we need to
bend this tube and then we go thought the assembly; then there is a crimping process
on the product, then quality test of the product and then we put the product in the
warehouse, that prepare it for the shipping.
The demand is 1380 per day. The delivery package is made from 30 pieces, we have
2 shifts per day.
We have 4 deliveries per day; the customer is called State Street Assembly. They are
sending a forecast to us. The daily order is managed by the production control unit,
that is pushing the order received by the customer into a daily schedule to the
shipping area, to ship the quantity every day. The quantity is spilt in 4 delivery per
day, as we said.
The finished product are expected to be always available in the warehouse for the
shipping.
The schedule that is received by every single process through the entire factory is
based on a weekly schedule. So weekly schedule for the processes and daily schedule
for the shipping.
The triangles are the warehouse between the operations. Every warehouse has a
number of units at stock.
For all the operations we have information related to the cycle time and to the
changeover.
We can calculate the lead time, thanks to the little law (quantity in the warehouse
divided by the daily demand). At the end we can calculate the overall lead time, that
is the sum of all the lead time, and the processing time that is the sum of the lead
times and the processing time of every operation.
Moreover, we have to buy raw materials (coils); we receive it from supplier Michigan
Steel co., twice per week.
This is the AS IS value stream map.
NOW, we need to go thought the 8 steps to design the future state design.
After going thought the process, the future state value stream map obtained is this:

We got the daily ship schedule; shipping is now using kanban to pick up the products
from a supermarket and the supermarket activate the production upstream by using a
kanban.
The tube bending, the sub assembly, final assembly, crimping and test, now are
unified in a unique cell (the assembly cell), with four workers in it.
The processing time of the cell is 159 seconds, that is the sum of 22+74+20+19+24.
The end forming and the tube extrusion remain the same.
We implemented the pull system with the kanban.

The implementation of the value stream map was able to increase the quality and
decrease the lead time, but the overall improvements are still not enough.
We see that we have a kind of improvement, the processing time has not been
modified because we worked at the plant level and not at the process level, we didn’t
improve the performance of the activities and we have just revised the layout and the
inventories.
In order to satisfy completely the company, we should have more improvements, and
in order to do so we have to act at the process level.

In order to have an improvement in the performance we need to zoom in every

different cell and line in order to re-design them. For example, we could take into
consideration the assembly line.
In the assembly line, we have the two bender, assembly one, assembly two, the
crimper and the testing.
We have 4 operators, instead of 6 of the previous situation.
The first operator is moving the tube which needed to be loaded to the bender (an
automatic machine); while the bender was producing the product sent by the
operator, the worker was taking the last finished product, which was brought to the
assembly 2, to the crimper and to the tester.
The worker 3 does also internal logistics activity. Worker 4 has to take the good from
the crimper, load it in the tester and pack the finished good.

According to the tack time, this should be the plan of the overall hourly production of
the pieces. 690 is the production of one shift.

If we look at the actual production in the future state, it is different from the planned
one, we are underperforming. Also the demand is not leveled. The fluctuation is very
high.
If we look at the distribution of the production, it goes up and down hour by hour.
We need to modify and redesign the assembly cell and, in order to do so, we need to
detect all the possible signals that are showing inefficiency
We have to ask ourselves: why there is so much variation?
Causes of variation:

Analysing the situation, we can see that there are at least four possible reasons for
this variation:
- Since the cell is too wide, the operator one has to leave its position every 25
pieces, he stops the flow.
 - There are inventories between operations: anytime we have inventories, first is
a waste, second you decouple the different phases and therefore you can create
fluctuation.
- Production operators are anchored to their machines. If you have a not well-
balanced assembly line and you keep the people anchored, you are creating
constraints that can create fluctuation.
- The physical layout is not nice, the u
shape is wide, it should be narrower.
We should focus on this points to try to
remove at least partially these possible
reasons of variation. Our target is to reach the
continuous flow.

The objective is to reach the target

performances, with the implementation of
some change: squeeze the space (the U
layout), reduce the assembly lead time,
reduce the number of workers and try to
implement the pacemaker. In so doing, we
will be able to have a higher level of
productivity.

How to divide the mix between cells?

The first question we need to ask ourselves is:
Is the layout the right one in order to reach a good level of productivity? There is a
better distribution of the items between cells?

