-...

' I
,,. l- J
r

[,

n
·r-,
L_J Creating
n Level Pull
A lean production-system improvement guide for
.n production-control,operations, and engineering professionals

n
, J by Art Smalley

Foreword by Jim Womack, Dan Jones, John Shook, and Jose Ferro

nL ....J

. [] A Lean Toolkit Method and Workbook

THE LEAN ENTERPRISEINSTITUTE
Brookline, MA, USA
n
l_ .J
www.lean.org

Version 1.0
ii April 2004
I I
l..,._ ..J

n
k .... ...J

·n:
L_ ... --'
I

;---,
i l
' I
L .J
D
0
D

r · ·1
LJ ACKNOWLEDGMENTS
r··-·1
I~· _,I In preparing this workbook, I've realized my debt the many fine supervisors and
to
sensei who guided my education during my early career with Toyota Motor Corp. in
r·····--i Japan. Without their patient instruction and generous sharing of their extraordinary
LJ knowledge, I would not be able to pass along these lean concepts.

Additionally, I'd like to thank my good friends and mentors Tom Harada, Russ Scaffede,
D and John Shook for creating learning opportunities for me over the past 15 years.
Special thanks are in order to my lean-expert friends and colleagues Elisa Martinez
[J and Raoul Dubeauclard. They volunteered significant personal time and energy to
review early versions of the workbook and supplied critical feedback as I crafted the
r ·1 final produce.
L.J
Most importantly, I'd like to thank my wife Miwa for her loving support and cooperation
while I wrestled with this project over many months. I promise to clean up the mess
on the dining room table now that the workbook is finished.

D
LJ
D
LJ
LJ
D
r--· ._.,,
1 !
L..-!
 FOREWORD

When we launched Learning to See in the summer of 1998 as the first publication of the
Lean Enterprise Institute (LEI), we urged readers to start down a path toward perfect
operational processes by mapping the value stream for each product family within the
walls of their facilities. Our objective was to raise the consciousness of many managers
from point improvements at the process level-creating cells, reducing set-up times,
implementing SS, improving the process capability of individual steps-to improvements in
the performance of the entire value stream. We called this progressing from process haizen
to flow kaizen.

In the years since the launch of Learning to See, we've introduced additional workbooks
describing how to introduce truly continuous flow in cellular production activities ( Creating
Continuous Flow) and how to implement a lean materials-handling system that supports r---,,
!
L____j
I
continuous flow (Making Materials Flow). We've also extended the mapping process for
product families far beyond the walls of individual facilities to encompass entire value
screams (Seeing the Whole). "·~J
I I

Now we are ready to move beyond the value stream for individual product families and
take on production control for all of the product families within a facility. We call this the leap
to system haizen because it ties together the flow of all products through a facility by means
of a lean production-control system. To do this, many facilities will need to convert from 0
traditional Material Requirements Planning (MRP) systems that schedule each activity within
a facility and push product ahead to the next activity. Others will need to move beyond simple D
pencil-and-paper schedules or homegrown pull systems that do not effectively control or level
production. In either case, the critical need is to transition to a rigorous pull system where
each production activity requests precisely the materials it needs from the previous activity
D
and where demand from the customer is leveled at a pacemaker process to smooth production
activities throughout the plant.
D
To help you make this leap, we have asked Art Smalley to share his years of lean implementation n
,_J

experience. Art was one of the first foreign nationals to be made a permanent employee of
the Toyota Motor Corp. in Japan where he worked at Kamigo Engine Plant, Toyota's largest
machining operation. In 1994, Art left Toyota to become director of lean production at
D
Donnelly Corp., an American automotive supplier with more than a dozen planes worldwide.
In 1999, Art moved to McKinsey & Co. where he was a subject-matter expert on lean
D
manufacturing and manager of McKinsey's Production System Design Center. In the course
D
,...-,
L__j
------ .,
of his duties over the past 20 years, he has advised hundreds of facilities worldwide across a
diverse set of industries on how to take a lean leap. In mid-2003, Art left McKinsey to spend
more time at home with family, to write educational material on lean manufacturing, and to
work directly with firms attempting a lean transformation.

We have warned in each of our workbooks that the step we are describing is harder than
the seeps required in previous workbooks, and we must offer this caution again. A truly lean
production-control system that rigorously controls production at every step and levels demand
from the customer has proved a great challenge for most firms. As a result, we usually see that
this is the last element attempted in a lean transformation. If this is true in your case, you are
in luck. In Creating Level Pull, Art has provided all of the basic knowledge you will need
to get started in creating a lean production system in your facilities. And he has carefully
constructed the workbook to be easily used by firms already far along with process kaizen
and flow kaizen.

On the other band, if you arc just starting your lean transformation you also arc in luck.
Veteran lean practitioners usually urge firms with sufficient process stability to start their lean
transformation by introducing lean production control with leveled demand as a system kaizen
before moving to flow kaizen and process kaizen. If you are in chis situation, we hope you will
summon the courage to take the leap. The benefits for your business will be enormous and all
of the knowledge you need is summarized here.

Given the nature of your challenge-wherever you are starting-we are anxious to hear
about your successes as well as your difficulties and to connect you with the Lean
Community at www.lean.org. Please send your comments to [email protected].

Jim Womack, Dan Jones, John Shook, and Jose Ferro

Brookline, MA, USA; Ross-on-Wye, Hereford, UK; Ann Arbor, MI, USA;
Sao Paulo, SP, Brazil.

www.lean.org
r
L.J
··1

D
r·1 CONTENTS

LJ Foreword

LJ Introduction

LJ Part 1: Getting Started

,;-····-··1
It__: I
Part 2: Matching Production-System Capability
to Demand

r ·---,
l~ i Part 3: Creating the Pacemaker
r ., . -,
IL____JI Part 4: Controlling Production Upstream
r·-· ··1

I..___...: Part 5: Expandingthe System

Part 6: Sustaining and Improving

LJ Conclusion

lJ About the Author

LJ
Appendix
[J
r 1
References
L

CJ
 D
INTRODUCTION
[J Continuous flow of materials and products in any production operation is a wonderful thing,
and lean thinkers strive to create this condition wherever possible. The reality of manufacturing
[J today and for many years to come, however, is that disconnected processes upstream will feed
activities downstream. Additionally, many internal processes are currently batch-oriented and
function as shared resources. The major challenge in this situation is for downstream processes
to obtain precisely what they need when they need it, while making upstream activities as
efficient as possible. This is where leveled demand and pull production are critical.

As I visit manufacturing operations around the world, I rarely see anything resembling level
[J pull production. Instead, I observe progress in introducing continuous flow as well as local
stability improvements at the individual process level by means of point kaizen (such as SS,
[J enhanced process capability, and set-up time reductions).

The reason for this is not mysterious: creating level pull production in an operation of any
[J complexity is not easy. Even within Toyota it cook 20 years of hard work and experiments,
between 1953 and 1973, to establish the system companywide. A successful transformation
r: -·1
I
l.__J
I requires the coordinated efforts of everyone in a facility looking at the needs of all the
product-family value streams. This calls for system kaizen of material and information flow
to support every value stream.

Fortunately, the basic methods needed for level pull production are well understood, having
been developed by Toyota and its affiliated companies over many years. In addition, there is
now a considerable base of experience in introducing these methods in firms outside of Toyota.
,....... --,
I : The challenge therefore is to provide a simple recipe for introducing these concepts in your
LJ
facilities. Based on my experience in converting facilities from push to level pull-based
r---·-·-:
I
. I
I production, I've developed 12 questions you will need to answer to meet the challenge. Not
L_J
every question will apply in every case, and you may need to deviate slightly as your
situation dictates, but I'm confident that by addressing these questions every facility can
improve its performance while moving operations to the next level of sustainable
achievement.

Only you can supply the courage and leadership to create level pull in your facilities. And
much additional point, flow, and system kaizen will be needed after your initial leap. But
this workbook provides all of the necessary knowledge to get started and to move you past
the critical threshold from erratic push to level pull. I'll be anxious co hear about your
LJ experiences, and I wish you smooth sailing on level seas!
-~,.-----·--,
./ J
L_J
Art Smalley
Huntington Beach, CA
April 2004
 LJ
D

[J Welcome to Apogee Mirror

Apogee Mirror is a typical discrete parts manufacturer, making exterior mirrors, interior
LJ mirrors, and door handles for the automotive industry. Several years ago, Apogee responded
to pressure from its customers for lower prices, higher quality, more frequent deliveries
LJ exactly on time, and more rapid response to changing market demand by taking a hard look
at its manufacturing operations.
r . .. ,
LJ Apogee managers took a value-stream walk to follow the manufacturing paths of its three
main product families. They drew value-stream maps for each product family-one of which
D is illustrated below by the map for the exterior-mirror product family. They soon were able
to see wastes of many sorts: long set-up times on molding machines; poor uptime in the
lJ paint booth; many disconnected operations for assembling the product; and long throughput
times with large inventories between every seep in the process.
[""l
Original-State Value-Stream Map for Exterior Mirrors

D Productl011Control

r ··1 MRP

LJ
.LJ
Cfl•40sea. CIT· 120mlrt. C/T• 100sec.. Cff·-40- Cff.(J()-,. C/T:120,ec.

D c10.1oom1n.
Uptlm4 .. 70%
!5%saap
Cl0·60-
Updmo•6SZ
25lr"""'*
CI0•20mln.
Updm••651
tO:rewri
c;o.1om1n.
Updmo•601
5% rework
CIO·IO-
Updmo•70I
51"'""'*
CIO• IOmln.
Updm•·75l
51""""*
2,J,11.. t:,hJft; 2,JrJfU 2sl,lfU 2,1,{fcs 2shffr.5

[J EPE• rweek EPEa 1week 20,,U:Jf,Q(S loperaw I operator ilop<f•toro

Produ,tlon _ :3@d
15day,; 5days feae/tlme - P)'6

r ·1 Proc.ss/n9
LJ time = 12@mlrr.

r ·-. 1
LJ
[j
Part 1: Getting Started 1
 D
D
By energetically pursuing both point and flow kaizen along the three value streams, Apogee
management and employees soon were able to achieve much better performance for all three
product families, as shown by their current-state map for the exterior-mirror product family.
r__J
LJ
Current-State Value-Stream Map for Exterior Mirrors
1:
LJ
Production Control

r--
L_ _j

:--i
L __J
CIT=40scc. CIT"=120m!n err= wo eec.
C!0=60min. Cl0=20min. CIO=Omin.
Uptlmo=95%

1ZGcra
2511ift5
Uptim<=90Z
15%rewcr.l::

!shift
Uptlme=92%

2% rework

2shlfts
D
EP£= !week EPE = t wee« tioperstors .

15dily6 5day,; 5dayo 5day,;

Production
lead time
= 30days D
40sec. I 120mln. r I wo ..c. Processing f25mln.
time = 40sec.

Apogee reduced changeover times in all of the processes, improved uptime in paint through
D
point kaizen, and created compact continuous-flow assembly cells through flow kaizen. Because
of this, Apogee managers were able to shrink throughput time and inventories while reducing LJ
effort and cost. They also were able to reduce the amount of manufacturing space required (see
Apogee Overhead Layout). D
0
D
D
'II
'~J
t

2
,- ..··,
! '
LJ
Apogee Overhead Layout

Legend

[J ~ ~ ~ ~
BBBBB ~ ~ Receiving
Dl:lO A!lsemblyCoB
DD
~ ~ ~ ~ EJ ~
DDDDDDDDDDDDDDDDDDDDDDDDDDDDD
BBBBB Dock
~

[§
lnjeclion Molding

DleC.est

~ Machiningcenter

D lnventorv

ODDO ODDO ODDO DODD

~~ ODDO DODD ODDO DODO
D ~~
ODDO ODDO ODDO DODO
...
DODO ······-··· ..
..
DODD ODDO ODDO DODO
ODDO .. ..
ODDO
DODD ~: Open
New 1 DODD DODD DODD ,···--····-·' ..
DODD ..
DODD t Space f:
.. Cell #3 Cell #2
g D g g D g g O g :: Opon
Cell#1
NCM' : Pe int

DODD
DODO
..
. ..
.
oo DD DD OD DD OD : Space ::
DD DD Do DD OD DD .. ..
DODD ·--·----·· 0 0 D t ...

DODD
D DODD
DODD DODD ODDO DODD i
Cell #4
...
DD CallD#6 DD DD Cell0#S D0 DD D DD i NewOpon
:
D oo oo DO OD oo OD
DD Do DD DD Do Do i Space
i
0 0 0 : ··-·-·····--······-···.:
r~
L__J
:~
ODDDDDDD DODDDDDD 000000 Finished-Goods
DDDDDDDDDDDDDDDDDDDDDD Storago

Cl
Like many companies today, Apogee avoided taking any action to more tightly link and
control the flow of information between the production departments-molding, die
casting, paint, assembly, and shipping. Apogee managers judged that modifying the
information management system linking these areas-which pushes products ahead to the
next processing step with the help of material handlers who respond as needed-would
D be complicated because the necessary systems would affect every value stream in the
plant. In addition, many managers wondered if this leap really was necessary. They
D thought that sufficient improvements could be wrung from point kaizen and flow kaizen.

D The ContinuingChallenges of Delivery and Cost

Initially, Apogee's managers were delighted with their achievements as a result of point and
[J flow kaizen. Morale in the facility was higher as a SS program brightened work areas and
employees participated in the kaizen activities. And direct-labor costs significantly declined.

Many dimensions of performance, however, did not improve as hoped. In particular, the
facility still needed considerable overtime and expediting of shipments to meet customer

Part 1: Getting Started 3

 D

requirements. And while total inventories had been lowered, they still were high. Equally
troubling, the reduction in direct-labor costs had not been matched by any change in indirect LJ
labor. Managers still were spending large amounts of time revising production schedules as
customer requirements changed. Meanwhile, an army of material handlers raced through the []
plant to get the right materials to the right place to meet changing customer requirements.

Even more troubling, performance in some areas seemed to be deteriorating as the initial D
excitement of the kaizen initiative wore off. In particular, the paint, assembly, and shipping
departments often reported that they could not provide what their customers wanted because D
of a lack of materials in the right place at the right time. This trend is shown in the box score.
ii
l__J
Box Score-Exterior-MirrorValue Stream*

Original After basic After cell Current

D
state stability flow kaizen state
Productivity
v <,
n
c. J
Direct labor (pieces per person per hr.) 9.0 10.0 11.5 I 11.0
) i'I
Material handlers supporting value stream 3 3 3 \ ..... 4 ./ L__J
!

Quality
Scrap 5% 4% 3% 2%
Rework** 25% 20% 15% 15%
External (ppm) 500 250 125 105
Downtime*** ~
D
Assembly (min. per shift) 40min. 30min. 10min. ( 20mln. )
Paint (min. per shift) 30min. 20min. 15min. - tcrtrun. --- []
Molding (min. per shift) 50min. 25min. 25min. 10min.
Inventoryturns
Total 8 11 14 (
~
12
--- )
~
On-timedelivery
To assembly 65% 68% 80% I
,.....- 75% ~ D
)
To shipping 80% 92% 95% \ 1--..... 85%
To customer 100% 100% 100% 100%
./
D
Door-to-doorlead time
Processing time (min.) 126.0 126.0 125.7 ~7 i I
Production time (days) 36 34 28
'
[7 30 <,
Costs ( D
Overtime costs per week $6,000 $5,000 $4,000 \ $5,000
r-, $2,000 _/
'
Expedite costs per week $2,000 $1,500 $1,500
- i I
* No major change in demand or product mix over this time.
** Rework is due to persistent inclusion problems.
*** Downtime is separate from changeover time and reflects only lost time in production due to mechanical problems or
material availability per shift.
D
4
 The pattern of visible improvements at many points but limited progress in the facility as
r--·1 a whole, along with ominous backsliding in some improved areas, seemed to suggest that
LJ
something was wrong with the total production system, not just the individual parts. Apogee
managers therefore decided to take another walk to focus on the flow of information and
D materials between production areas and to look at the entire production system involving
all three product families. What they saw was quite startling.

TraditionalScheduling in a 'Lean' Facility

LJ The management team started its walk in the shipping area, following the value stream

c for exterior mirrors. It quickly learned from the Production Control manager that customer
schedules were forecast well in advance and formed the basis for the weekly schedules sent
to each production area by the computerized Material Requirements Planning (MRP) system.
[J However, the weekly schedules bore only limited resemblance to the daily releases from
customers that determined what was actually shipped. Because the throughput time in the
plant from raw materials to finished goods was still several weeks, the frequent change in
customer orders, as reflected in the daily releases, often meant:
D
,-,
L.....J
•
•
The wrong items-too many and too early-were being produced far upstream.
Downstream processes, such as assembly, lacked the correct parts despite holding large
inventories of many parts.
• Downstream processes had no effective mechanism to let upstream processes know
r--1 what parts they needed next, short of supervisor intervention.

To deal with these problems, Production Control spent most of its time revising schedules
D and expediting parts within the plant. Yet during a normal shift only 75% of orders were
ready to assemble on time, and only 85% were ready to ship on time. Because no automotive
D supplier can risk stopping its customer's assembly plant, Apogee dealt with the problem of
products arriving late at the shipping dock by running large amounts of overtime every day
r -1
(to get the product out of the plant that night) and by using expensive air freight. Senior
LJ
managers also discovered on their walk that production capacity for each process was greater

LJ than the average demand. This meant that expensive overtime mostly was caused by scheduling
problems rather than capacity constraints.

To illustrate the production-control problem Apogee was facing, the management team drew
a simple graph (see Apogee Demand Variation for Exterior Mirrors on page 6) in which the
,- . ---1
variation in orders for exterior mirrors was plotted. The solid line shows the actual variation in
LJ
weekly demand for units from the end customer over the most recent 13-week period for one
·1
r'"
I ' of the two assembly cells in this value stream. (The two cells were identical. One produced
LJ
right-side mirrors and the other produced left-side mirrors.)
r :
LJ
,----,
LJ Part 1: Getting Started 5
 n
L.. ...• J

Apogee Demand Variation for Exterior Mirrors

500
~--.::;-~-- - --- 550 units
:t10%
450 units

Average 400
customer
300
demand
per shift
200 D
100
D
2 3 4 5 6 7 8 9 10 11 12 13 Weeks

L_J
The chart reflects only variation in the total number of mirrors demanded from the left-side D
exterior-mirror cell. The situation became more interesting when mix variation for mirrors
by colors and configurations was included for the same 13-week interval. A sampling of the n
L.J
top 10 part numbers that ran through the same assembly cell in the exterior-mirror value
stream generated the chart below (see Apogee Demand and Mix Variation). The obvious
conclusion was that total demand varied only slightly, but mix varied substantially.
D
Apogee Demand and Mix Variation
D
% Variation Maximum
demand
D Minimum demand D
+100 +100%
D Maximum demand
D
+50 -
D
0 - - - - - - .- - - - - Average demand
for part number :l
L_j
Minimum
-50
demand
-75% D
-100 r i
[J
Top 10 finished-goods part numbers

.---
I I
LJ

n
~---1

n
L._ __ J
6
D
More surprising, however, was the realization that variation in both total demand and mix
D got progressively worse through the plant. By collecting data on actual production orders at
each process step, the Apogee team soon was able to see that the variation in daily release
D amounts was less than the variation in actual orders sent to the two assembly cells for exterior
mirrors, and this variation was less than that experienced by the most upstream production
step for this product family (molding). In short, Apogee faced a modest challenge from
D erratic customer demand-one no worse than what most facilities face-but its internal
f ···1 practices made the problem much worse than it needed to be.
LJ
Demand Transmission and Amplification
D Demand

D Fina!
Assembly --~
~

D 100%
f month
50%

2weeks
25%

1week
10%

2·3days
Schedule amplification

Inventory

D
The Need to Switch from Erratic Push to Level Pull
D As the Apogee managers reflected on their walk, they suddenly understood some simple
ideas they had read about but never really grasped. They had discovered that their customers
were only modestly erratic with their orders, but Apogee's internal scheduling practices were
making the situation much worse by transmitting customer variation to every step in the
production process and then increasing the variation. They also realized that the centralized
production-control methods, which attempt to schedule every point in the production
r:
L_J process, were pushing products ahead to the next production area on the basis of the
forecast rather than the actual needs of the next process. This was causing inventories to
r···1 pile up ahead of every step.
LJ
r-·-r The team realized Apogee had a cognitive scheduling system that pulled all information up
I :
LJ to a centralized point for decision making when they really needed a reflexive production-
conrrol syste~ permitting each production point to signal its needs to the previous production
operation. In biological terms, Apogee was transmitting all information to its brain for processing
when it really needed to let its reflexes take over. When we put our finger on a hot stove,
we don't methodically review the situation and propose the best course of action. Instead,
our reflexes do the right thing by pulling our finger away. This is the simplest way to think
about the difference between push and pull.

Part 1: Getting Started 7

 r-1
L .l
ii
L _ _j
Similarly, the team could see that their facility was exposed to the full brunt of customer
orders, as if built on an unprotected coastline and exposed to storm waves. Yet humans D
always have sought to locate important sea-related activities in safe harbors rather than on
unprotected coasts. The purpose of the harbor and its breakwater is to prevent disruptive waves
from reaching the docks, even though the level of water still rises and falls over time in the
D
harbor to match the average level of the ocean. Apogee's location on an unprotected coast
was allowing waves of customer demand and mix fluctuations to flood through the plant
D
unchecked, becoming even more turbulent as they passed from department to department.
[J
What Apogee needed was a way to level and smooth external customer orders to protect the
activities within the plant from chaos while still serving the customer and while letting every []
production activity pull the materials it needed next from the previous process. They needed
to create level pull!

Level and Pull to Smooth Demand Amplification

n
Li

Fina(
Assembly
D
8uffer wall of Inventory

Internal variation minimal

to absorb variation
D
D
As we soon will learn, there are many places to locate strategic buffers that protect operations
from demand waves while serving the customer better and improving operational performance. n
L__J
Indeed, much of this workbook will be devoted to locating and precisely sizing the appropriate
buffers. The key point now grasped by the Apogee team was that inventories at the right
points could greatly improve productivity and customer response.
D

Do You Have SufficientStabilityto Embrace Level Pull?

D
An important question for you to ask at this point is whether you have sufficient
stability in your operations to move forward with a pull production-control system. ,-----,
I I
L__J
In general, if individual processes have uptimes of 75-80%, as they did at Apogee,
you can move forward on pull. However, if the output in many of your processes is ,-------,
less stable and predictable, lead times internally will vary tremendously and pull I
!L-J
l
'

production will be very hard to implement. In these cases you probably will do better
to spend a bit more time on point and flow kaizen to improve stability before n
attempting to make the leap to level pull. LJ

D
8
 At the end of their value-stream walk, Apogee managers resolved to install a truly lean
production-control system for every value stream in their facility. This workbook will show
how they did it, describing the questions they asked, the actions they took, the performance
targets they set (as shown in the chart below), and the timeline adopted for the initiative.

Box Score-All Value Streams

Original Current Target

LJ state state state

( ··1
Productivity

LJ Direct labor (pieces per person per hr.) 78 10.2 12.5

Material handlers per shift 24 25 15
Quality•

Scrap 5% 2% <1%
-

[J Rework 25% 15% <5%

External (ppm) 500 105 <50
r · ··1 Downtime
LJ Assembly (min. per shift) 40 30 <5
Paint (min. per shift) 30 15 <10
Molding (min. per shift) 20 20 <10
Inventory turns
r 1
Total 8 10 30
L..J
On-time delivery
r·· .. ,
To assembly 60% 75% 98%
LJ
To shipping 85% 85% 100%
To customer 100% 100% 100%
Door-to-door lead time
r: ···-, Processing time (min.) 126.0 125.7 125.7
LJ Production lead time (days) 36 30 12
Other
r ·1
I ,
i____J Overtime costs per week $30,000 $25,000 $0
Expedite costs per week $12,000 $9,000 $0
* Quality issues will not be directly addressed in this implementation effort. These targets represent long-term
goals for the value stream.

r
I '!
c...-1
r--··,
LJ

LJ
Part 1: Getting Started 9
 D
Implementation Approaches As Apogee set out to create level pull, it needed much more than
performance targets and the right questions. With a little experience []
I often encounter debates
you will find that setting reasonable targets is the easy part, and the
about which implementation
path to take-narrow or questions to answer are always similar among facilities. Apogee's [J
broad, fast or slow-and l most critical needs were the correct management team to spur the
always say, "It depends." It
depends specifically on:
transition, a clear plan to guide everyone's actions, and a reasonable
scope and timing for their efforts.
D
• Your level of knowledge
and experience as you
The Transition Team
D
start;
• The level of acceptance of
the concept within your
Apogee knew it was important for everyone in the facility to be D
involved, and created a special team for the transition. The team was
implementation team;
• Your need for quick
led by a dedicated leader from Production Control, the organization
that will operate the system over the long term. The team included
D
results as opposed to the
one manager from every area of the operation-shipping, final
need to get it right the
first time while educating assembly, paint, molding, die casting and machining, receiving, D
a larger number of materials handling, industrial engineering, and human resources. The
individuals; detailed implementation work then was done by a small staff, which D
• The nature of your worked full time on the project and reported frequently to the team.

•
production assets; and
Your tolerance for making Apogee set a six-month timetable to get the job done, which is
D
mistakes. reasonable in all but the largest facilities. The timetable listed every
Firms with limited knowledge task to accomplish, established times to start and complete each task, D
and experience, ambivalent and assigned clear responsibility for each task to a specific member
managersin some key positions,
and limited tolerancefor errors
of the team. D
will do better to follow the
incremental path described in The Scope and Timing D
this workbook. Other firms
Apogee could have started with the whole facility, done lots of
with more knowledge(perhaps
including an experienced planning, and switched from erratic push to level pull on a given D
external sensei),more buy-in, Monday morning. (And some facilities actually can do this. Yours may
and lots of courage may be one.) However, Apogee was attempting this conversion with no D
follow the all-at-once path prior experience operating a level pull system. In addition, Apogee had
and will gain the full benefits
of the transformation sooner.
limited resources-they could devote only a few full-time staff to the [l
project. The area managers on the team needed to perform their normal
However, the end objective
and the methods to employ
are the same. With the
tasks in their areas and could devote only a few hours each week.
D
Apogee decided to proceed in stages: They started with only one
information provided in this
workbook, you can successfully
product family-exterior mirrors-and began their implementation at D
follow either path or some the shipping dock for this one product. They then worked backward
path in between that best fits
your circumstances.
D
n
L__J
10
 to the two assembly cells for exterior mirrors, then to the paint booth, and finally to injection
molding. At the end of the first two-month phase they had created a level pull system for only
this product family. It was not very efficient from a total plant performance point of view
because the rest of the facility, including a fraction of each shared process (paint and molding),
was still operating on the old production-control system. But it worked and it demonstrated
the concept. Based on their learning and the growing acceptance of the concept among
formerly ambivalent managers, Apogee then transitioned the rest of the plant in a disciplined
manner over four months.

c·-·······..,
Apogee faced a considerable challenge in tackling system kaizen to complement previous
I i
LJ initiatives with point and flow kaizen. But the benefits were enormous. So, let's get started
and follow their progress.

D
D
D

[J
r·-1
LJ

Part 1: Getting Started 11

 r··---,
\ I
LJ
i j
I:._____j;

r·---.
1 i
I '
"'---'

Matching Production-System Capability to Demand

To implement a level pull system, Apogee's improvement team needed to start with one
product family-in this case, exterior mirrors-at the point closest to the customer. This
meant thinking first about the point of shipment at the end of their internal value stream.
lJ
LJ The Need for Simplicity
r·· . -, Two assembly cells were needed to meet the volume requirement for exterior mirrors,
LJ one cell for left-side mirrors, the other for right-side mirrors. This division simplified
fixturing and speeded changeovers, and it also served customer requirements, which
r---1 called for receiving lefts and rights separately.
LJ
Apogee began its implementation with only one of the assembly cells, the one for
left-side mirrors. This is the path we will follow until the end of Part 3, when we add
the second cell.

As you implement your own level pull system, it is very likely you will need to make
LJ similar simplifications. Remember that if you have limited experiences or resources
it is better to start with a simple path and make rapid progress than to tackle too much
complexity and get bogged down.

r l
I '
L.____J
Apogee always served its customer with timely shipments, which also was the price of entry
f""
I
---,
; in its industry because massive auto-assembly plants must never be stopped due to lack of
L...J parts from a supplier. Apogee tackled this problem by trying to maintain large buffers of
r··· r finished goods for every part number.
l ;
...____!
When parts shortages emerged, despite the large stocks of finished goods, Apogee compensated
r: · -,
'I II first with overtime and then by expediting products through the production process. In either
"------'
case, smooth operations upstream in the facility were disrupted. Because the existing system
iI :I
I '
'----'

,
I1..-l :
. '

r··· ·--i
!
i.......-....l

i
'.._j
Part 2: Matching Production-System Capability to Demand 13
 D

worked poorly and Apogee had new insights gained from the value-stream walk, the team
asked three simple questions: D
1. Which products should Apogee hold in its finished-goods inventory and which products
should instead be produced only upon receipt of a confirmed customer order?
n
L_J
2. How much of each product should Apogee hold in finished goods to protect both the
customer and the facility from disruptions? D
3. How should Apogee organize its finished-goods area to make management of the
inventories easy? D
These are the first questions you will need to answer as you start to implement your own
level pull system.
D
D
But Aren't All Inventories Waste? n....J
L..

Isn't inventory one of the seven wastes in lean manufacturing? Yes, Toyota does list
inventory among the seven forms of waste.

So why should you maintain or even increase inventories of at least some finished
goods? The reason is because not having a sufficient inventory of material can create n
L_J
even bigger wastes all the way up the value stream in the form of waiting, excess
transport, excess effort (expediting) and overtime, and intermediate goods inventories.

While it would be ideal to have a 100% build-to-order system with practically no

l.~J
lead times, no production shortfalls, no capacity constraints for dealing with demand
surges, and no finished-goods inventory, this is not practical for most industries with n
L-•
short customer-lead-time requirements. Inventory in the right place is a powerful
tool to buffer against surges in external demand as well as against internal process n
i.. .J
instability. Until you develop the capability to build 100% on-time with practically no
production lead time, and you have worked with your customer to smooth demand
surges, you are likely to require finished-goods inventories for many items to act as
a buffer.
D
Extra capacity and longer response time also can act as buffers, but both carry costs, n
L.J
either for you in the form of extra investment or for your customer in the form of wait
time. Attempts to dramatically eliminate all inventory will meet with failure unless i)
you first eliminate the problems that require you to hold inventory. Put another way, L __J
the popular saying, "lower the water level to see the rocks," sounds great, but all
you will gain is a nice view of some rocks and a hole in your boat unless you have a -l-,

L_J
way to remove the obstacles before proceeding. Above all, be practical in the initial
stages of creating level pull in your operations.
[]

n
~-~-.J

14
D
Question 1:
D Which productsshould you hold in a finished-goodsinventory,and
which productsshouldyou produce only to a confirmedorder?
D Like many manufacturers, Apogee had varying demand and mix for the different end
products flowing through its exterior-mirror value stream. Twenty-five different left-side
mirrors were made, varying by color and complexity (e.g., manually adjusted and power
mirrors, heated and unheated mirrors, and color combinations). The chart below shows the
D distribution of demand for each part number as a fraction of total orders.

Distributionof Demand by Part Number

D % 100
90
D 80
70
60
D 50
40

D 30
20
10
D Part number

D A items B items Citems

The bars in the diagram show the fraction of total demand accounted for by each part number. The curved line running
from left to right stacks the orders to show the fraction of total demand accounted for by any given number of products.
For example, the first five part numbers account for 60% of total demand and the first 10 account for 80%.

D
Apogee used these data to conduct a product segmentation that lean thinkers call an
D "ABC production analysis." (Do not confuse lean ABC analysis with a common practice at
many companies of categorizing inventories as A, B or C by annual dollar volumes.) In doing
D this they noted that five of the 25 part numbers accounted for 60% of demand and were ordered
every day by the customer. These were the five A items or high runners. A second group
of five products accounted for another 20% of demand and were ordered frequently but
not every day. These were the five B items or medium runners. The third group of products
consisted of the remaining 15 part numbers and also accounted for 20% of demand, but each
D was ordered infrequently and in highly variable amounts. This group-C items or low runners-
included infrequent color and build combinations, special-edition items, and replacement
D parts, many of them for out-of-production vehicles. (Some manufacturers apply different
names to their A, B, and C items, such as "runners," "repeaters," and "rogues," respectively.)

D Part 2: Matching Production-System Capability to Demand 15

 D
Apogee had been treating all three types of parts in the same way through the same ,,....---,

centralized scheduling process and maintained considerable stocks of finished goods for L_J
every part number. However, due to the batch sizes in the plant and the lengthy lead time
to produce parts, the plant still encountered out-of-stocks in finished goods, necessitating ii
LJ
expedited production and shipment.