Option 1: instead of having 2 separate sells, we could try to produce both of the
products in order to smooth the higher volume in production A and few volume in
production B in a way that they can compensate in the same cell.
Option 2: we could decide to focus the production of the cell on a particular product
(e.g. we could try to separate the products and make them work in different sells). In
this case, in order to saturate the cell we must have (approximately) the same
volume of product A and B.
How to choose which items to produce in the same cell? We need to take into
consideration some aspects:
- Flexibility
- Variation in total work content: if you have two products with different work
content, that will be more difficult to put them in the same cell.
- Similarity in processing steps and equipment: more similar they are, easier is to
merge them in the same cell.
- Takt time: the takt time should be similar too.
- Customer demand
- Customer location

Takt time
Note: how to measure the takt time and the real production time in order to check if
we are below it.
It is very important every time we work at process level to do measurement.
Guidelines for the measurement: we need to collect real time of the processes (do not
rely on standard time); position yourself where you can see the operator’s hand
motions; time each work element in a sperate way; time several cycles of each work
element; always separate operator time and machine time (machines and people have
to be measured separately); select the lowest repeatable time for each element (look
at what is the bottleneck, that will be removed); remember shop floor courtesy.

Implementing the approach to our example, we need to use the so called paper
kaizen:

We have listed all the processes that the workers and the machines are doing in the
cell. In the column “work elements” we put all the machines, the workers and the
other elements; next to it we put the time, which is the result of out timing (that we
measure many) expressing the average of what we measure.
Form the paper kaizen, we design the so-called operator balance chart.

We need to take each of the 4 columns (stations) and we go one by one and consider
only the activities performed at the manual level, the purpose is to see if the workload
of each operator is somehow balanced and if it is below the takt time.
Here we don’t have to put the task performed by the machine because it is the
OPERATOR balance chart.

How much automation?

After clustering all the activities done by each operator in each station, we can now
think of removing the waste. For example, is there any activity that we can try to
make more automated than it already is? Can we have some of the activities
performed by the workers which can be performed by the machine and thus reducing
the number of workers?
To answer to these questions, we have to consider that there is a kind of cycle of the
activities in our cell:
According to this, we can define what is called the different levels of automation. We
have 5 levels of automation according to which of the previous task is automated and
which is not automated.

In the first level, we load the machine manually, the machine cycle is done manually,
the unload is done manually, and the transfer part is done manually.
In the second level…
In the third…
In the 5th level everything is automated.
You have to decide at which level you will stop when you redesign the process.

Possible improvements at the process level that we can be do:

 Eliminate wait while machine cycles: wen could try to redesign the system in
such a way the worker is not waiting anymore.
Looking at what is normally
happening in a cell, if you have
three stations in the cell and
you have three workers, the
worker 1 is loading the machine
one, that push the machine two
to work. Then the worker 2 do
the loading the machine 2 and
push the machine 3 to work.

In the new view, instead of having three workers we can create a situation where
there is only one worker that goes thought the stations to do the work and load
and unload.
 The operator loads the machine one and
starts the machine working. After he
goes to machine 2, load it and starts it.
the same is for the third machine. When
the third machine is started, he goes
back to the first one that has finished
and starts load the new batch. We don’t
care if the machine waits for the
operator: the goal now is saturating the
operator not the machine.

 Introduce auto-eject at Ass I, Ass II, Crimper, Tester (unclamp, remove, set
aside). We move from level 2 of automation to level 3, if we introduce auto-
ejection system we can improve the unloading part done by the machine. By
doing so, some activities in the APEX PAPER KAIZEN could be eliminated since
the operator is not doing them anymore.
 Convert out-of-cycle work (loading the bender with a batch of 25 tubes) to in-
cycle-work of loading one tube every cycle. Allow to load the machine while the
machine is working, for example we could create an autoloading system like a
buffer.

If we do all this changes, the new paper kaizen will be this:

Form a total of 118 sec, we remove wastes, and
we reach total time of 88 sec:

Meet the takt time

We need to ensure that we get the takt time:
sometimes it can happen that we are not
meeting it. If we don’t meet the TT, we can try
to:
- Reduce the set up time;
- Increase availability;
- Kaizen the load, start, unload process:
increase the quality and decrease the
number of scraps;
- Eliminate waste in the machine cycle itself;
- Split apart some of the tasks a bottleneck
machine is performing and use more than
one machine;
- Install 2 machines of the same type in parallel: instead of producing 1 piece you
produce 20 pieces, so the production rate is still 10 seconds for each machine
and overall;
- Create 2 cells instead of one: if there are more than one stations that is over
the TT is better if we split the two cells;
- If everything else fails: remove the bottleneck equipment from the cell,
decouple it and operate in batch.

What kind of machines? What sizing?

Another question we could ask ourselves is: what kind of machine is it better to use
for each production process? Do we prefer to have multi functions machines (1) or a
situation where we have more specialized machines (2)?
1) Few multifunctional machines: in this case
the machines will be few but they will be
able to perform complex processes.
Pro: economies of scale.
Cons: if we have an over demand, we can
be too close from the takt time.
2) Many and specialized machines: in this case,
each machine would be specialized in a
specific procedure.
 Pro: higher room to absorb demand
variation.
Cons: reliability (because if one machine
stops working, the entire system stops
working), mix flexibility is lower because
each machine is specialized.