D
Deciding about Finished Goods vs. Make-to-Order
With their ABC production analysis in hand and knowledge about the ability of their D
production process to deliver on schedule, the Apogee team was ready to decide which
products to hold in finished goods and which to make-to-order. They constructed the
following chart as a way to list the logical options.

Options for Finished Goods vs. Make-to-Order

D
Apogee Apogee
D
Options Pros Cons situation decision
1. Hold finished-goods
inventory of all products
Ready to
ship all items
Requires
inventory
Finished-goods
stores and
Not practical due
to physical layout
D
(As, Bs, and Cs) and on short for each shipping unable constraints and ,----,
make all to stock- notice part number to hold all
replenishment pull system. and much items
number of
end items
LI
space
2. Hold no finished-goods
inventory and make all
Less inventory Requires
and associated high process
Production lead
time too long
Not practical with
current lead time
D
products to order- waste stability and and paint and capability
sequential pull system. short lead
time to
process too
unstable
[J
produce

3a. Hold only Cs in inventory Less inventory Requires Daily stability a Possible second LJ
and make A and B products mixed concern step for future
to order daily-mixed pull production
system. control and D
daily stability

3b. Hold A and B products Moderate Requires Most applicable Best fit for today D
in finished-goods inventory mixed to current
inventory. Make Cs to
order from semifinished
components-mixed pull
production
control and
visibility on
situation
D
system. C items
D
Option #1 would be the safest and easiest to implement, but this required too much space
(and to many inventory dollars) in finished goods tO hold adequate amounts of every part D
number. In addition, Apogee needed to create more space in the facility for product launches
in the upcoming year. D
16
 [J

,-.--~,
I ;
._____,
I :
Replenishment Pull System

Holding finished goods for every type of product and using withdrawals by the customer
D to trigger production-as Option #1 requires-is known in lean manufacturing as a
replenishment pull system: only consumption of the end items triggers replenishment
of product. In this case the production instruction would be sent to the final assembly
line from finished goods via the heijunka leveling device (to be explained shortly),
and then back upstream from assembly.
[J
Production Control

Q Production instruction kanban

O Withdrawal kanban

Option #2 was impractical at present because of the 30-day production lead time and
process instabilities, particularly in the paint area. No customer was willing or capable of
r--···, placing 100% firm orders so far in advance, and Apogee couldn't guarantee that all items
I I
I i
.____. could be made on time even if they did.

[]

r··-~

LJ
r'"······-1
I
'----..J
!

Part 2: Matching Production-System Capability to Demand 17

 n
l_ ___J

D
Sequential Pull System D
Producing all items to customer order-as required by Option #2-is known as a
sequential pull system: items are paced and built in accordance with demand, with
the build instruction sent to the first process step at the beginning of the value
stream. This type of pull system is more demanding to manage than a simple
replenishment pull system because it is hard to pace the flow of operations to takt
time. Unless your facility has a short and steady internal lead time and high
equipment availability, this option will be difficult to maintain. Even in lean
companies, such as Toyota, sequential pull is employed only when the operations
D
have demonstrated high capability, and special build-to-order situations are
required by the downstream process or customer. D

g
Production Control
==::::: D
.....----- -11NJ----
joxox]
. D
.
0.
t
I,~~~ I \W1 ii
L...i

LJ -FIFO-
LJ -FIFO-
G l l
D
11
_: _J
Option #3a was attractive because it minimized inventories of high-volume A and B items.
But it required the assembly cells to receive good parts from paint in time to consistently
deliver A items and B items on schedule to the finished-goods area. The team decided that
D
this could be a goal for the future if production lead time could be sufficiently reduced and n
:_,
the capability of the paint process improved. But this option carried too much risk for the
present situation.

Option #3b was chosen as the best fit between the needs of the customer and the current
performance of the facility. This is a version of a mixed pull system where some items are
built-to-order while others are replenished to stock. This option required holding appropriate
finished-goods inventory for the A and B items, while assembling the 15 C items only to
customer order. (Parts 3 and 4 will explain how the Apogee team handled the C items.)
D
n
L~i

n
I l

18
D
[J

Mixed Pull System

The scheduling required for Option #3 utilizes a mixed pull system: aspects of both
D sequential and replenishment pull are utilized in conjunction. This is particularly
useful when the majority of items requested are frequent repeat orders, but many
infrequent items are required as well. (Note that Apogee chose to assemble C
D items to order from parts held in the market after the paint process. The diagram
below, by contrast, shows the more common situation in which orders for C items
D are sent to the beginning of the value stream at molding.}

Production Control
D
D 8
D D ~-----E-~!:!: -~ox I
-FIFO- -FIFO-

r---,
LJ
LJ3
. G.
I
:,
I I
:
I I
:,
I I
~ 1•

1
·--- 0 :: D . : 0
__ ! ·---- , __ J • __ .. ..! :: D:: D-
·---- ·---·
:... --

AandBitems
O Production instruction kanban

D O Withdrawal kanban

i~
I
L___.l
I

Choosing this option meant that Apogee spent less time managing the 80% of volume
D accounted for by A and B items, as long as they were consistently replenished on a daily or
every-other-day basis. This enabled Apogee managers to concentrate daily efforts on the
D more management-intensive C items.
r--·-
i i
L_J

~'
L'

r--:
,·-,
LJ

Part 2: Matching Production-System Capability to Demand 19

 []

n
LJ
Handling A, B, and C Items: Together or Separately?
You may need to take a step back and think hard about whether C items should be
produced in the same assembly cell as A and B items. If the work content between
the items is quite different or the changeover pattern complex, you may want to
n
consider putting C items in a highly flexible assembly cell dedicated to low-volume
production runs. This C cell also could include low runners from other product families. D
Having a dedicated cell for C items may simplify the pull system for the A and B items.
Additionally, in industries where the profit margins on C items, such as spare parts, D
are much higher than for A and B items, it may make economic sense to produce
them in separate cells even if the cost is a bit higher.
l I
In Apogee's case, because the work content and the changeovers in assembly for C
items are not disruptive, C items will be produced in the same assembly cell. In your n
[__J
own implementation you will need to think carefully about which option to follow.
Either way, taking the time to weigh the options and collect the necessary data on
current plant performance and customer demand patterns will pay large dividends
in the performance of your level pull system.
D
n
L.._J

Question 2: iiI
L._J'
How much of each product should you hold in finished goods?
Having decided to employ a mixed pull system and hold finished goods for the A and B !1
I I
~.)

items, the next question for the Apogee ream was how much of each of these items to hold.
After some discussion, the team adopted a simple formula to calculate the initial finished-
goods inventory levels.
[]

Finished-Goods Calculation D
Average daily demand x Lead time to replenish (days) Cycle stock

+ Demand variation as% of Cycle stock Buffer stock

+ Safety factor as% of (Cycle stock+ Buffer stock) Safety stock

= Finished-goods inventory
~
LJ

20
D
Finished-Goods Calculationfor Part #14509
1~·1
160 x 5* Cycle stock 800
r-·--,
200
LJ + 25%** of BOO Bufferstock

+ 20%*** of (800 + 200) Safety stock 200

r-1 = Finished-goods inventory 1,200
LJ
* Production Control schedules production of this part number once a week and average daily demand is 160.

D ** Reflects two standard deviations of demand and thus approximately 95% of normal order variation. If necessary, more
standard deviations can be taken to cover a higher level of variation.
"** Reflects the worst-case example of scrap, rework, and typical downtime amounts at Apogee.

D
Because of the importance of this formula, it is worth walking through the calculation
[J briefly, as Apogee applied it to part #14509:

The first step is to determine average demand per day. Apogee specified this using a
[J three-month time span. (Of course, you can use a longer or shorter period depending on
seasonality and likely changes in average demand due to market conditions.) This amount,
r- · 1 160 pieces for part #14509, is then multiplied by the number of days between scheduled
replenishment (the lead time to replenish) for this part number-five days. Because the
[J part is produced only every fifth working day, the maximum cycle stock inventory
immediately after the part is produced must be five days worth or 800 pieces (160 x 5).
Otherwise, supplies of chis part number would be exhausted before the next scheduled
[J
production date. During the five-day cycle the amount on hand will fall steadily-it will
r-···-~ cycle down and thus the term "cycle stock"-until it reaches zero just at the point of the
I
..____.;
next replenishment (see Finished-Goods Inventory Example on page 22) .

D What if there is a surge in demand during the five-day period before the next replenishment?
The second factor in the calculation, buffer stock, deals with this problem by adopting the
I ~: statistical technique of standard deviations. Standard deviations are used to calculate how
likely it is that demand will surge by more than a certain amount during the replenishment
period of five days. Apogee decided to set their buffer stock at a level of two standard
deviations above average demand, which means the odds are 19 out of 20 that demand will
not exceed the amount set in the buffer. (You, of course, can pick higher or lower odds.)
Two standard deviations in demand equated to 25% above 800 pieces, or an additional 200
pieces. (Any introductory statistics text gives the formula for calculating standard deviations
r·~1
and there are simple functions in spreadsheets that will do this automatically for you as well.)
LJ
r·,
I .
L..J

,-1
I I
L__J
Part 2: Matching Production-System Capability to Demand 21
 D

What if there is a shortfall in production from downtime or quality losses and the expected
number of new parts doesn't arrive in finished goods when planned? The third and final term
in the formula, safety stock, deals with this problem. Based on data collected from the previous
three-month period, Apogee calculated that the maximum possible shortfall was 20% of the []
combined cycle stock and buffer stock (20% of 1,000), or 200 pieces. Apogee's paint department
represented the least stable part of the production process, and had worst-case rework rates i-.,
L_._____J
of up to 15% on any given day. Additionally, minor downtime and scrap losses totaled about
another 5% in the plant. (You may choose a lower or higher percentage, of course, based on
analysis of your production reliability and capability to respond to problems. Just be sure co D
rely on real data from the recent past rather than promises from production managers that
I'
they will be more reliable in the future. Remember, all inventory amounts can be adjusted L_J
downward over time as improvement occurs. Starting too high is much less of a problem
than starting too low.) D
By adding the maximum cycle stock (800), buffer stock (200), and safety stock (200), Apogee
calculated that the maximum number of pieces of finished goods for part #14509 was 1,200 D
pieces. It also was easy to see that the minimum number of pieces-at the point just before
the weekly replenishment-would be 400 pieces if everything was running normally. D
Apogee selected this finished-goods inventory formula for two reasons: First, it was simple
and everyone could understand it. Equally important, it emphasizes internal production lead D
time to replenish (frequency of replenishment), which highlights the importance of reducing
lead time as the best internal way for operations to reduce inventory. This focus on production D
lead-time reduction will be critical to the success of the project as Apogee moves forward.

Finished-Goods InventoryExample
D
Finished-goods
inventory levels
Part#14509 Key drivers []
Cycle stock*

---- I
Amount required Average demand
-1

l__ ,

J
800 to cover normal
demand Lead time to replenish
--------
Buffer stock
Amount required ----
Demand volatility D
to cover customer- -----.._
1--- ........
-...L.._-4
! induced variation

Safety stock
Amount required ----
Forecast/MRPerrors

Downtime losses
Ii
L __J

200
to cover internal
losses
-----.._
Quality losses D
Time ~
CD
:,;-

" Assumes an average customer demand and draw down during each of the three weeks. Greater than average demand or
production shortfalls due to quality problems and equipment downtime will require Apogee to consume some buffer and/or
n
L_J
safety stock.

,...---,
' I

' I
'---------'
22
.--··----,

i
l....--1
i

After doing their math for all the required part numbers in the value stream, Apogee calculated
an overall need for finished goods for the exterior-mirror value stream that was about the
same as the total finished-goods inventory currently on hand-but the inventory now was
practicallyall A and B items. In the future, the only C items in the finished-goods area will
be products already completed as part of a large customer order.

D To transition to the new system, Apogee's Production Control Department issued instructions
to Operations to run overtime on the weekend to build up the necessary inventory levels for
I '

LJ the A and B items. The existing C items were put into a special location and labeled carefully
for use in the interim until they were made only to order.
D
Question 3:
How will you organize and control the finished-goods store?
Apogee's project team also decided to alter the way they stored finished goods. The plant
had relied on a system that scanned the inventory as it entered the finished-goods area.
The product then was placed in any open location on the shelves, and the material handlers
[J scanned in the location as well with a hand-held device. Although fine in theory, the system
was prone to human errors and parts were frequently "lost," only to turn up later in an

D incorrect location. Also, first-in/first-out (FIFO) was hard to maintain because there was no
visual queue as to which items to ship first. Apogee needed to combine the benefits of
scanning with the simplicity of a visual method to manage the finished-goods inventory.
[~l
To do this, Apogee took to heart the lean concepts of visual control and workplace organization.
r··1 The team decided to create dedicated locations for each part number, which were carefully
designed to ensure that the oldest parts were shipped first. The improvement team also hung
signage above the storage locations clearly indicating the maximum quantities for each finished
item. This enabled the team to organize finished goods in a logical and visual fashion that
allowed Apogee to help distinguish normal from abnormal circumstances. It also eliminated
the daily search for products and improved the efficiency of the Shipping Department as
they picked the daily order for the customer.
[J
As a tracking and daily management tool, Apogee's Production Control Department created
a spreadsheet for finished-goods inventory. Each part number was divided into the three
categories: cycle stock, buffer stock, and safety stock. All three levels for A and B items were
tracked and updated by the shipping department on a regular basis at the end of the first
shift. The numbers constantly moved (especially in the cycle-stock category, as the customer
took inventory away before replenishment at the end of the period), but it still was easy
to see whether inventories were within the normal range. This tool helped Shipping and
Production Control agree on the status of the inventory, what needed to be shipped next,
and the high-priority items for production.

Part 2: Matching Production-System Capability to Demand 23

 .~)
L_J
Tracking Spreadsheet for Exterior-MirrorFinished-Goods Inventory j/

L _,I
I

Finished-goods inventory

Safety Buffer Cycle

Part# maximum/actual* maximum/actual* maxi mum/actual* Tues. 8:08 p.m.

A items D
14509 200/200 200/200 800/900 Over/;,y100

14504 175/175 175/175 700/600 OK

J
l_ __

14506 150/150 150/150 600/450 OK D

14502 125/125 125/125 600/300 OK

14508 100/100 100/50 600/0 Into buffer stock

I i
B items []
14505 100/100 100/100 300/100 OK

14503 100/100 100/100 200/50 OK

n,~

14507 100/100 100/100 200/50 OK

D
14501 100/100 100/100 200/100 OK

14510 100/100 100/100 100/0 Out of cycle stock

C items**
!]
"
**
The actual amount is recorded at 3 p.m. each day after the first-shift daily shipment and completion of first-shift production.
C items are built to order and not held in finished-goods inventory. Under the current temporary plan, C items are
blended into production between runs of the A and B items, using overtime when necessary.
n
'--__j

To simplify flow within the facility, all finished goods were scored after assembly near the n
L-------'

shipping dock. The buffer and safety inventories were marked clearly on the holding racks
and kept separate from the cycle stock. Control rules were adopted as well:

• Buffer inventory could be removed from finished goods by Shipping only with the
authorization of the Production Control Department.
• Operations and Production Control needed to jointly review usage of buffer inventory r---,

to determine if it represented a fundamental change in customer demand or merely a L_J

one-time occurrence.
• The plant manager needed to authorize any use of the safety-stock inventory. Depletion
of finished goods to the safety-stock level indicated a serious production problem requiring
immediate countermeasures followed by investigation for the root cause.

24
r· ......
I .
I
L____.;
i

When shipping releases were received, the Shipping Department picked the necessary pieces
from the finished-goods storage location for the A and B items several hours in advance of
every shipment. For the time being, Production Control manually scheduled the items that
i-i
LJ needed to be replenished by final assembly after the daily shipment, basing this decision
upon calculations involving the depletion of the finished-goods inventory.

For the short term, Production Control also handed the production supervisor a list of the
r··, C items to produce, giving the supervisor the list several days before they were due. Using
LJ available production time between A and B items during the week and overtime when
necessary, Operations made the C items. Shipping collects C items as completed in Operations
D and places them in the shipping lane along with the correct A items and B items for the
next delivery truck.
[J Because the assembly cell was building batch sizes larger than the amounts normally taken
away by the customer each day, Apogee could not yet build directly to dedicated lanes in
[J shipping and in small amounts. However, the plane was moving closer to building to daily
replenishment for A and B items and closer to build-to-order for C items. (Note: This does
not represent the final Apogee solution. As Apogee progresses, we'll see how they improve
the handling of daily production control and devise system improvements to refine the
process, but only after a sufficient foundation is in place to move forward.)

,···· ·,
LJ
Make It Clear: Normal vs. Abnormal

If possible, organize and segment your inventory as Apogee has done. The main
goal is to make clear to everyone whether inventory levels are normal or abnormal.
r· ·1 Doing this correctly pushes down the ownership and control for inventory closer
Ii___J l
to the assembly cell and surfaces problems on a real-time basis for the managers
and employees. Additionally, it can eliminate the need for extensive computer-
generated reports, which typically are kept out of sight or entirely within the
production-control system.

LJ
(" ..... '
LJ

1·····-··1

LJ
f .. -·1
L__j

lJ Part 2: Matching Production-System Capability to Demand 25

 r
n
L_ __..,

D
D
:------.
Matching Production-System Capabilityto I
_ _J I

Demand-Keys to Success
• Always go to the gemba (the production floor) to get [l _j

facts, data, and anecdotes. Don't rely upon second-

hand information.

• Determine if your production processes are stable

D
enough to work around internal delays by buffering.
If not, find the root cause and solve the problems
before proceeding with your level pull system.
D
• Segment your demand by A. B, and C items, and
determine what type of pull system works best for you.

• Carefully calculate the amount of cycle, buffer, and

safety stocks for each finished-goods part number
required to ensure on-time shipments to the customer
without overtime or express freight.

• Practice good workplace organization and visual

D
control in the finished-goods store and shipping area.
[]
• Strive to make the status of finished goods-normal
vs. abnormal-clear at all times to everyone.
n
__ J

r-.
' I
L_J

r--.
LJ

(ii__J

26
 ii
I '
i__J

D
[J

Creating the Pacemaker

Apogee managers now had a firm grasp of their customer demand and the complications it
was causing the plant. As a first step to stabilize the situation, they had determined which
parts they would hold in finished-goods inventory, how many of each pare to hold, and how
to organize the finished-goods area for better visual control.

After one week, the results of their activities in finished goods were encouraging: the number
of expedited orders during the week fell by about one-third and the amount of daily overtime
r·, attributable to expediting decreased from one hour per day to 40 minutes. (This was after
Il..--l I the one-time buildup of A and B items required to change the composition of the finished-
.r----
goods inventory.) Finished goods were ready to ship 92% of the time, up from 85%.
i !
' I
L...-l
The Apogee improvement team was pleased to see that the symptoms had eased, but they
knew they had not resolved the root cause of the problems. Moving material in a timely
LJ fashion between departments and into finished goods was still difficult because the assembly,
r .. -.. --, paint, and molding processes were responding to multiple schedules.
LJ
To address the root cause of their inability to ship on time, the Apogee team needed to
r 1 determine where to schedule the value stream, how to level production at this location, and
LJ how to convey demand to this location. Doing so would aid in controlling production
throughout the plant.

r ··--,
I ,
,_J
Question 4:
Where will you schedule the value stream?
~·· 1 The team reviewed Apogee's value-stream map for the left-hand exterior-mirror value
LJ stream, this time focusing on the flow of information.
r · 1
L

r .J
i I
~
Part 3: Creating the Pacemaker 27
 r--1
I '
L.. J

·L_J
~
Left-Hand Exterior-MirrorValue Stream
iJ
L

Production Control
~ I 12·Weetf0f'l:Ca~t I
[1NJ-----::z.....~- ~
~ D
~·-~-· I~ ~ l__l
<, I o.,~::::Md·~ I C=b
,. 1
,..·····~~
. ,..·····~···.\!} .
[]
&..+ LJoldinq
15days
~
5days
u-lnt ~ """i"'" ....... ~ LJhlpplng
5days 0Gl © 0
D
Q]ql em
5days

The map showed three distinct locations where scheduling occurred on a weekly basis; four
locations if you counted the daily ship schedule. (Because molding and paint were shared
D
with other value streams, their schedules were particularly complicated.) As indicated by ~
1 I
I \
the telephone calls from each process back to the previous process-sending instructions to ;______;

ignore the schedule and produce the parts needed immediately by downstream processes-
the more scheduling points in a value stream the greater the chance for errors. n
L-

From their earlier data collection, the team had gathered a number of anecdotes from the
team leaders and operators on the shop floor about the actual use of the schedules by
molding, paint, and assembly:
LJ
• "The schedule changes at least five times per week, so I don't put
much faith in it at all."-assembly cell operator
• "I wait to be told what to make from paint; that way I know it is

•
urgent." -molding operator
"There are a few part numbers that really matter so I run them in the
n
__
[ ,

biggest batch possible until I have to change over."-molding operator

• "The schedule is more of a guideline; from experience we know what

•
needs to be made."-paint department supervisor
"I get measured more on productivity than schedule adherence, so I just
n
L__,

try to make lots of parts."-assembly cell operator

D
As shown by these comments, scheduling information throughout Apogee quickly became r--,
outdated. And even if it was correct, it often was ignored. Multiple scheduling points combined L_j
with inherently unreliable forecasts, long lead times, large batch sizes, and department-

0
n
I~

28
[J

centric metrics led the Apogee molding, paint, and assembly departments to a state of
"What is best for me?" instead of "What does my customer need next?" The result was
excess inventory of most parts combined with a failure to produce all of the right parts in
the right amount at the right time for the customer. To overcome these problems, the
Apogee team needed to designate a single point in the value stream as the pacemaker to
receive the schedule from Production Control.

Traditional SchedulingCan't Keep Up, But Don't Throw It Out

Apogee's performance and the comments of production operators and first-line

managers are typical of facilities that employ traditional scheduling techniques. In
r ·, theory, sending multiple schedules from a central MRP system to each department
I i
L___j
should keep everyone informed and working to the same cadence. In reality it rarely
does. Problems inevitably enter into the equation when assumptions for lead time,
scrap and yield rates, and other inputs are wrong.

The early-generation MRP programs have algorithms that assume infinite capacity in
the system, a condition that never exists in the real world. They also frequently build
in extra buffer or safety inventory at each step of the process. Recent systems are
more refined, but the shop floor is still a dynamic place. It changes minute by minute
LJ throughout the day while MRP systems typically work with a time fence of anywhere
from a shift to a week. The MRP needs to be continuously updated about the actual
status of production on the floor, but this is difficult to achieve. Often schedules,
production status, and inventory levels are only updated overnight in a batch program,
making them useless for resolving problems arising throughout the day.

Does this mean you should yank out your MRP system or develop an anti-IT bias?
Certainly not. But you should recognize the inherent weakness and limitations of
existing systems. Manufacturing companies always will need some type of MRP
system to hold the bill of material, create rough-cut capacity plans, handle forecast
r ··1 information, and complete other useful tasks in production planning. However, even
L.J the most advanced software systems poorly execute real-time, shop-floor control
for production between processes. This failure is what we have observed at Apogee.

lJ Pull production in a value stream as regulated by a pacemaker has great advantage

over most standard software applications: With shop-floor production control,
employees can sense and react more rapidly to changing production dynamics.
There is no delay while waiting overnight or until the next schedule can be created
and transmitted to the floor. Response to a problem can oe nearly instantaneous.
D Additionally, shop-floor control puts responsibility and capability for solving problems
in the hands of those actually running the process.
r .. - l
I I Information technology continues to advance, but it remains a challenge for standard
L_j
scheduling systems to incorporate the full logic of lean.

~-,
LJ Part 3: Creating the Pacemaker 29
 LJ
Guidelines for selection of the pacemaker process
Apogee needed to select one process to function as the overall pacemaker in the value stream.
l_J
This would eliminate confusion over the "right" schedule, and it would ultimately enable
everyone to march to the same beat-take time at the pacemaker process. Coupling this beat D
with shop-floor control cools, especially kanban, would facilitate a smoother pull of material
on time to the customer through the value stream. D
Apogee kept the following in mind when selecting their pacemaker:
• In replenishment pull, final assembly will be the pacemaker in virtually every case.
D
• In sequential pull, the pacemaker often is the first process at the beginning of the value [J
stream. If possible, however, the pacemaker should be located further downstream just
before inventory types proliferate.
D
At Apogee, it was easy to see that the final assembly cell should be the pacemaker for the A
and B items. The C items, however, presented a dilemma. One approach was co schedule the
C items at the molding process, but the instability of the paint process between molding and
D
assembly meant that the parts might not show up on time in assembly. (This often is a problem
with sequential pull because you must maintain rigid FIFO control and pace the flow of parts
D
to sync with take time in final assembly. Otherwise you wind up with sequential push instead
of pull!) n
To simplify the problem given current stability, Apogee decided to create an inventory of
painted parts for C items in a market between paint and final assembly, and to assemble these
D
items only when an order was received. (To do this, Apogee will use a scheduling method that
soon will be described.) This greatly simplified the mixed pull system by creating only one
n
L_ __J

schedule point-in final assembly-for the value stream.

Question 5: n
L J
How will you level productionat the pacemaker?
Perhaps without realizing it, Apogee had already taken the first step in leveling output at n
L__J
the pacemaker. This had happened two years earlier when Apogee created its continuous-
flow cells working to take time with standardized work. The operations team had balanced
the two exterior-mirror cells and their daily output to a take time of 54 seconds, meaning
that 500 pieces were produced during every 450-minute shift. The previously erratic daily
output had been replaced with output leveled in terms of the quantity produced per shift. D
(Of course, when customer demand soared for several days or plummeted briefly to a much
lower level, the cells either worked some overtime or stopped early. The key breakthrough
was getting the cells to produce at a level rate whenever operating and to stick with that
rate until takt time was changed when long-term demand changed.) '\I .
L.J

30
 ,·-·-1
Now Apogee needed to level production by mix by reducing the batch sizes produced of
each part number in the cells to better reflect the amounts actually requested by customers
in their daily shipping releases. A quick look at the chart (see Apogee Orders vs. Apogee
Build) showed that Apogee's current batch size of 500 pieces for every item was far out of
sync with typical daily requirements.

Apogee Orders vs. Apogee Build Schedule

D Customer Final-assembly
requirements build schedule (left-hand cell)
11
LJ Part#
Demand Mirror Monday Monday Monday
category description orders 1st shift 2nd shift

14509 A Black heated 140 500 0

[J 14504 A Black unheated 110 0 500
14506 A White heated 120 0 0
D 14502 A White unheated 120 0 0
14508 A Red heated 110 0 0
D 14505 B Silver heated 70 0 0
14503 B Yellow heated 60 0 0

D 14507
14511
B

c
Bronze heated
Purple unheated
70
100
0
0
0
0
14512 c Gold unheated 100 0 0
Total 1,000 500 500

[J
Apogee managers historically had run the cells with large batch sizes because they fundamentally
believed that the plant ran more efficiently this way. Prior to the lean initiative, a changeover
had required a substantial stop of production tO change fixtures and ensure that all necessary
j[ parts were in the correct positions at the cell to run the next part number. This stop was
LJ
perceived as a "productivity loss" that could be minimized with large batches that reduced
the need for changeovers.
D
The belief that efficiency was maximized as changeovers were minimized was even more
r·, I
!
L_j strongly held by managers in the upstream batch processes in paint and molding departments
where changeover times were much longer. The paint department always sought to make
,---. at least 1,000 items of a given color once it was scheduled and often ran two or three times
LJ more of that color once it was running smoothly to avoid changeover problems in the
automated cleanroom. Molding then would double the paint department's planned batch
0 size (to 2,000 pieces) to further minimize downtime due to its changeovers, which were
even more time-consuming.

Part 3: Creating the Pacemaker 31

 [i
L__ J

D
At each process step up the value stream, production became less level in relation to customer
needs, which was one right and one left mirror for each car moving down the customer's D
assembly line (and often in a different color or style for the next car). This is shown in the
diagram (see Sample Lot Size Comparisons Across the Extended Value Stream). D
Sample Lot Size Comparisons across the Extended Value Stream
D
Quantity
2,000 right
2,000 left
D
2,000
D
1,500
1,000 right
1,000 left
D
1,000 -
500 right
5001eft
D
500 - Average part
1 right 10 right number order
1 left
-
10 left
------------------------·------- ......... ------- ------- ------ - .......... ------- quantity = 50 D
Customer
usage
Customer
pack
Apogee
assembly
Apogee
paint
Apogee
molding Processstep D
quantity quantity quantity quantity

LJ
Based on what they had learned, the Apogee team now could see that the previous definition
of efficiency had been far too narrow. In order to operate each point along the value stream D
efficiently-in final assembly, paint, and molding-Apogee was building large batches far
removed from customer needs. Point efficiency at each process step was producing much
larger system inefficiency in the form of inventory carrying costs, space requirements,
expediting of missing parts, and general management overheads. [J
The Apogee team now also realized that the biggest gains from reducing batch sizes would
be in inventory reductions because batch sizes determine the replenishment cycle for D
finished goods. The team could see this by referring back to their formula (shown on page
20) for calculating the minimum amount of necessary finished goods. Shortening the "lead
time to replenish" from the pacemaker, which could be done only by reducing batch sizes,
would directly lead to a decrease in finished-goods inventory. []
For example, if Apogee could build all five A items consistently on a daily basis (rather than .-------,
once a week) and all five B items every two days (rather than once a week), they could reduce L__ J
finished-goods inventories for these items by 80% and 60%, respectively (see Shorter Lead
Times Reduce Finished-Goods Inventory). n
L_ _J

:---1
L_J
32
 []
Shorter Lead Times Reduce Finished-Goods Inventory

Days of 5
cycle stock
finished goods
4
[J
3 60%

2
80%
r 1
L 1

Before After Before After

A Items B Items

The team now realized that batch sizes at the pacemaker should be minimized rather than

D maximized, subject to three constraints:

A. Work content differences between products

B. Changeover requirements between part numbers
C. Production pitch interval

A. Work content differences between products

Even though mix and demand may be level at a pacemaker, wide variation in work content
for different products moving through a pacemaker assembly cell can create an unreasonable
burden for production operators in the cell. Fortunately, Apogee had carefully evaluated work
content in setting up its final assembly cells two years earlier to be sure that the work content
for all of the products assigned to the cell varied by only a small amount and that no product
had work content above takt time.

The easiest item to produce was an unheated, manual mirror requiring 45 seconds of work
content per cell station. The most difficult item was a heated, powered mirror requiring 54
seconds of work per station. The weighted average work content for all items moving through
the cell was 50 seconds per workstation, which was well under the customer-based takt time
of 54 seconds. Work content difference therefore posed no barrier to Apogee in producing
r -1
L.____j
in smaller lot sizes in any sequence. And, in fact, reducing the batch size for the heated,

:-,
.
L-.J
.'

Part 3: Creating the Pacemaker 33

 powered mirrors with the 54-second takt time from the previous minimum of 500 would make
it easier to operate the cell smoothly through the entire shift because it would free up a bit D
of time to deal with unavoidable variations in output.

As you think about the batch sizes to use in your facility, the move to level pull is a great
time co reevaluate the work content of the part numbers going through each operation.
Significant variations that have gone unnoticed in a push-scheduling environment, particularly
L__]
if work content exceeds take time, will come painfully to light in a tight pull system.

B. Changeover requirementsbetween part numbers

If changeover times were long in the assembly cell (the pacemaker), it also would have been
hard to significantly level the pacemaker by reducing batch sizes because a substantial
amount of production time would have been lost. Fortunately, two years earlier Apogee also
had reduced changeover times at the pacemaker cell to zero. (This included both the time
for fixture changes and the time to get materials in place for the next part number.) This r
L__J
meant that there was no minimum constraint on the Apogee batch size due to changeovers. The
management had simply failed to translate this improvement into a reduction in batch sizes,
choosing to use the freed-up time from shortened changeovers for more production time. D
C. Production pitch interval
[]
When work content variations and changeover times do not present obstacles to reducing
batch sizes, production pitch determines the maximum extent to which the pacemaker can n
L-,

be leveled by mix. Pitch is a lean concept and is calculated by multiplying take time by
pack quantity (the number of products per container transferred to finished goods from the []
assembly cell). For example, the completed mirrors produced in the exterior mirror cell are
packed for shipment to the customer in containers with exactly 10 items. This means that
with a take time of 54 seconds, a pack of mirrors ready to ship to the customer is produced
every nine minutes, for a pitch (or pitch interval) of nine minutes. n
IL. _ _J
I

Production Pitch Calculation

n
L__ J

Takttime x Pack quantity = Pitch r--,

1 I
54 sec. x 10 pieces 540 sec. (9 min.)
L__J
=

r---ij
Because pitch is the bridge connecting customer pack quantity with take time, it makes no L~-
sense to produce quantities in the assembly cell that are less than pack size, or in any batch
size other than a multiple of pack size. This is because there is no way to convey smaller L_J
quantities or partial quantities to the customer. Therefore the minimum batch size in the
D
34
 D
D
Apogee final assembly cell was 10 items. (This could change in the future, of course, if the
pack size was changed, but this seemed unlikely at Apogee because the pack size was already
small. However, in your facility, reducing pack size may be an appropriate step in introducing
D your level pull system if the current quantity in your packs is very large.)