Guidelines for supporting in choosing between multifunctional or specialized

machines:
- Small equipment for specialized machines or multi-task equipment for
multifunctional
- Introduce auto-eject (Level 3 Automation) whenever operators must use both
hands to handle the part.
- Install one-touch automation where possible.
- Avoid batching.
- Incorporate sensors to signal abnormal conditions and even automatically stop
machines if necessary, so operators do not need to watch machines during their
cycle.
- Design in maintainability.
- At the pacemaker process, strive to device machine changeovers between
different end items that take less than one takt time cycle.

Layout of cells
Which is the best layout of the cell?
Normally there is one important guideline: arrange machines. Workstations, and
material presentation devices as if only one operator makes the product from the
beginning to the end.
The best configuration of the workstation is the U cell: people are close to the other
and waste less time in the movement.
This is the layout of a cell:
Guidelines for the layout:
o Place machines and workstations close together to minimize walking distance
o Remove obstacles from the efficient operator walking path
o Try to keep the inside width of a cell at around five feet to allow flexibility in
reallocating work elements among team members.
o Eliminates spaces and surfaces where work-in process inventory can be
accumulated: if you don’t give space to people, they will be forced to find
solutions
o Maintain consistent heights for work surfaces and points of use
o Locate the leadoff and final processes near one to another
o Avoid-up and down and front to back transfers Guidelines for cell layout
o Use gravity to assist operators in placing parts and moving materials whenever
possible.
o Install flexible utility drops from the ceiling to make layout adjustments easier.
o Absolutely ensure safety and good ergonomics
o Keep hand tools as close as possible to the point of use and orient them in the
direction that they are used by operators.
o Use dedicated hand tools instead of tools that require bit changes, and combine
two or more tools wherever possible.
o Keep manual, operator-based work steps close together to allow flexible work
element distribution and value-added operator work.
o Segregate level 5 automation and continuous cycle operations from manual
operators or operator-based work flow as shown in the diagram below. We
prevent incident and increasing the safety of the worker.

Before implementing all of these suggestions, the layout is as follow:

As we can see there is still a problem, the space is too much; in fact, the tube bender
needs modification since the operator needs to move really far.
By implementing the suggestions:

We can remove a lot of inventories in a way that machines can get closer and worker
from assembly 1 has no obstacles, the beginning is close to the end, so the U shape is
closer. Dealing with batch-oriented equipment:
- Separate batch process from continuous flow with a supermarket or a FIFO lane
- Keep in the continuous flow if equipment is moving-conveyor type (and operator can
drop a single piece at start, and pick up one at end within TT);
- Ignore equipment batch capability and use it single piece;
- Transform equipment from batch to single;
- Replace batch equipment with one or more single piece inexpensive equipment.

How many operators should we have after the improvements?

We got to 88 seconds, so the total work content is 88 secs and takt time is 40.
The required operators are 2.2: 88/40 = 2.2 operators.
In this sense:

But how can 3 workers work with 4 stations? We could solve this problems in different
ways:
1) SPLIT THE WORK:
Solution: we have a line there are many stations in the line, there are in fact 17
tasks to be performed ONLY with three operators. We distributed the work
among them, the first operator does the work in on the first stations then he
does one of the last one. We can decouple the number of station by the number
of workers. The takt time can be covered by less number of workers.

2) THE CIRCUIT: Each operator is performing all the tasks and moving all around;
in this case we need multifunctional operators and it could be a bit less
efficient. We cannot put too many people because otherwise there would be
confusion.

3) REVERSE FLOW: Workers are going the other way round. The operator flow is
the straight line and the material flow is the dotted line. Material flow goes to
machine one to machine 2 but the worker goes the other way around, we need
higher flexibility of the workers and they are working on the same product, so
we can increase the quantity.
 4) COMBINATION: Worker 1 and 2 are going in circuit and worker 3 is working by
himself.

Material management
People outside the U helps the operators
inside the U to maximize the value adding
portion of their time through making things
easy available for them. The operator can
bring the components that he need and do
the assembly. The material handler brings
the containers full of components and put
them on the shelves. When I finish a box I
take it out and the material handler bring
them out of the cell. The components are
always in the same position. On the boxes is
described what is in, and the quantity. The
material handler goes with the box to the supermarket and shops the new
components, refill the boxes, goes there and put them on the right shelves. There are
different shelves (slides) in the station because the cell need different boxes.

How will the pacemaker react to changes in customer demand?

It is also very important that we are able to
create flexibility according to a possible
change of the demand. Whenever we
design cell, we need to be open to be
flexible to volume changes.
Level and pull 11/05

What is going to happen when we have a not stably demand in the market? How we
manage the variability in the demand?
We need to find out a way to smooth the impact that the variation of the demand in
the market are creating in our system. We need to do this because, in order to create
a continuous flow, we need to have as much as possible a stable demand and
therefore a flow which is not creating high variation.
We need to create the right level of demand in order to create the pull system.
We are going to discuss this by an example.

Value stream of the AS IS situation:

Future state:
Layout:

Story of the improvement done:

We can see that there
are some improvements
that are being worsening.
We should find the
reasons why.
We will look to potential
reasons outside the
factory. What is
happening outside? Is
there something
happening in the market that may be the reason of our worsening in performance?
Let’s analyse the demand.
If we analyse the overall demand, we see that there isn’t a big variation (+-10%), is
not a big issue.

But if we go to analyse
not the overall
demand, but the
demand of every
single item, we can
see that the variation
is much higher.
The demand variation
of a single item can
reach the 75%.

What we need to be aware of is that going upstream the variation is amplified; this
lead to higher lead time. This is called Forrester effect (or Bullwhip effect):

Why we have this kind of amplification?

- Lead time: delays increase as the lead time increases;
 - Batch size: suppose that we have 1/2 pieces that are asked; maybe one unit is
missing but since we produce in batch, even though only one unit is missing,
we produce an entire batch of products. This creates an amplification.
- Unpredictability (scrap, downtime, …)
- Wrong expectations, fears, … (psychological effects): if we expect that the
demand will increase, we will order much more and variation will be more
heightened.

How to smooth this amplification? We need to LEVEL and PULL.

How to do it? Let’s put a ‘filter’, a warehouse with inventories, to decouple what is
happening in the market downstream from what is happening upstream in our
system. This can avoid the propagation of the variation.

Is true that in lean philosophy warehouses must be avoided, to not crate waste, but
without this solution the wastes would be higher.
Where to locate the buffer? And how large the buffer should be?
The suggestion is to not implement something new everywhere from the first day,
always start with a small pilot: choose one or two products, focus on them, test it and
then deploy the result to all the other areas.
To mitigate the bullwhip effect, we should go thought these 9 steps. They are
guidelines to implement the LEVEL AND PULL approach.
Note: points 1, 2 and 3 will relate in general to the product A, B and C.
Points 4, 5 and 6 will relate to products A and B (make to stock).
Points 7, 8 and 9 will relate to product C (assembly to order).

Now we analyse these points.

1. What kind of production strategy we use (make to stock, assembly to

order ATO, MTO)?
To decide, we need to build the ABC curve of the demand from the market.
We look at the overall demand for each of the items that we have and then we built a
distribution curve by putting all the items in decreasing order according to the
quantity that each item is requested by the market.

We create a first class with the first n items such that the overall volume of that
makes 60% of our production volume. Then you create a second class, that should
sum up to another 20% of the overall demand. The remaining items will make the
calls C, which will be the remaining percent to reach 100%.
We need to decide the distribution of all the percentage, it isn’t fixed.
o Class A: runners, ordered every day, large overall quantities, small percentage
variations in volume.
o Class B: repeaters, ordered frequently (not every day), medium overall
quantities.
o Class C: strangers, ordered from time to time, small overall quantities, very
large percentage variation in volume.
We do this distinction to be aware that we don’t have a “unique” demand, we have
three different demands with different characteristics, and we should respond in
different ways to the different markets.
How to manage these classes of products? Are we going to out class A, B and C
products all in the inventory that we put to absorb variation? Are we going to put only
one of them in inventory?
We have 4 possible options.

1st option: hold all

the finished goods
in inventory of all
the three A, B and C
classes. Make all of
them to stock. This
solution is called
replenishment pull
system. How it
works?

All the products are

in inventories. Every
time we have consumption, we put the kanban in the Heiunka box (system already
discussed).
Why “replenishment”? We go to replenish the stock every time inventories are
consumed.
Why “pull”? Because the system is activated by the customer.

2nd option: hold no finished goods inventory and make all the product to order. This is
called sequential pull system:

We wait for the customer  the customer asks for a delivery  the request go to the
production control system that release a production from the beginning, and then
production starts. FIFO approach is applied to the step of production.
Why “pull”? Because the system is activated by the customer.
Why “sequential”? because we work with FIFO approach through the system.
So sequential pull and replenishment pull are the two extremes.

3rd option: hybrid option. Hold only C class product in inventory and make A and B
products to order daily, this is called mixed pull system.
For class C is used replenishment pull approach, for A and B sequential pull approach.
The pross of this solution is that we have less inventory, since only product C is in
inventories, that have lower volume than products A and B. The constraint is that
requires mixed production control and daily stability.
Why to put C in the inventories? Because C is more disturbance; we don’t know when
they are going to ask for this product. Once on a while we produce a batch of C, we
put in inventories and then we focus on A and B that are more profitable for us.