With clear knowledge of available daily production time, daily customer demand by part
number, and minimum batch size, the Apogee team now was in a position to level production
by mix. To do this they simply needed to divide the available production time (450 minutes)
by pitch (nine minutes) to calculate the number of pitch intervals (SO) available to meet demand.

Pitch Interval Calculation

D Time available + Pitch = Intervals

450 min. -e- 9 min. = 50 intervals

lJ
But how should Apogee allocate the 50 pitch intervals available by part numbers? From
[J their ABC analysis, the Apogee team knew that 60% of demand was for the five A items,
20% for the five B items, and 20% for the 15 C items. They therefore decided to allocate
the available production time by category of part number as shown in the chart.
lJ
lime Intervals per A, B, and C Items
[_J
%of
LJ Total interval x productionmix = Intervals per item (equivalent time)
50 internals x 60% = 30 reserved for As (9 min. x 30 = 270 min.)
r· · ·1 50 intervals x 20% = 10 reserved for Bs (9 min. x 10 = 90 min.)
50 intervals x 20% = 10 reserved for Cs (9 min. x 10 = 90 min.)

LJ
Using the pitch-interval logic just described, 60% of the available time was devoted to
D replenishing A items during each shift. Similarly, 20% of the time (10 pitch intervals) was
allocated to B items, and the remaining 20% of the time (10 pitch intervals) was reserved

.. for the C items.

Multiplying pack quantity (10 items) by the number of pitch intervals available for each
[J category of parts shows that Apogee will now produce 300 A items each shift, 100 B items,

D
D
Part 3: Creating the Pacemaker 35
 D
D
and 100 C items. This contrasts strikingly with the current situation where the cell
produces 500 items of only a single part number each shift. D
But which A, B, and C items? As a first step, Apogee determined to build each A item
every shift because customers ordered each of these items every day. By dividing the five
D
part numbers into the available capacity of 300 for A items, it was easy to see that the A
items should be built in batches of 60 (or six pitches) for each pare number as a first step. D
The next step was to consider batch sizes for the B items. One approach would have been
to conduct a changeover between part numbers after every two packs (20 pieces) so the five
D
part numbers could be built every shift. However, Apogee managers were worried about
changing over so often in a new system that was untested. And they also noted that B items D
were on average ordered only every other day. So to get started, they decided to make B
items in batches of 50, with the result that every part was made every 1.25 days. (As you D
approach your own decision on batch sizes, above all be practical. Setting an ambitious
target for batch sizes in an immature system actually may lead to slower progress than D
setting a less ambitious target, successfully achieving it, and then quickly reducing batch
sizes again.)
D
The final step was to set batch sizes for the C items. This was easy. In the block of time
reserved for C items on each shift, Apogee would try to produce a whole order requested D
by a customer and, if necessary, carry this order over into the next available block of time
reserved for C items at the end of the next shift. Alternatively, if a customer order was small D
enough Apogee could produce several different C items during the time interval reserved
during a single shift.
D
With these decisions made, it was easy for Apogee to plan how available production
capacity could be distributed by type of item (e.g., A, B, or C) and by part number (e.g., D
14509) as shown in the chart on the next page.

D
D
D
D
D
D
36
D
r·-,
LJ

D
(Please note this is not a schedule; the actual schedule will be based on customer
D withdrawals from finished goods, daily releases, and cell performance.)

D Planning Production by Product Type and Part Number

D A items
270 total min.
Bitems
90 total min.
Citems
90 total min.
30 time intervals 10 intervals 10 intervals

D Shift 1 6
#14509
6
#14502 I I 6
#14504
6
#14506 I
6
#14508 I
5
#14501 I
5
#14510 I To be determined

D Shift 2
6
#14509
6
#14502 I
6
#14504
6
#14506 I
6
#14508 I
5
#14507 I
5
#14503 I To be determined

Shift 1
6
#14509
6
#14502 I
6
#14504
6
#14506 I
6
#14508 I
5
#14505 I
5
#14501 I To be determined

D
6 6 6 6 6 5 5
Shift 2 To be determined
#14509 #14502 I #14504 #14506 I #14508 I #14510 I #14507 I
Shift 1
6
#14509
6
#14502 I
6
#14504
6
#14506 I
6
#14508
I 5
#14503 I
5
#14505 I To be determined

D Etc.

Average shift demand 300 units 100 units 100 units

Total finished-goods types SA items 5 B items 15 C items
D
Apogee recognized that further leveling of the schedule was indeed possible and desirable
so that every part number could be produced every day, then every shift, and then even
more frequently if necessary. As a first step, however, leveling production by mix to this
11 extent was both feasible and a tremendous improvement.
LJ

Question 6:
How will you convey demand to the pacemaker to create pull?
D Apogee had now answered five of the questions necessary for creating a level pull system.
The assembly cell was officially designated as the pacemaker process for the exterior-mirror

D value stream, and the daily quantity and mix of production were leveled by Production Control.

The next question to answer is how Production Control will convey demand information in
D the form of production instructions to the assembly cell to trigger pull production. (As we will
see, answering this question also will answer the question of how finished goods will be
D conveyed from the cell to the finished-goods market.)

In lean production, the specific tool for communicating production instructions and for
D regulating materials conveyance is the kanban. As a downstream process consumes product,
signals are sent back to the upstream process via kanban to replenish the amount consumed.
D When communicating over long distances, some form of electronic kanban often are used
in place of simple kanban cards, but within the four walls at Apogee kanban cards are a good
way to govern information and material flows.

Part 3: Creating the Pacemaker 37

 I l
D
Rules for Using Kanban Types of Kanban
[]
• The downstream process In-process kanban for
pulls from the upstream Production
[ scheduling flow processes
D
[
process. instruction
Signal kanban for scheduling
• The upstream process batch processes

replenishes the quantity

taken away.
Kanban
Interprocess kanban D
[
for internal purposes
Parts
• No defects should be withdrawal

allowed to pass on to
Supplier ka nba n for
extern a I purposes D
the next process.
• Kanban must be attached
and conveyed with the
Apogee now needed to create a set of production-instruction kanban
and specifically in-process kanban to communicate production infor-
D
part or container. mation from Production Control to the assembly cell.
• No parts can be produced D
or conveyed without a
Example of In-Process InstructionKanban

•
kanban instruction.
The quantity indicated on
D
Finished-goods Product Production
the kanban must equal
storage information information area area
the actual amount in the
container (to ensure
D
Finished-goods Part#: Line address:
accuracy of information). market

1145091
Assembly Cell #1 D
NOTE: Readers not fully
comfortable with the kanban
concept or completely
Supplier:

Internal
Part description:
LH Exterior Mirror
D
knowledgeable about the Heated Black Location:

different types of kanban Final Assembly

should refer at this point to Market location: Quantity; 10 Department
A4
Appendix: Kanban.
R2 D
D
Note that each in-process kanban card is an instruction to produce
one pitch (one container) of a given part number, in this case part D
#14509. Note also that the kanban provides all the information the
cell and the material conveyance system will need to make the right D
amount and send it to the right place:

• Storage location for finished goods in the market (on the left D
side of the card);
• Production location of the cell (on the right side of the card); and D
• Part number, part description, and the quantity to be made
(in the center of the card). [J
38
r·· -,
LJ
Apogee produced these cards for every part number to be run in the exterior-mirror cell.
D But how should they be sequenced, and at what pace should they be delivered?
r---,
\ ! One method is to collect all of the cards for the shift and deliver them to the cell at the
L_J
beginning of the shift, leaving the operators to decide in what sequence to produce products
and at what pace to produce. And this is what traditional schedules often do in mass production.
However, lean thinkers discovered long ago that a better method, and the one adopted at
Apogee, is a heijunka box. This simple device uses time intervals across the top (the column
labels) to visually array the production instructions-the in-process kanban-in a way that
makes completely clear what part number to produce next and at precisely what time.
c Apogee initially set the time intervals across the top of the heijunka box co correspond to
pitch intervals of nine minutes. The box was sized to accommodate all SO pitch intervals in
n
' !
L.___J
a shift and all of the part numbers to be run during the shift. The box then was loaded by
Production Control with all of the kanban cards for items to be produced during the shift.
(The box shown on the following page is loaded for the beginning of the first shift on
Monday at Apogee.)
r--·-1
r·----
I
.___I Heijunka Box Origins
While we commonly associate the heijunka box with production control for oper-

D ations, the first application of the heiju nka box concept at Toyota actually occurred
in maintenance. Many years ago, Toyota managers found it useful to create boxes
with hourly time intervals for scheduling preventive maintenance activities. By care-
fully timing the work content for each activity, drawing up a work sheet specifying
how much time was needed to complete the task, and placing the sheets in a highly

D visible box with time intervals clearly marked, Toyota helped its supervisors to pace
work while remembering to schedule all of the necessary tasks and avoiding
scheduling them all at once in a way that might interrupt production. The heijunka
device leveled maintenance effort (a concept you should try in your own operations).
From this initial application, finer divisions oftime evolved and the heijunka idea
,-1 migrated to pacing production work. The most widespread application has been
with suppliers, where the box acts as a tool to pace withdrawals of inventory and to
tightly link the output of the supplier's assembly cells (usually manual) to the takt
time of Toyota's assembly lines.

11
LJ

Part 3: Creating the Pacemaker 39

 n
c..._ ____j

D
Heijunka Box-Nine-Minute Interval
r:
L_J
Shift 1 7:00 7:09 7:18 7:27 7:36 7:45 7:54 8:03 8:12
n
L__J

Cell #1

D
But how are the cards to be conveyed from the heijunka box near Production Control to the D
assembly cell? Apogee realized that the best method is to deliver the production instructions
just-in-time to the cell, ideally one pitch of work (one container) at a time. Doing this introduces D
a strong sense of pace at the exterior-mirror assembly cell (which lean thinkers often call
"takt image.") It prevents working ahead and it also notifies Apogee managers immediately
if the cell is not keeping up with the output needed to meet the customer requirement.
D
The way to introduce just-in-time delivery of instructions is by means of a conveyance D
operator who picks up the cards at the heijunka box and takes them to the cell. This operator
also can push or pull a cart and pick up an empty container needed for holding the finished i' _ _J

goods indicated on each kanban being delivered. The container can be taken to the cell
attached to the kanban, where the conveyance operator can take away completed products and
deliver them to the finished-goods market. Doing this turns out to have the major advantage
D
of precisely linking material and information flow to and from the cell while removing almost
all empty containers and finished goods from the area around the cell. D
To establish the proper configuration of the conveyance route, Apogee needed to run several n
!__J
time trials linking the in-process kanban, the heijunka box, the empty-container storage area,
the assembly cell, and the finished-goods storage area. (You will need to do this as well ii
LJ
because there is no secret formula for determining conveyance routes.)

Conveyance Route for Information and Materials

., , ® ··----------------------------------:
,,,,' t D
DD DD DD DD DD ©
DD DD DD DD DD Ce/1#1
DD DD DD DD DD @]
@' D DD DD DD DD
0Q)Q©0
•
I
'
Empty·contalnerarea
~~ QJQ) ©[D r-,
' I
'• Kanban LJ
@ post

CD-+--------------® .... ©-+------------- 0 _,_ J D Empty container

IN#NN
Finished goods
DD DD DD DD DD DD
DD DD DD DD DD DD
7:00 7;09 7:18 7:27 7:38 7:45 7:54 8:03
Heijunko box
DD
DD
DD
DD
DD
DD
DD
DD
DD
DD
DD
DD
D
Finished-goods store area
/'1
L __)

40
 0
After several time studies, the work elements and flow of the conveyance route were
D determined to be as follows:

Apogee Conveyance Route ActivityList

Step Activity Time

1. Pick up instruction kanban from the heijunka box. 10sec.
2. Travel to obtain empty finished-goods container. 1 min.
3. Obtain empty finished-goods container for delivery to the 30sec.
assembly cell.
r ·1
L_j 4. Travel time to the assembly cell 30sec.
5. Drop off an instruction kanban and an empty finished-goods 1min.
[J container at the cell, pick up finished goods from the previous cycle.
6. Travel time to the finished-goods market. 1 min.
7. 30sec.
[J Drop off finished goods to their correct storage location.
8. Log product into finished-goods inventory (e.g., scan bar code). 1min.
9. Return to heijunka box for next instruction. 30sec.
Total time 6 min. 10 sec.
r ·1
L.....J
(Note: The LEI workbook "Making Materials Flow" provides a more detailed
description of conveyance route design that you may find useful.)

As established, the route took only six minutes and 10 seconds for the conveyance operator
to run-not the nine-minute pitch interval. Also, despite trying to adhere to a takt time of 54
seconds, the assembly cell output often was varying by up to one minute (plus or minus) off
r · -1 the targeted nine-minute pitch time. This fluctuation meant that sometimes the completed
Ll pack of finished goods was not ready for pick-up by the conveyance operator. Additionally,
one of the stations in the assembly cell had a labeling machine go down for 10 minutes
during the trials, an occurrence that rarely happened but in fact did on this occasion.
r ... ..,
Apogee's improvement team was learning that even improving the level and pull aspect of
LJ scheduling final assembly would not solve all production problems. (Clearly, more point kaizen
was needed to stabilize production output from the assembly cell.) More importantly, Apogee
learned that they could not react to problems in final assembly as fast as the nine-minute pitch
required. The downtime in assembly, although infrequent, meant that a nine-minute pitch
interval was faster than they realistically could sustain at their current level of stability.

r ·1
L-J
The team, therefore, made several short-term adjustments. They increased the pitch interval to
18 minutes instead of nine (meaning there would now be two production instruction k.anban in
each pitch interval per cell in the heijunka box). This way the cell was more likely to have one

f. 1
I '
l__J
Part 3: Creating the Pacemaker 41
 D
D
finished-goods container completed when the conveyance operator arrived, and there was more
time for the supervisor to react to minor problems that might occur during assembly. (You are
likely to encounter similar issues. As always, be practical as you start your implementation.)
D
Heijunka Box-18-Minute Intervals

Shift 1 7:00 7:18 7:36 7:54 8:12 8:30 8:48 9:06 9:24
n
Cell #1
D
.----,
LJ

Production Instruction Pitch

D
How short should your pitch interval be for production instruction? The answer
depends upon several factors.

One purpose of the pitch and conveyance withdrawal cycle is to create a sense of
pace for a manual assembly line. If you want a strong sense of pull and flow, a short
n
LJ
pitch of 10 minutes or less is better. This requires clockwork precision and stability
in final assembly, which most companies do not have as they start this process. n
[____J
The other purpose of the pitch interval is to create a tool to see if production output
is maintaining the scheduled amount throughout the shift. The in-process instruction 'IL_J
kanban automatically prevents overbuild of inventory because the delivery timing is
regulated by the conveyance operator. If the assembly cell is behind, then the pitch
interval can become a time bucket for the team leader to manage, determining if
the team is keeping up with the schedule. If the cell frequently falls behind, some
D
type of visual signal can be created, such as a red light or accumulation of instruction
cards in a collection post at the cell. Toyota calls this type of signal an andon, of
n
L__J

which many varieties exist. In this way, the interval of pitch can become a visual
tool for managers to assess production status and react to problems. n
L___,
Most companies can't assess production status every 10 minutes at the outset of
their implementation. If you calculate a mathematical pitch as short as 10 minutes,
you should carefully weigh the benefits vs. the risks. Most companies are probably
D
better off with an initial pitch of 15 minutes to 30 minutes. Any longer than that and
you are back to taking hourly or mid-day production counts and not discovering and []
reacting to problems early enough on the shop floor.

D
[]

42
n
r---1
L_j
The new pitch left the conveyance operator with only six minutes of work during an
18-minute interval, which obviously was inefficient. As the Apogee team considered this
problem, they realized that it now was time to add the second cell for exterior mirrors (for
[J the right-hand versions) to their level pull system. This required a bit of work to calculate
the appropriate finished-goods inventory for each part number and to add these part
r~-
. I numbers to the finished-goods market. But once this task was done, it was easy to add
LJ
a section to the heijunka box to regulate the output of both exterior-mirror cells.