4th option: A and B produced to stock, with replenishment pull approach; C sequential
pull approach. Why? A and B have high volume, we always sell them, so we have
minor wastes. C is produced anytime there is the need to not create wastes.
In this solution you have a little more inventory, since A and B have higher volumes
than C.
The company of our case decided to go for this solution (4th option).
Actual system:

A and B products are in the stock (replenishment pull system). C product is made with
the sequential pull system.

2. If we use MTS make to stock, how large the stock should be, how much
inventories for each MTS product?
We create three types of stocks (for always the same item), to respond to three
different phenomena which are affecting the total level of inventories.

- Cycle stock: stock to be created to fulfil the daily demand, to keep the system
running. CT = average daily demand * lead time to replenish.
- Buffer stock: created to protect ourselves from the demand variation. Higher
the variation is, higher the buffer stock is. We calculate it as a percentage of
the variation.
- Safety stock: to protect from the variation of the lead time of the production.
The overall finished goods inventory is the sum of these three types of inventories.
Example:

From the graphical point of view:

3. How should we organize and control the finished goods store?

They have:
- dedicated locations per part number
- easy to reach locations for heavy items
- near begin of aisle for high volume demand
- stock control rule:
o buffer stock can be used only with production control department
authorisation, because is the production control that know if there are
customer demand variations;
o operations and Production control department review together the buffer
stock usage to identify fundamental changes;
o safety stock can be used only with plant manage authorisation.
Spreadsheet of the finished goods inventory:
The C item don’t have this table because is produced by MTO system.
Key to success:
- Always go to the Gemba
- Determine whether your production system is stable enough to work around
internal delays by buffering
- Segment your demand by A, B, C and define what type of pull system to adopt
- Carefully calculate the amount of Cycle, Buffer and Safety stock to satisfy
customers and minimise expedite, overtime etc.
- Practice good workplace organisation and visual control in the finished goods
store and shipping area
- Strive to make the status of finished goods / system clear: Normal Vs Abnormal

4. Where will you schedule the value stream?

In replenishment pull, the pacemaker will be close as much as possible to the market.
In sequential pull, the pacemaker often is the first process at the beginning of the
value stream. It can be located further downstream, just before inventory types
proliferate (product design gives a fundamental contribution). In the first position
before the first FIFO.

5. How will you level production at the pacemaker?

In this table we have the overall demand that we have every day for the items A, B
and C.
If we decide to produce based on schedule, the result is that we are producing in
batch.
In the morning we produce 500 units of the first item A. In the second shift we
produce other 500 units of the second item. At the end the overall daily production is
1000, which is equal to the demand; but we don’t need all the pieces produced, so we
will put in inventories the overproduction and then maybe we will use it tomorrow. We
satisfy the demand of the item that we not produce daily with the inventories of the
previous days.
We don’t like this type of system because this batch production, also because it create
bullwhip effect:

How to manage this? We need to understand that setups are losses.

Someone can say ‘since the setup is a loss, let’s minimize the number of setups,
increasing the batch size’. But the drawback of this is that we optimize our production,
but increasing the batch size, we create losses in other departments, upstream,
propagating the bullwhip effect. We optimize the packing for instance, but this creates
issues upstream, don’t optimize the overall system.
What we need to do is try to decrease the production, the change over time to
minimize the fluctuation.
Reduce the lead time: if we reduce the lead time, we reduce inventories. We can
reduce the lead time reducing the setup.
Batch sizes should be minimised as much as possible according to the lean approach.
This does not create upstream the bullwhip effect.
Batch sizes should be minimised subject to the following constraints:
- Work content differences between products (Best if all items are below TT)
- Changeover requirements between part numbers (Effective cycle time)
- Production pitch interval (Typically TT * Pack size, but consider also material
handling effort and need of control)

Example: takt time 54 sec.

Pack quantity to deliver = 10
pieces per pack.
We have to produce a pack
every 540 seconds (pitch
interval).
If in a shift we have an
available time equal to 450
minutes, how many pitches we
can do? 50 pitch intervals. We
can divide our shift in 50
intervals; in each interval we
produce one pack of 10 pieces.
Which items we produce in these 50 intervals? Let’s go back to the A B C curve.
According to the percentage of product of class A, B and C, we assign a certain
percentage of the interval to class A, a certain percentage to class B and a certain
percentage to class C.
How many pieces we produce for each item:

Result:
Every day we produce all the 5 items of A for instance, in small batches.

6. How will you convey demand to the pacemaker to create pull?

What kind of kanban we need? In what sequence?
There are different types of kanban.

The withdrawal kanban is used to pick the product from the supermarket.
The production kanban is used to activate the production upstream.
For the class A and B in our case we use the in process kanban.
Example of in-process kanban, how it looks:
We put the item number, the description, the quantity. On the left we put the
information where to stock the product once the product has been produced. On the
right there are the information of where to produce (in which cell).

At this point, it remains the problem about how to schedule the production.
We will schedule the production as we have seen thought the Heijunka box: in
correspondence of each shift, the vertical columns represent equal production interval
(in the example 9 minutes interval).
Every 9 minutes we have a kanban.

Conveyance route for information and materials:

Every 9 minutes this path of the pacemaker is going to repeat. The pacemaker go
thought these phases:

The total time of the tour is 6 min and 10 sec, so it’s ok (< 9 minutes).
But it could be that this is too fast, is too risky. 9 minutes may represent a too narrow
interval (there would be problems in each control, with the impossibility to recover
from any inconvenience). For this reason, we can decide to lengthen the pitch interval
to 18 minutes.
Instead of picking 1 empty box per each travel, we can pick 2 containers for each
travel, two kanban. The pacemaker passes every 18 minutes, bringing 18 minutes of
work.
However, this solution lets a cycle of 6 minutes every 18 minutes, which represents
an inefficiency, since the water spider is inactive for a relevant period of time. In order
to avoid this inconvenience, we may think to let him work in more than one cell,
aggregating the work and distributing the scheduling and the movements among
more flows.

Every 18 minutes, the pacemaker pick up 4 kanban, 2 for the first cell, 2 for the
second. Now the tour of the pacemaker will last longer than 6 minutes, since there are
more areas to be visited.
 7. How will you manage information and materials flow upstream from the
pacemaker?
So far we were concentrating on the pacemaker, now we need to go upstream.
A system based on the use of kanban should be supported with proper mechanisms
for the management of information and materials between the upstream processes
and the downstream ones, in order to avoid the possibility of incurring in
overproduction and in order to synchronize the different departments.
In our system, the big problem lies in the fact of having adopted a "hybrid" system, in
part managed in push
logic and in part
managed in pull logic; as
a result, there is a
strong lack of
coordination between
departments, which
means that, often, the
cell remains inactive due
to the absence of some
components that feed its
production.

In fact, if we look at the upstream stages, we can observe how they are characterized
by bigger batches, given the greater level of automation. However, bigger batches
lead to a higher EPE and to a lower response time.
Below are reported the possible causes of delays in assembly.

Upstream process review:

In order to realize the standard components, required in high volume, it is necessary

to create a production system supported by supermarkets.
A formal market mechanism is needed for upstream processes with different operating
patterns.
In building it, usually it is possible to recognize some principles:
- Position the supermarket close to who/what produces: since stocks represent an
overproduction, it is appropriate to position them near the production
department, in order to push this phase in improving, reducing costs and
recalling its inefficiencies (traditional Toyota);
- Shape them in order to make the components needed instantaneously available
to the subsequent stage and feed its production (centrally located market);
- In the case of many internal warehouses, it is obvious to maintain all the
typologies of components in a unique supermarket; then the water spider will
be designated to distribute them to the different departments;
- For all the products that come from the outside (purchased parts), it is required
to reserve a smaller supermarket in proximity of the receiving area/docks

In any case, we need to keep in mind to:

- Make clear that producing process has ownership of inventories;
- Ensure that downstream processes pull what they need, when they need, on a
regular basis;
- Ensure there are signal exchanged between the processes regarding what has
been taken away.

After all these considerations, the updated layout with market area results in the
following in our case:
 8. How will you size the markets and trigger withdrawal kanban?

Three tasks to successfully trigger pull from the assembly cells to the central market:
1. Set a standard amount of inventory of each part to hold at the assembly cells,
based on the nature and frequency of the conveyance route
2. Create a separate withdrawal kanban for each container stored in the cells
3. Determine the right amount of inventory to hold in the central market
Workflow for withdrawal material-handling loop, from the central market to the cell:

Note: here we look at the tour of the material handler from the central market area to
the assembly cell, to take the component to the cell. We are not referring to the tour
of the pacemaker, which is running at the same time, but in a different part (between
the assembly cell and the shipping)
SIZE OF THE MARKET 15/05
How to calculate the stocks at the level of inventories that we need in the central
market to guarantee the delivery of the demand?
These inventories are decoupling the assembly (the “customer” of the central market)
from the upstream machines, like the molding and the painting.
Also in the central market we should use the cycle stock, the buffer stock and the
safety stock.

The formula are the same used for the finished product.
The buffer stock here is much lower, because production downstream has been
levelled.

After the implementation of the changes, this are the improvements:

The 9% gap should be analyzed; most probably is related to the molded parts which
are coming from upstream, or from the C item (that are unpredictable in the
volumes).
How to manage the class C item?
At the beginning we decided to produce them on order (MTO). Now we can decide to
implement a different strategy: assembly to order. We create the molds, we paint
them and we assemble them when we get the order.
 9. How will you control batch processes upstream from the market?
Actual situation:
o At Apogee, lot size is presently defined on the base of EOQ
o Operators run longer than planned lots
o Operators and plant manager are measured on machine utilisation
The system based on the EOQ (economic order quantity) sometimes could lead to too
large batch sizes, that may create bullwhip effect.
there is a different approach to manage the class C product in a better way?
There is an alternative, based on the kanban approach. We will use a triangular
kanban.