Heijunka Box-Exterior Mirror Cells

Shift 1 7:00 7:18 7:36 7:54 8:12 8:30 8:48 9:06 9:24
[]
Cell #1

Cell #2

D
By combining instructions and conveyance to both cells, the work time for the conveyance
operator increased to 10 minutes. This percentage of work still was inefficient, but for the
short term the objective was to keep the implementation moving ahead; the team decided to
operate the route for two or three weeks while documenting their learning.
D As soon as possible Apogee will expand the system to all six cells in final assembly, developing
[-1 more complex routes that should increase conveyance-operator efficiency. In the meantime,
the conveyance person will collect information and document how many times the cells

c-1 completed their production on time as well as causes of delay. (This information will help
prioritize point kaizen in the cell and on the conveyance route.)

[J

[J

0 Part 3: Creating the Pacemaker 43

 ) J

Alternative Heijunka Methods 0

Apogee employed a combination of instruction kanban loaded into the heijunka box
with a fixed-time conveyance route. This option represents one way to use the
heijunka box, but it is not the only way or necessarily the best for your situation.
11
For Apogee, the actual ship quantity and mix can change fairly late in the sequence
of events (even hours before the delivery truck arrives) and so Apogee must hold
D
finished-goods inventory of A and B items. This scenario dictates a high amount of
Production Control observation and intervention to schedule the assembly cell via
[]
the heijunka box for each shift. Production Control schedules the heijunka box based
upon knowledge of what was removed from finished goods (and now requires
replenishment) and any insight into the special C items that might be required that day.
D
In more stable customer-demand environments and ones where all end items are
held in finished-goods inventory, a different method may be better. Instead of putting
D
instruction kanban in the heijunka box, use withdrawal kanban:

1. From the heijunka box the material handler removes a withdrawal kanban,
D
which is a signal to remove material from a market.
I------,
2. The material handler travels to the finished-goods store. LJ
3. From the store the material handler obtains the proper item.
4. While at the finished-goods store, the material handler also detaches a production D
instruction kanban from the selected item in the store and puts this in a central
kanban post in shipping. (This instruction kanban must first be on the item in
the market for the system to work. It also must go to the assembly area quickly,
I I
either via the material handler or another person to signal a direct
replenishment instruction back to the cell.)
5. The material handler delivers the finished-goods item to the dedicated shipping
lane for preparation for the upcoming customer shipment.
6. The cycle repeats.
D
Using withdrawal kanban in this manner has several advantages:
• Delivery of material to the shipping lane is paced and timed evenly across
the shift.
• Withdrawal of finished goods automatically triggers replenishment build D
instructions for the assembly cell.
• The need for frequent direction and intervention by Production Control is D
greatly lessened because the system self-schedules.

D
D

44
[J There are, however, several difficulties in this scenario:

• All items in the finished-good store must have an instruction kanban on their
r r container.
• The actual ship schedule must be 100% firm in advance of loading the heijunka
box with withdrawal kanban.
• Pacing the delivery of product to the shipping department may not be efficient

r-1 if only one cell is involved.

Please experiment with both types of basic options for combining kanban with the
heijunka box. Variations of each style also exist.
D
[J Pacemaker Summary
The Apogee team now had made several important decisions. They had designated a single
pacemaker point for the value stream-the final assembly cells. They had identified both
the frequency of production based on product mix at assembly (how level) as well as the
[J pitch for delivery of the production instruction. And they had decided to use a heijunka box
scheduling tool in conjunction with a fixed-time withdrawal conveyance route to deliver the
build instructions to the assembly cells. Instead of adopting the lowest possible pitch of
nine minutes for the conveyance route, they temporarily chose a pitch of 18 minutes until
the system was more stable.
[J
During the next few days, Apogee built up the instruction kanban needed, set up the heijunka
boxes, timed the conveyance cycle for delivery, and trialed the system to debug minor
problems. Then they put the system in operation.

D
c--·1
r--1
D

Part 3: Creating the Pacemaker 45

 I'
'--·_J

D
;--,
J
L_,
I

j
L1

D
Creating the Pacemaker-Keysto Success n.
'

• Use the guidelines provided to select the best location

for your pacemaker.
D
• Don't unplug the entire schedule system once you
have identified the pacemaker process-there is a lot 11
~
more work to be done.

• Set a final-assembly lot size after carefully considering ii

work content differences, changeover times, and pitch i..,
intervals. As a rule, lot sizes should not vary much
from average customer order size.

• To help pace the assembly cells to customer demand,

D
link the pitch interval for instruction to the cycle of
convey a nee operations.
n
LJ
• Select a pitch size for the pacemaker that is close to
your capability to react to problems in operations. If
pitch is too short you will not be able to respond
D
within the time period; if it is too long you may not
notice small problems until they are big problems. i\
L_____,
• Create standardized work for conveyance of
information and materials based on the pitch cycle.
D
D
D
n__
[ _)

iiL_J

n
L_.J

n
L_J

46
D
D

r:
L...J

ControllingProduction Upstream
Apogee had made key strides toward their goal of achieving level pull. The Apogee improve-
ment team had determined what items the facility would hold in finished-goods inventory,
,- ··-1 established how finished goods would be replenished, and identified the single point-the
LJ pacemaker-to schedule production.

r· ·1
These decisions coupled with level scheduling of the pacemaker cut overtime and expediting
I1.....-...al
by half. It also improved the ready-to-ship rate from 92% to 97%. But more improvements were
both possible and necessary because Apogee had worked only on improving the information
flow to the assembly cells. The next challenge was to create production pull upstream from
the pacemaker process.
c
Question 7:
[J How will you manage informationand material flow upstream
from the pacemaker?
D As a first step, the improvement team collected data on production at the pacemaker. For
several days the team focused on the cell's ability to deliver finished goods on time to the
finished-goods conveyance operator. They discovered that usually the assembly cells kept
pace with the production instruction on each pitch interval. But several times per shift the
D cells failed to produce on time, and the team created a Pareto chart to list the causes.

[J AssemblyDelays

% 100
D
75

D 50
40%
35%
[J 25 20%
5%
I I

Molded Painted Purchased Minor Categories

parts parts parts downtime

r··-~
I I
'_j
Part 4: Controlling Production Upstream 47
 l I
Ii
LJ
The simple chart showed that the leading cause of delays in assembly was the supply of parts
coming from paint, molding, and purchased parts. Downtime in the cells was only a minor LJ
contributor to delays.
f-]
Why were parts from paint, molding, and purchased parts not getting to the cells on time?
Additional investigation quickly identified the fundamental problem: One end of the plant
-from the shipping dock back to the assembly cells-now was working with a pull system, D
while the other end of the plant-from the receiving dock forward to assembly-still was
working to a push schedule. Push met pull at the final assembly cells with unacceptable results. l_ _J

Apogee's Push/Pull Conflict D

Production Control

[J
D
-;--=-~ \ joxoxJ
r-1
L_-1
~1!!~1!=Z
Celf#1

L__J L__J 0-fJdays 0-7aays

0
0QJQ,(90
[iJ Q) (9 [I] ~
G
D
D
Push Pull

Apogee now needed to regulate production at the upstream processes and to pull products
forward from paint, molding, and purchased-parts storage.
LJ
In an ideal world, upstream production of parts would be conducted one piece at a time at
the same rate as final assembly, and parts would flow directly to the pacemaker without
interruption. But at Apogee this clearly was impossible given the batch nature of the paint
and molding processes and the long distances to the external suppliers of parts. D
D
D
LJ
[]
;-----,
\ i
L... .l
48
\:
\_)

Because the upstream processes had different operating patterns compared to final assembly
cells, significant changeover times, and in the case of purchased parts significant travel distances,
a formal market mechanism was needed to properly regulate flow.

Exterior-Mirror Process Review

0 Long Different
Supplies Batch
exterior-mirror process with distance to shift
D Process value stream? changeovers? next area? pattern?

Paint Yes Yes No Yes

Molding Yes Yes No No

Purchased parts Yes No Yes No

The Apogee team realized that they would need to create controlled markets to buffer
the flow of the three main types of parts upstream from the assembly cells because careful
management of chis inventory would be the only way to guarantee smooth output in the
r. ·--, pacemaker cells. Additionally, a tool known as the withdrawal kanban would be needed
' I
I I to regulate the movement of parts between these markets and the assembly cells.
----'

Based on this decision, Apogee's improvement team proposed to incorporate into the
exterior-mirrors value stream markets for molded, painted, and purchased parts, with pull
loops between each downstream process and the upstream parts markets.

Apogee's concept in implementation was to store a minimum amount of inventory for

each A and B item in the assembly cells and to store the remainder (including the parts
for C items) in the markets.
r· ..·1
' I
LJ

IL.J 1

[J
[J

r --,
,L ;
L_J
Part 4: Controlling Production Upstream 49
 n
i..___J

D
Making Materials Flow Apogee took the following steps to set up the markets, which they

For detailed information on

located together to aid management and minimize the length of D
conveyance routes (see plant layout on page 52).
establishing markets, especially
purchased-part markets-
• A spreadsheet was created listing every part used to assemble
including the Plan for Every
exterior mirrors. This spreadsheet-referred to as a Plan For
Part (PFEP} and examples of
material handling-please refer
Every Part (PFEP)-listed every relevant detail about each part D
to the LEI workbook, Making and its path through Apogee. This helped Apogee confirm that
rl
Materials Flow. The focus the right parts were in the market, as well as estimate size LJ
of this workbook is on the requirements for racks and shelves.
system-level mechanics
required to implement the
• Space was marked off on the production floor for the market in D
level and pull methods that
one central area, and racks and shelving were constructed to hold
make material-handing the material. A precise address system was created (including the
address of the delivery locations at the assembly cells) so that a
D
tasks effective.

•
given part was stored in only one, precisely defined location.
Over the next weekend, Apogee's improvement team moved
D
the material from many locations into the central market location
and tested delivery to the pacemaker cells using a conveyance
n
route with a tugger pulling carts of material.
D
Markets and Point-of-Use Addressing n
LJ
Central Market Area Assembly Cell

I I
D
D
Market address
A2 R3 51
Cell address
C1 S2 R2 D
A2 = Aisle #2 C1 = Cell #1
R3= Row#3
S1 = Shelf #1
$2 = Station #2
=
R2 Rack#2 D
[_]

50
D
D
As noted, Apogee established a basic addressing system for the assembly cells and the
D central markets. This information will be needed for the withdrawal kanban and helps
material handlers identify where exactly to go for delivery and pick up of material.
D
D Locating the Markets

Different options for locating markets should be considered. Traditional Toyota

D practice has been to locate them directly at the end of the producing process. This
helps the process see exactly what and how much is being made of each item.

D Downstream customers {i.e., processes) are expected to come and pull what they
need on a regular basis (either on a fixed-time or fixed-quantity conveyance route).
The producing process then simply replenishes whatever is taken away. This
D method may work well for you as a corrective action if you detect a tendency to
overproduce and push items downstream, provided of course that you have room
to store all parts near the process.
D Facilities that have many internally produced parts, which is the case at Apogee,
or that have layouts that can't accommodate all items at the end of the producing
D process can try centrally located markets. These ideally locate all components in
proximity to both the producing processes and the downstream operation {the final
D assembly cells in Apogee's case).This option may increase overall material-handling
efficiency and minimize the distance traveled on the parts withdrawal loop.

D Alternatively, in plants that do only assembly and bring many parts in from suppliers
on large pallets, a good location is near the receiving dock. This market position
will favor ease of unloading and storage of material into a market and minimize the
D conveyance distance for the receiving team, although it may lengthen the travel
distance to the assembly cells.

D Whichever option you chose, be sure the following rules apply:

• Make clear that the producing process has ownership of the inventory it produces.
D • Ensure that downstream processes pull what they need when they need it on a
regular basis.

D • Ensure there are signals exchanged between the processes regarding what has
been taken away {a topic we will cover shortly}.

Part 4: Controlling Production Upstream 51

 0
D
Updated Plant Layout with Market Area
D
Legend

~ ~ ~ ~ ~ ~ BBBBB ODD AssemblyCell

DD
D
~ ~ ~ ~ ~ ~ BBBBB Receiving
Dock
~

~
InjectionMolding

DieCast
DDDDDDDDDDDDDDDDODDDDDDDDDDDO ~

D
MachiningCenter

Inventory
D
CentratMarkets
1111 Exterior-mirror

~ ~
DODD
DODD
DODD
DODO
DODD
DODD
.....
Painted Molded
Perte
value-streamparts
D
r=i ~ DODD DODD
Purchased
Perte
Empty
Containers

DODD
DODD D
DODD
DODD DODD :·······-·-·
: :
DODD
DODD
Open
Space Cell
#3 Coll
§ D g : D : : D :
Cell
#2 #1
:
• Open
:
•
[]
oo DD i i
DD DD •DD••o o· •Do:
Space
DODD : :
•.
•D D : ;:
DODD
DODD
DODD
D
• Paint
[]
..
DODD ........................
DODD ODDO DODD DODD
Cell#6
Cell #5 Cell #4 .. .
,
..
g D g g D g g D g
oo oo oo oo oo oo : Open :

..
: Space :
DD Do oD Do DD Do :
.
:
D
D D D : --------------········: Shipping
Oock

D DD D DD D D O D D D D D • • • • • • • • Finished·Goods
D
DDDDDDDDODDDDD•••••••• Storage

D
Question 8:
How will you size your marketsand trigger withdrawal pull?
Apogee now needed to implement methods to regulate and pace the flow of information to
the central market and the flow of material from the market to the assembly cells. This will
ensure the right parts are available at the right time. Specifically, Apogee needed to employ
withdrawal kanban. These govern what to deliver to the assembly area from markets and D
support the assembly of product as dictated by the production instruction kanban arriving
at the cells from Production Control.
D
In-process kanban for
Production
instruction [ scheduling flow processes
D
Signal kanban for scheduling
batch processes
Kanban

Parts
withdrawal [ Supplier kanban for
external purposes
fJ
__ , 1

52
n
L_
 [J

Apogee employed interprocess kanban between the assembly cells and the parts market. (The
LJ other type of withdrawal kanban is the supplier kanban, which controls the information loop
from the purchased-parts market back to external suppliers. Apogee will need to deal with the
D supplier-kanban loop eventually, but we will restrict the current exercise to replenishment of
parts within Apogee's facility, as regulated by the interprocess kanban.)
D
Example of Interprocess Withdrawal Kanban

Supplier Product Point-of-use

information area information area area
r··-·
I
·, I
L..J
Supplier code: Part#: Line address:

r-, ABC Station #1

. I
i. Flow Rack#2

Supplier: Part description:

Ajax Springs Spring
Line location:

C1
Market location: Quantity: 40 S2
A2 R2
n
LJ
R7
S1

Apogee needed to complete three tasks to successfully trigger pull from the assembly cells
to the central market:
[J
A. Set a standard amount of inventory of each part to hold at the assembly cells based
,-------,
I .
LJ on the nature and frequency of the conveyance route.
B. Create a separate withdrawal kanban for each container stored at the cells.
C. Determine the right amount of inventory to hold in the central market.

D A. 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.
How much inventory to hold at the assembly cells (called "inventory line side") depends on
how frequently this inventory will be replenished, the minimum pack size for parts, and the
operation of the delivery route. For example, if material is delivered only once per shift, then
a minimum of one shift of inventory needs to be kept at the assembly cells. But with eight or
more hours' worth of material stored line side, the cells will look more like a storage facility
JI
LJ than a production site. In addition, finding the right item will be difficult, floor space will be
wasted, and the large area required for each cell will lengthen the conveyance route.

r··-·,
LJ
Part 4: Controlling Production Upstream 53
 D

Instead, material could be delivered in small containers and at a much quicker rate. The
majority of the 30 parts that Apogee stored line side per cell came in containers holding six
n
L_J

to 25 items (exceptions were fasteners, which have many more items per container). Apogee
theoretically could continually deliver all basic items to the cells as the parts in a single con-
tainer of a given part number were used. One container always would be at the cell, and
D
another would be in circulation with the parts withdrawal conveyance operator.
D
Lean facilities with only a small number of parts and tight assembly areas can adopt this two-
bin style of conveyance. However, with many part numbers and small quantities inside each D
container, two-bin conveyance can lead to excessive material handling due to the need for
ultra-frequent runs by the conveyance operator. I !
Because Apogee has many parts line side and the internally molded items have as few as six
items per container, a conveyance method was needed that was somewhere between the two- D
bin style and the traditional method that placed one shift of inventory at line side.

To achieve this, the Apogee team first needed to determine the nature of the conveyance route
and conducted a number of small conveyance time trials between the cells and the central
market, using mock withdrawal kanban on index cards. D
Withdrawal Material-Handling Loop
D
Exterior-Mirror A,;,;em/Jly Cells
n
L..._j

//······;;.········ -.

/./~// i
B:0 Q)
q -, -~ ·,-,
(9 0::8 ~;~ ~
D
D
Central Market Area ~ :~;-,,
n
LJ
©·······························<ii:=. B:OJ Q) (Q 0:8 ::. ~
~--; ~ Y!
z: @-,~~ ····················--·:

D
P.alnt.uJPart!r

~~¥ · · · · <;
\LJI QJ \DJ -. <.
0 T Q T :
© z: B:0
D
EmptyContalne~ .~ Q) (9 ITJ:8 :,W;, ¢
®· •.
,/ ri, "- c::=J--111
. 'y c::=J--L..'...J
~'"
cJ :!)
'O '5'-c::::J '""
~ L...::.J-c::::J '""
r.,
~

L------~-©._·····-_··-·_····--=..J-~f i_ ~------------------------······-~-------J

54
 For experimentation and learning purposes, Apogee first placed four containers of every item
LJ at line-side for a total of about 120 containers for each exterior-mirror cell. Apogee knew this
was too much inventory for some part numbers and too little for others, but chose this level
for their trial.

Then Apogee placed the mock withdrawal kanban on each of the 120 containers stored line
side at each cell with the basic information necessary to identify the part number (see Example
of Interprocess Withdrawal Kanban on page 53).

Simple kanban posts were erected at each station of the assembly cell and production operators
r·-1 were told to detach the withdrawal kanban each time they began consuming a container of
LJ material such as nuts, bolts, wires, brackets, and housings to build mirrors. The detached
kanban then was placed in the kanban post.

After several trials between one of the assembly cells and the central markets, Apogee
determined the following time and work standards for the material-handling route:
D
1-·-·· Standard Work Flow and limes for Withdrawal Material-Handling Loop
I
L__;
I

Conveyance
Step Standard work elements time estimate

Cell operator withdraws kanban card from a container upon NIA

removing the first part out of the container. The withdrawal
D kanban cards collect inside a kanban post at each cell station.

The material handler travels from the central market and 2min.
1.
arrives at assembly cells.

The material handler visits and picks up all the withdrawal 4min.
kanban at each station of each assembly cell as well as any
2. empty containers, delivers parts to cells, and returns to the
central market.

D 3.
Once back at the store, the material handler sorts the cards
for the best pick order.
15sec.

Empty containers picked up on the route are dropped off in 1 min.

4.
the specified location in the market.
Any necessary bins for small items, such as nuts, bolts, screws, 1 min.
5.
washers, etc., are obtained.

Parts are picked from the store and placed on the 3min.
r- -; 6. material-handling cart (an average of 10 withdrawal items each 20sec.
I I at 20 sec. per pick).
LJ
The withdrawal kanban cards are placed into their 15sec.
7.
D respective containers.
11mln.
Total route time
50sec.

Part 4: Controlling Production Upstream 55

 n
~_J

Apogee calculated that on average 40 withdrawal kanban cards would be triggered per hour
of production by the two assembly cells for exterior mirrors. The actual travel time between [J
the market and cells was only four minutes. This meant that each conveyance and delivery
cycle would contain the four minutes of travel time plus a variable amount of work represented
by the number of the kanban cards per trip.
D
But how frequently should the withdrawal material-handling loop be operated? The Apogee D
team decided to start with a 15-minute interval to gain experience with the new system.
n
LJ
They calculated that about 10 containers would be required by the two cells every 15 minutes
(40 withdrawals per hour divided by four loop operations per hour) and calculated that this ,---,
would require seven minutes and 50 seconds of work in addition to the four minutes of travel LJ
time between the parts market and the cells. This meant a total of 11 minutes and 50 seconds
of work on each loop. Trial runs quickly showed that the conveyance operator using a tugger D
to deliver parts to both cells could easily manage this workload. (Eventually, an additional cell
can be added to this loop to use the three minutes and 10 seconds of extra time.)
D
Every time a set of withdrawal kanban was picked up from the cell (approximately 10 kanban),
the material was replenished on the next delivery cycle. With a 15-minute route, Apogee D
could store as little as 30 minutes of inventory at the assembly cells-far less than they had
in the past. But because Apogee did not want to risk shutting down the assembly cells while
the withdrawal pull system was being implemented and refined, they conservatively chose
to hold one hour of material at the assembly cells and reduce this amount over time.

Next Apogee determined what one hour of material equated to for every part number at the
cell. For some items it was one container (e.g., fasteners) and for others it was several containers. D
Flow rack lengths and positions at the cells then were adjusted and finalized based upon
this decision.

D
11
l. J

n
L._J

LJ
n
L____J

56
 Two Types of Conveyance Routes
So far we have examined one type of conveyance route- fixed-time variable-quantity.
With this method the conveyance operator performs a precise sequence of steps

D standardized by time, much like standardized work for an assembly cell operator.
The time for the route is fixed, but the amount of material moved during a given
withdrawal interval will vary based on the previous usage by the production area.
D In contrast to fixed-time variable-quantity conveyance, fixed-quantity variable-time
.,. ----"-----, conveyance uses a quantity-based trigger to signal the need for material movement
IL__JI between locations. The time to deliver the quantities fluctuates as needed, as does
the route traveled. Since periodic route times cannot be created with this method,
the key to success is a clear visual and/or audio signal that identifies when material
D needs to be moved to a production process.

Lean operations often need both styles of conveyance. One size does not fit all.
When the goal is to frequently move material to a production area in conjunction
with a pitch interval (as at Apogee}, fixed-time conveyance is best. When intervals

D are highly infrequent or parts are exceptionally heavy or cumbersome, such as

automotive windshields, fixed-quantity conveyance may be better.

B. Create a separate withdrawal kanban for each container stored at the cells.
D The Apogee team now was ready to create formal interprocess withdrawal kanban for each
container at the assembly cells in the exterior-mirror value stream. The system can work only
[J if there is a physical withdrawal kanban for each item stored line-side at the assembly cells
(see Example of Parts Interprocess Withdrawal Kanban on page 53).
D Each kanban must contain address information on the storage point for the part in the central
market and address information for where material will be delivered in the production area.
Apogee had already completed the former task, and now transferred both types of information
to withdrawal kanban.
D
Apogee's exterior-mirror parts withdrawal loop consisted of 30 different part numbers with

D one to six containers stored line side for each, representing one hour of inventory per item.
The total number of withdrawal kanban in the loop for the exterior mirror cells was 110 cards.
(Part 5 will show how Apogee can reduce the number of cards and containers in circulation as
D improvements occur.) For control and management purposes, Apogee incorporated this infor-
mation into its parts-tracking spreadsheet. (Often companies use their PFEP for this purpose.)

Part 4: Controlling Production Upstream 57

 lJ

C. Determine the right amount of inventoryto hold in the central markets.

The final issue for the Apogee team to resolve for effective management of the central market D
was the sizing of inventory at the markets. As parts were moved from many locations in the
plant into the central markets, the Apogee team discovered chat the plant had anywhere from D
one day to two months' consumption of different part numbers! They obviously needed a more
precise rule for setting inventory levels in the markets.
D
The basic rule they developed was similar to the formula used for finished goods on pages 20
and 21. The demand variation, though, was much lower than with finished-goods inventories D
because external demand variation from the customer already was being absorbed by the buffer
stock in the finished-goods market. (This is how adding to inventories at one point in a facility- D
in this case in finished goods-can reduce total inventories in the facility.)

Apogee also had leveled the assembly cells, which reduced internal variation during a normal
replenishment cycle. Because of this, a lower buffer-stock factor could be assumed since this
variation also would be absorbed by the safety inventory in finished goods. D
Market InventoryCalculation D
Average daily demand x Lead time to replenish (days)* Cycle stock
D
+ Demand variation as% of Cycle stock Buffer stock

+ Safety factor as% of (Cycle stock+ Buffer stock) Safety stock

D
= Market inventory

D
Market InventoryCalculation for Part #14117 (Painted Bracket)
D
100 x 5* Cycle stock 500
+ 10%** of 500 Buffer stock 50
+
=
10%*** of (500 + 50) Safety stock

Market inventory
55

605
D
*
**
Lead time for paint process to replenish this item
Two standard deviations of internal variation for this part number
D
*** Average scrap rate and downtime for this part number

D
r-----i
. l
L-----'

[J

58
,··1
·i-_J

Apogee used this calculation and logic for all items in the central market. At the end of the
first full week of running the exterior-mirror cells with the parts withdrawal kanban route,
Apogee's ready-to-assemble performance for this value stream improved from 75% to 91 %.

Apogee Ready-to-Assemble performance

% on time %
Target Gap assessment
100
D 91% t Ga
75% 75%
-
LJ
20%
5%

LJ Previous Current Molded Parts for Minor

parts C items downtime

Within the first week of trial operations, the withdrawal kanban loop largely eliminated part
shortages at the exterior-mirror assembly cells, which ran overtime on only one day and for
only 30 minutes. Apogee now was shipping and delivering to the customer with 100% on-time
performance. The remaining problems in the exterior-mirrors value streams were related to
getting a few internal parts to the central markets exactly when they were needed and in the
LJ right amount, as indicated by withdrawal kanban. Although still short of their final goal, the
Apogee improvement team took a moment to celebrate their success!

A Pareto analysis (as shown above), based on data collected during the week by the parts-
r·--·. ···1
withdrawal material handler, provided insight on the remaining problems. The most important
LJ insight was that molded parts still were not always in the market when needed by the parts-
,·- .. l withdrawal material-handling loop and, thus, were not ready for assembly. The assembly cell
I i
' '
'------" now was pulling from the market based upon a fixed-time withdrawal cycle, but the flow of
material into the market from the fabrication processes still was regulated by a push schedule.
r- · ·1 The conflict between push and pull had migrated once more and now was focused on the
batch processes.
-- l
I '
LJ

LJ
Part 4: Controlling Production Upstream 59
 r-;_
L _ _j

Managing the Flow for C Items

When creating the central parts market, the Apogee team faced an important choice
n
L_.J
with regard to parts for C items. The team already had decided that no inventories of
C items would be maintained in the finished-goods store. Now the team needed to
decide whether to build C items to order from the molding process forward or hold
all of the part numbers needed for C items in the central market and assemble the ,---,
items when an order came in. LJ
Apogee realized the latter option would be much easier to manage-particularly at
this stage in the implementation of a level pull system-than manufacturing all of the
needed parts to order each time. Apogee therefore purchased a month's worth of the
purchased parts for C items and molded and painted a month's usage of the internal
.-,
parts. This was a much higher level of parts inventory, measured in days of usage, LJ
than for the A or B items, but the actual number of pieces (and dollars of inventory)
was not that large {e.g., generally fewer than 100 to 200 pieces per part number}. D
For your situation you should consider all possible solutions regarding production
of C items. If there is enough volume and associated business margins, the C items
may belong in their own value stream or dedicated assembly cell.
D
If, however, C items must be made using the same set of production assets as A and [J
B items, you need to decide between two options: Either make each C item completely
from the first process in the value stream on through to finished goods or make each
C item from some midpoint in the value stream where you hold partially completed
C items. This latter option is what Apogee chose to follow, and, as long as it does not
result in a large amount of slow-moving inventory, is an acceptable solution.

Question 9:
D
How will you control batch processes upstream from the market?
The last question Apogee needed to answer to complete their new production control system
was scheduling of the processes producing batches of parts upstream from the central markets.
As shown previously in the Pareto chart, the molding process now was the largest source of
problems in terms of supplying material on time to the assembly cells for exterior mirrors.
n
I I
Therefore, the team decided to start with these parts. -~-'

LJ
n
~-J

D
60
 Apogee Pull Against Push

D Batch processes Instruction and withdrawal

pushing into markets pu//loops

c Parts wtthdrewet
material-handling loop
Production
instruction loop
Haijunka box

!oxoxl
Empty-ContainerArea

DD DD DD
DD DD DD
DD DD DD
D DD DD
!JcolOrs
5daye; QJ
Finished-GoodsStoreArea
[I] Q) Qi (9 0
D 0 Q) (9 [I]
DD DD DD
W/Pth,-. DD DD DD
D Purchased-Parts
Ce/1#2
DD
DD
DD
DD
DD
DD
Market 0 DD DD
[I] Q) \(Ji© 0 D DD
D QJQ) ©ill
Finished goods 1-2 days
W/Plhr.

r-1
I :
L-1 For scheduling batch processes lean thinkers employ a set of tools known as signal kanban.
The common feature of these tools is that they aggregate the small increments of demand
communicated by withdrawal kanban-which arrive at the central market from continuous
processes downstream (final assembly in Apogee's case)-inco larger batches of demand
corresponding to feasible run lengths for the batch processes.

D Types of Kanban

In-process kanban for

[J Production
instruction [ scheduling flow processes

Kanban
Interprocess kanban
for internal purposes
Parts [
withdrawal
Supplier kanban for

D external purposes

[J

D Part 4: Controlling Production Upstream 61

 0
[]
For Apogee's molded parts, the team decided to use a type of signal kanban known as triangle
kanban (two other options are described at the end of this Part). D
Apogee batch sizes in the molding department were a minimum of 2,000 units (and often far
higher). With average cycle time across all the machines of 40 seconds and only one part pro-
D
duced on each cycle, a "small" run of a part number for Apogee still required more than 22
hours or close to three shifts.

Four machines in Apogee's molding department were dedicated to the value stream for exterior
mirrors. Three of the machines made the visible exterior components, including the base and
D
the mount, which were painted as well. The fourth machine produced three molded but
unpainted items for both right- and left-hand mirrors that were inside the mirror housings and D
invisible to the customer. For simplicity, Apogee decided to start signal kanban implemen-
tation with this machine and its three part numbers. []
Apogee's Production Control Department previously scheduled molding by using an economic
order-quantity (EOQ) model to determine the efficient lot size to run in production. As the
D
team investigated the actual run time and changeover patterns, however, they found that
supervisors ran most part numbers longer than called for by the EOQ-based schedule. LJ
They were measured on machine utilization, and the best way to increase utilization was
r-:
to do as few changeovers as possible. L__ J
To change the molding-machine operating paradigm the Apogee team eliminated the 1----i
traditional efficiency measurements for the trial period and told operators simply to make
LJ
parts in the amount and sequence called for by the new triangle kanban. This subtle but
important change significantly reduced the incentive for the supervisors to encourage
nJ
l __

overbuilding of molded product.

D
Production ControlUsing Triangle Kanban

Batch process
D
machine running
part #15455 Signal kanban rail Inventory D
J t7
Part #15456

Trigger D
point 7
®---- Part #15454 Part #15455 D
11
L_J

62
I'~--J
0
f-··1
With the triangle (signal) kanban, the removal of inventory from the market triggers
D replenishment through the following process:

1. Inventory is picked from the storage location and taken to assembly.

2. Once a predefined trigger level is reached in the central market (after a number of
withdrawal kanban and containers of material have been picked) the signal kanban for
D the part number is removed from the market and placed on a rail in front of the molding
machine. (Note that there is only one triangle kanban in the system per part number.)
3. The operator follows the instructions on the signal kanban and builds precisely the
required lot size indicated on the triangle kanban, which is the amount needed to
replenish inventory in the market.

[J To implement this type of system, Apogee had to determine the maximum amount of inventory
to have in the market (the lot size) and the level of inventory that triggers the release of the
~-·-i triangle kanban for conveyance to molding. An example of a triangle kanban is shown below.
I I

LJ

r·, Example of Triangle Kanban

I
\.....__)
I

Part# Part description Market location

D 15897 Internal bracket A2 R6 S2

Date triggered* Lot size

Nov.23 3p.m. 400

[J
Trigger point
200

D Tool#
14

[J

* The point in time at which the kanban is placed on the signal kanban rail. This aids production management in assessing
status of the process and reduces the possibility of getting kanban out of sequence.

,·-·-,
I '
LJ
Part 4: Controlling Production Upstream 63
 (--.
I
c__ _ _J
l
The improvement team started the conversion process by gathering basic information on
customer demand and the #4 injection-molding machine. D
Injection Molding Example-Machine #4 Data l_J

Average
demand* Cycle time Average
Required
run time
Average
changeover
n
--
Part# per day (pieces) per piece scrap rate per day** time
Ii
15897 400 20 sec. 1.5% 136 min. 55min. LJ
15898 600 22 sec. 1.3% 223 min. 55min.
15899 1,000 20 sec. 1.5% 339 min. 55min.
[]
Total 2,000 698 min.

* To make 1,000 left-hand mirrors and 1,000 right-hand mirrors in assembly cells, as dictated by customer demand
D
** Adjusted for average scrap rate and rounded up to nearest whole number

The team then took the following four steps to implement the triangle kanban pull signals:
ii
LJ
Step 1: Determine time available for changeover work.
Step 2: Set the number of changeovers per day.
D
Step 3: Determine production lot size.