Assume that we have three types of products in the inventories (454, 455, 456). How
do we replenish the inventories in the central market asking to the upstream process
to produce them?
We use a kanban, called signal kanban; we define a trigger point (the minimum level
of inventories that we have in the central market, below which we trigger upstream
the production of a specific quantity of parts). We have to define the trigger point,
and, whenever we trigger a production, we have to define how much we have to
produce.
If we are consuming 454 item and we go below the trigger point, we take the triangle
kanban and we put it in the signal kanban, in a queue of kanban of parts that are
waiting to be produced. So we use the
FIFO approach.
Why don’t we put the kanban of 454 in the
first place of the queue, with the other
454 item kanban, to save a setup of the
machine? This is wrong, we will se why.
In the triangle kanban are represented
these information:
- The name of the part (the code)
- Where we have to deliver in the
central market
- How many we have to produce
 - The trigger point. Whenever we reach less than 200 in the inventories, we
produce 400.
How to calculate the trigger point? How to calculate how much pieces to produce?
We start form this data:

**total run time to be able to produce the required demand.

We will follow this procedure:

 Step 1: determine time available for changeover work

The total time required to produce the demand is 698 minutes per day. Then we have
sometimes which is not used from the machine to produce, that can be used for doing
setup; if we do more setup, we reduce the batch size, therefore we reduce the
inventories  no wastes.
Let’s see how much time we have available for the set ups:

 Step 2: set the number of changeovers per day

With an estimate of the time available for nonproduction work, the team could
establish a target for the number of changeovers per day.
How will we use these setups? We have 3 products  1 setup per product.

 Step 3: determine production size

There are two basic ways in lean manufacturing to set lot sizes with triangle kanban:
- Option 1: one is known as the fixed-time variable-quantity method. It calls for
producing a constant (fixed) interval of demand for each part number run. For
example, one shift’s worth of customer demand might be run for each part
number. Because demand per unit of time (e.g., per shift) differs by part
number, this means that the process will make a different (variable) quantity of
parts for each part number run. In thinking about this, be careful not to confuse
machine run time with the demand interval (for instance, demand interval
equal to 1 day, machine interval could be different)
- Option 2: the other option calls for running the same (fixed) quantity of parts
for each part number run, producing parts for varying intervals of demand for
each part number. For example, a process might run every part number for
three hours before changing over to the next part number. This would create
lots of perhaps a day's worth of consumption for some parts and two days' for
others. (Note that with this method production is adjusted to demand by
running some part numbers more frequently than others.)
The first option builds lot sizes more closely linked to customer demand and leads to
lower inventories. The second option may be easier to manage because the pattern of
changeovers is steady and predictable. Because Apogee felt its molding management
was equal to the challenge and because it wished to supply the central market for
molded parts in a highly responsive manner, it chose to implement the fixed-time
variable-quantity method for the triangle kanban lot-size calculation. A lot size of one
shift's worth of demand for each part number was selected, as shown in the diagram.
  Step 4: specify a trigger point for reorder

Now we calculate the trigger point for each of the three products.
In order to calculate it, we should start form the takt time.
900 minutes is the available time that we have in one day; dividing it by the daily
demand of each product, we obtain the takt time of each product.
 Kanban are picked from bottom
part to upper one. If we exceed
the line, we start producing,
because it is the indication that
the specific item has enough
kanban to batch and to
produce enough quantity to
reduce changeover. Doing too
many setup, we will not be able
to respect demand. The
advantage is that we see how
demand is arriving, while in the other case we do not have information (the inventory
in the market is not visible at all times since there is no batch board with cards).
In addition, in this case we can decide to produce even if we have a small batch.

Once we have set up a pull system that meets all the prerequisites outlined, logic
dictates that the same organization is adopted also by other product families.
This, however, creates a problem: if there are more than one pull system that coexist
together, the "water spider" will be obligated to give a rhythm to the pacemaker cells
that are characterized by different takt time and pitch interval; that is why to manage
a situation of this type, we can think to be organized as follows (although it is not so
levelled):
Manage variability 17/05

Lean is mostly known for the wastes hunting

(Muda), but the concepts of Mura
(Variability) and Muri (overload) are equally
important.
Let’s clarify the 3 categories considering the
transport of 6.000 Kgs of material by a
pickup truck that can move 3.000 Kgs: how
is it possible?
• 3 trips of 2.000 Kgs  MUDA
• 1 trip of 6.000 Kgs  MURI
• 1 trip of 4.000 Kg + 1 trip of 2.000 kg 
MURA

MURI problem is clear Less clear is MURA: what are the problems?
- Inefficiencies
- Queues/stocks

A system which is not stable is hard to be

controlled and improved. In order to
manage variability, we need to do a trade-
off between three dimensions: demand
variability, WIP level and resource
saturation.

We have a given demand variability, in

order to satisfy the demand of the market,
we have to make a trade of, find the right balance between resource saturation and
the amount of stock (WIP level). In order to guarantee the capability to respond to
the demand, we have to define a specific level of stock and a given level of resource
saturation.

The production cost is too high, and we have the

target to decrease the cots of the resources. How
can we manage this? If we reduce our capacity, we
have to make a trade off: if we increase saturation,
in order to defend ourselves from variability, we
have to increase the stocks. It is the same concept
of the queuing theory, in which when we increase
saturation, we have more queue.

We can decide to decrease the saturation of the

resources in order to decrease the stocks. If we
want to move to a lower level of WIP, we have to
reduce the resources .
If we reduce variability (process and demand
variability) we are able to change the conditions of
the system and we can reduce the stocks and
saturation. We can work on the line or modify the
line, making it less variable.

Lean manufacturing is also focused on improvement of the situation (kaizen principle).

The first thing to do when we want to improve something, is to define the target.
How?

A first way is to define the mean performance of all the operators and try to define the
improvement target according to the percentage increase of the average performance.

We define the average performance value, that in this case is more or less 15 calls per
hour, ad then we fix an improvement target in percentage.

According to a managerial approach, this method can create frustration in the

operators, because maybe the management doesn’t provide means to reach the
target objectives, this can demotivate employees.

On the other side, setting an average target can create frustration in the less
performing operators, because they can see the target too much high, and on the
other hand the overperforming operator would tend to reduce their performance.
So, we should define a target according to each specific situation. We have to ensure
that the target can be reasonably achievable for each operator.

This process of fixing target is a process affected by some degree of variability. There
are many causes of variability:

- Special causes, not common, symptom of something that is not expected tom
be and might be a responsibility of the operator.
- Common causes, like random evens that happen during the day, this are
matter of managers, because managers have the overall view of what’s
happening in the system. If we let operators to respond to this causes of
variability, this will lead to an increase of the variability instead, because each
operator will find its own solution. So, managers have to understand how to
deal with variability.

There are many tools to manage variability, that referrer mainly to statistical control.

To see if a process is in control, we have to set an upper and a lower limit of

performance; if the values of performances measured lies the belt defined by upper
and lower level, the system is under control.

If you take samples from a population:

• The sample mean is a proper estimator of the population mean
• The distribution of the averages of the samples tends to a normal distribution • The
variance of the averages of the samples is equal to the population variance divided by
the sample size
• The variance of the sample can be used to obtain more information

Control charts:
If there are points outside the limits
the system is not in control. It is not
stable. If an item is outside the
limits, there are less than 0.15%
chance that it is a random effect ->
worth investigating.

Although the system is not in

control, you can identify the limits
and they show the points out of
control.

We can set different bands; we can

have a narrowing of the bands
between lower and upper limits, for
instance by increasing the number of
the sample selected. We can also
decrease the bands by reducing the
confidence level that we want to
achieve.
If all the points fall in the narrower band, it means that the process is super in control
and is affected by a very low variability.

Once we have defined the target and we measured it, who is responsible of the action
for the improvement? If the system is in control, the responsibility is of the managers,
the person who is in charge of the quality of the output.

If the process is in control, it doesn’t mean that we cannot act in order to improve that.
The important part is to improve on the methods that will result in a decrease in
variability. We can focus on the improvement of the process, rather than the final
output.

Usually, 95% of the errors depend on the system, only 5% is caused by the
operators. 95% of managers’ attention is towards operators (and their management),
and only 5% to the management of the system. The implementation of a good
process is far better than the selection of good people.

Focusing on method, it’s also a way to avoid stress and over boarding of people.
Measurement of performance are a good way to evaluate people, but operations
management should not expect a change in the performance of the operators percent
but should expect changing performances of the operators after an improvement of
the method on the process itself.

If managers tend to ask for an increase in the performances to operators of lower

level, without an suggestion on how to improve the method, this will result in a faking
of the result.

When we want to manage variability, we can either operate on demand side or

operational side.
If we want to deal with the demand, first we have to understand what our customer is
expecting to have and what is our customer.
We know that, in base of the frequency of the demand, it can be classified in three
different ways: runners, repeaters, strangers.
How to manage the variation?
In a service systems variations have to be managed at the front line, that means
company should manage variation from the beginning, not postpone the management
of variation to upper levels, to avoid the bullwhip effect. The earlier we manage it, the
better it is.

The higher the variability, the higher the competences and autonomy that operators
are expected to have.
The more autonomy we give to operator, the lower is the level of command and
control that we need in the system.