Step 4: Specify a trigger point for reorder. n

L__;

Step 1: Determine time available for changeover work. D

lime Available for Changeover Work

Total 1-shift production time available (net breaks and lunch) 450 min.
Number of shifts x 2

Time available for production on 1 machine during 1 day = 900 min.

n
L. ..J
Time required per day to meet average demand* - 698 min.

*
Net time available for set up and changeovers per day

From Machine #4 Data

= 202 min.
n
L____.j

Ii
On an average day, with a volume of 2,000 total pieces needed from machine #4, there are LJ
202 minutes available for nonproduction activities, including set-up and changeover work.
D
\\
L__ J

64
 r·1
'-'·
Step 2: Set the numberof changeovers per day.
With an estimate of che time available for non production work, the team could establish a
target for the number of changeovers per day.
r·--, I
;
~
Possible Changeovers per Day (Two Shifts)

l--··1
:...__;.
Nonproduction time available 202 min.
Average downtime per day (not including set-up and changeover - 30 min.
r--~
1 !
times; each shift had 15 min. downtime)
j I
L..__;
Time available for changeover work on 1 machine during 1 day = 172 min.
Average changeover time -r- 55 min .
[J Possible number of changeovers per day = 3*

r- 1 * Rounded down to nearest whole number from 3.13

The team subtracted the average downtime per day to determine how much time was available
[] for changeovers, and then reversed Apogee's long-standing management practice: Instead of
seeking to minimize changeovers, the Apogee team determined the maximum number of
changeovers permitted by the time available and set the number of changeovers per day at
this number (three changeovers).

Step 3: Determine productionlot size.

Given that Apogee would now execute three changeovers per day, the team needed to set
the correct lot sizes (the amount of parts called for by the triangle kanban) and the timing
of changeovers.

There are two basic ways in lean manufacturing to set lot sizes with triangle kanban:
r ··:
I I
L-.J
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
D 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
D (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.)
D The other option calls for running the same (fixed) quantity of parts for each part number
r-··, run, producing parts for varying intervals of demand for each part number. For example, a
LJ process might run every part number for three hours before changing over to the next part

D Part 4: Controlling Production Upstream 65

 n
I _ __J '
L.

L !
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 l~-I
running some part numbers more frequently than others.)

The first option builds lot sizes more closely linked to customer demand and leads co lower
D
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 D
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 n
l...____J

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. n
L___J

Fixed-Time Lot Sizes and Changeovers

D
One day

Part#15899 Part#15898 Part#15897

1,000 pieces
339-min. run time
55 min. 600 pieces
223-min. run time
55 min. 400 pieces
136-min. run time
55 min.
n
L

t
Fixed-time method
D
D Changeovers (e.g., each lot size= 1 day of demand) !I
L __J

For molding machine #4, with three products and three changeovers each day, the production D
lot size can simply be set at one day's average demand (three products -;- three changeovers
= one day of demand).
D
Note that with more frequent changeovers the lot size called for on the triangle kanban naturally
becomes smaller. For example, if six changeovers per day were possible, as changeover times D
are reduced then only one-half day of inventory would be needed per part number.
D
Lot Sizes with Fixed-Time Variable-Quantity Method
n
L___,
Demand per day
Part# Lot size (pieces)
15897 1 day
x
x 400
=
=
Lot size (pieces)
400
D
15898 1day x 600 ;;;;;
600
15899 1 day x 1,000 = 1,000 D
n
LJ

66
0
 Because each part is stored in containers of 10 parts and each pare number is being produced
to meet consumer demand every day, Apogee will need space for 200 containers in the market
to hold the 2,000 pieces of total inventory for the three required part numbers. Previously,
[] when Apogee was producing large lots with infrequent changeovers to obtain high machine
utilization, there often were 5,000 to 10,000 pieces per pare number (5 to 10 days) of these
[J items at various storage locations in the facility, yet these still were not consistently delivered
on time to assembly!

D With lot sizes determined, the operator of the molding machine will know precisely how many
parts to make each time a signal kanban for a lot of parts arrives at molding. This leaves as
D the only remaining question, "How do you set the trigger point in the central store for sending
the signal kanban to the molding machine?"

D Step 4: Specify a trigger point for reorder.

D Setting of the trigger level for inventory replenishment with a signal kanban is simple, but it
does require several steps: You need to determine the elapsed time necessary for the molding
machine to make a standard batch of the product (including changeover time), add the time
[J needed to deliver the first container of parts to the market, and then divide by takt time. (Adding
a buffer for external-demand variation is not necessary because this is now covered by the buffer
D stock in finished-goods inventory and the internal schedule is level.)

For machine #4, the time to replenish once the triangle is triggered can be estimated as follows:
[J
Lead lime to Replenish

D Changeover 1st container Total lead time

Part# Run time* + time + time** = to replenish

15897 136 min. + 55 min. + 10 min. = 201 min.

15898 223 min. + 55 min. + 10 min. = 288 min.

D 15899 339 min. + 55 min. + 10 min. = 404 min.

* Run time equals the part cycle time multiplied by the lot size.
** Assume 10 min. to make enough parts to fill one container and to convey it to the market for use.

D Of course, each of these individual replenishment times assumes that no other product is
running in the machine or waiting to run, and this is clearly not realistic. With three part
numbers running on machine #4, it is normal that one part number will be running, one part
D number will be waiting on the signal rail to run, and one part number will be in the central
market not yet triggered. It is statistically unlikely that all three part numbers ever will be
triggered at once because there is always some lag between them. Therefore, Apogee assumed

Part 4: Controlling Production Upstream 67

 D
D
that only one triangle kanban will be in front of another when it is triggered. For an extra
margin of safety, Apogee assumed chat the triangle with the highest demand for molded items D
(and the longest run time) is at the front of the triangle kanban rail and set to run.

The longest replenishment time for machine #4 therefore is 404 minutes, since that is the
D
longest lead time possible before replenishment of the next part number can begin. Of
course, this extreme assumption is appropriate only for two of the three part numbers-the LJ
calculation for the third part number can use the second-longest lead time, since a part
number cannot be in front of itself. D
Part Takt lime Calculation D
Time
available Demand
D
Part# per day x Seconds = -'- per day = Part takt time
15897 900 min. x 60 = 54,000
sec.
+ 400 ;;;;;
135 sec.
(2.25 min.)
D
15898 900 min. x 60 = 54,000
sec.
-'- 600 = 90 sec.
(1.5 min.) D
15899 900 min. x 60 = 54,000 -'- 1,000 = 54 sec.
sec. (0.9min.) D
Trigger Point Calculation
[]

Longest Trigger point** D

Part# lead time* + Part takt time = (number of pieces left in market)
,--,
15897 404 min. + 2.25 min. = 180 pieces LJ
15898 404 min. + 1.5 min. = 270 pieces
15899 288 min. + 0.9 min. = 320 pieces n
* The longest possible lead time for another product that theoretically could have entered the run queue for that process in
front of the required product
** Rounded up to nearest whole number of 10
[l

For each of the three part numbers running through molding machine #4, Apogee was able co D
calculate a new lot size as well as the signal point at which they should start replenishment for
each item. For part #15897 the market would start with 400 pieces of inventory for this item []
(40 containers). The triangle kanban trigger point would be placed at 180 pieces or 18 containers
of product. When the inventory decreased to this level, the kanban would be moved to the
injection-molding machine to act as a signal for replenishment. The replenishment calculation
n
1 . J

n
I
,..__ .:I

68
11
LJ

D
shows chat molding can produce the produce and return it co the l\lotes on Trigger Points
D market just before the remainder of the inventory in the market is
BatchP,rocess running many

.. consumed.

Summary of lot size and signal pointsfor

part numbers: If Apogee had
needed to run more part num-
bers through machine #4, then
the lead time to replenish any
signal kanban
given partwould need to be
Using these formulas and rules, Apogee established the new required sethiqher to compensate for
inventory quantities for machine #4 and then for each of the other the other parts numbers that
molding machines in the value stream that supplies parts to the central may be in front and waiting

D market for exterior mirrors. to be run. In general, no more

than half the kanban ever will
For each part number in the value stream, Apogee created one signal be in front when the produc-
D kanban and also reorganized the material in the market. As inventory
tion cycle is running in normal
condltions, A conservative
was taken away from the central market by the withdrawal macerial-
estimate for the number of
handling loop, it triggered replenishment build instructions for the kanban on the rail would be
molding machines. to divide the number of signal
r····1 kanbanfor that asset by two
The improvement results were drastic: from 50% to 90% of the and then subtract one kanban:
previously existing material in the market could be eliminated. More =
(total kanban + 2)-1 kanban
[J importantly, the replenishment time in every department along the on the rail.
value stream now could be measured in hours instead of days as
D before. Apogee realized chat this was a major breakthrough in terms Unstable production: In a less
stable production environment,
of enhancing the production system's capability to deal with changes
you may want to establish an
D in demand and the ability to supply products on time in the future. earlier trigger point. The Apogee
explanation is a lower limit
Looking ahead, Apogee will be able to further reduce replenishment
calculation that could cause
D time in molding if changeover times can be reduced from the current the market to run down to the
average of one hour. Indeed, if set-up times can be taken down to final container of product in
D single minutes, as many molders are now doing, it should be possible some rare instances when the
full lead time to replenish the
to add a fourth changeover on machine #4 and cue lot sizes by
,·
I
1
; another third. previous item is required.
L.J

,··-·,
i ;
L_J

D
69
 ) I

Other Choices in Signal Kanban for Batch Processes D

Triangle kanban is the standard method used in lean manufacturing to schedule a
batch production process. However, it is only one type of signal kanban. The other
common types are pattern production and lot making. Usage depends upon your
D
individual circumstances, capability, and the type of process you are operating.

Pattern Production: Pattern production, as the name implies, creates a fixed sequence
D
or pattern of production that is continually repeated. For example, in an eight-hour
cycle part numbers are always run A through F. The difficulty of your changeovers (for D
example from light to dark paint or from light- to heavy-gauge steel) may dictate this
order for your processes. Inventory in the central market is a function of the length
of the pattern-replenishment cycle. A one-day pattern implies one day of inventory D
must be kept in the market. A one-week pattern requires one week of inventory.

There are several advantages to pattern production. It is simple in that every part is D
made every cycle (which you must determine based upon average demand}, and the
changeover sequence is standard and predictable. Companies with high downtime n
l_J
and other problems may find this a much easier way to begin than with the triangle
kanban. However, as with any tool, the pattern and quantity run must be analyzed
and updated over time to best match production. D
The main disadvantage of pattern production is that the sequence is fixed. You can't
jump from making part C to part F. Instead, you must finish the sequence. Also there
is little incentive to improve changeover time and reduce inventory. The pattern just
keeps repeating to replenish material in the market location.
D
Apogee utilized a form of pattern production to schedule its paint department. A color
sequence on a rotating wheel was used to indicate what color to paint next. The basic
paint sequence was set for the whole week, for the 50 colors required each week. One LJ
week's worth of material for each part number was kept in the central market. The
wheel also contained several planned open spaces that were reserved for items such
as infrequent service part items and conducting planned maintenance tasks.
D
As the sequenced color came up to run, the paint supervisor confirmed the amount
of inventory in the central market that had been consumed during the previous week
D
and scheduled this amount. Although a visual system, the method was highly effective.
If you have the capability, you can use instruction kanban and attach them to each D
container of painted items in the central market. As the items are consumed during
the week, the instruction kanban are detached, collected, and taken back to the paint
shop. When the sequence indicates it is time for the color to be run, the cards are
D
counted to determine the paint quantity for that cycle.

Lot Making: Lot making is another effective method, although a more difficult way
D
of scheduling batch processes. Unless you are already adept at the two previous
forms of signal kanban or have no other alternative, I suggest you avoid it. D
11
L_,

70
D
r-··-1 With lot-making methods, a physical kanban is created for every container of parts
that is controlled in the system. This can equate to hundreds of cards depending
upon on how many part numbers and containers are in the central market. As material
D is consumed from the market, the kanban are periodically detached and brought back
to the producing process and displayed on a batch board that highlights all part
numbers and displays an outlined shadow space for each of the kanban cards in
the system.
,-.
I
.,;
L__j
Lot-Making Batch Board

Part #1 Part #2 Part #3 Part #4 Part #5 Part #6

Market full

Triggered! Trigger point

(can vary

D Next?

• Kanban cards returned from central market indicating product consumed

D Empty slots

A returned kanban card placed on the board in the shadow space indicates inventory
has been consumed in the market; unreturned cards represent inventory still in the
market. As predefined trigger points are reached, the production operator knows to
begin making product to replenish the material in the market.

The advantage of the lot-making method is that it allows information to come back
to the production process more often. It signals what actually has been taken away
and uses smaller increments than with the signal kanban. It also provides a more
visual representation of inventory consumption and highlights emerging problems
in the central market.

The difficulties of this system are that it requires many kanban cards that must be
brought back in a timely and reliable manner if the batch board is going to reflect an
accurate status of inventory in the market. Additionally, it takes resolve on the part
of schedulers and supervisors not to look ahead, see what might be required next,
and then start building it in advance of what actually is needed. Looking ahead is
always a double-edged sword because you may avoid running low on inventory in
[J some instances but are likely to overproduce in many others.

D
Part 4: Controlling Production Upstream 71
 (!
r t
L_._J

I!
l __J

n
L_j

n
L__J

Controlling Production Upstream -

Keys to Success
• Determine how you will install pull production n
L_____J
upstream from the pacemaker.

• Consider carefully whether you need a

fixed-time conveyance cycle or a fixed-quantity
conveyance cycle.
ji
L J
• Standardize the role of the material handler.

• Organize inventory in a central market location

and size the inventory carefully. D
• Use withdrawal kanban to pull inventory from
the market to final assembly. D
• Use signal kanban, such as the triangle kanban,
to control batch processes upstream from the
market.
LJ
D
n
L__J

I l

\I
LJ
ii
L.J
r---i
LJ
[l

72
 c------,
I I
L_J

c Expanding the System

D Apogee had now converted their exterior-mirror value stream co a level pull system. From
customer demand through assembly, painting, and molding, the pace was being driven by
take time with material replenished just in time.

Information and Material Flow for Exterior Mirrors

'r
LJ Parts instruction by
signal k1nban
Parts instruction by
pattern production
15·min. parts
withdrawal
18-mln. productron
lnstn.tc:f;ion lcep
Heljunh box

.----o • loxox!
(tri•ngtel meteriet·hendHng loop for finished goods

D :------. ~ - ---:,--------:

~G LJ~G~
' Empty-Cont,,lner Area
:
DO DO DO
D Ce/1#1
DO DO DO
DD DD DD
D DD DD

LJ ......... ····~ sz: 0QJ-Q,©0

0
Flni5hed-Goode StQre Area

-----------. :-----v----~ : ' [:]QI

WIP1hr.
©0 OD
DD
DD
DD
DD
DD
DO
DO
DD

r----1
'E1Jjv 0
09-Q,©0
Cel/#2
DD DD

DD
DD
DD
DD
DD
DD
DD
DD DD
_ -------~ ~parts
Iday
[:]qi IP 0 DD DD
DD
WJP1hr. OD
D Purcha5ed-Parts , .,._ .. I
DD DD
DD DD
DD
DD
DD
Market DD

LJ 20p;,~s
DD
DD
DD
DD
DD
DD
DD
DD
r····, 5day
'~ .... - ......... - .
LJ Central Markets

Multiple benefits to the value stream now were evident. On-time delivery to shipping and
on-time delivery to customers were stable at 100%. The assembly cells were ready to assemble
[J product 98% of the time-only isolated instances of minor downtime in upstream processes
had prevented the cells from having a perfect week. For the first time in the collective memory
of the improvement team, all the equipment in the value stream had run without overtime, and
[J
no expediting of material had occurred during the week. Dramatic increases in available space
,-., and dramatic reductions in inventory also were apparent: Inventory had been reduced by 70%
\,I__J! for work-in-process and 70% for finished goods since the start of the pilot.

Part 5: Expanding the System 73

 I(
I \
._____J

n
L....J
Box Score-Exterior Mirrors
D
Start of Current
level pull state Comment
Productivity

Direct labor (pieces per person per hr.) 11.0 12.5 14% improvement - reduced overtime
Material handlers 4 2 50% improvement - but low
[]
utilization
Quality []
Scrap 2% 1.5% Rapiddiscovery of downstream
problems
Rework 15% 12% Rapiddiscovery of downstream
problems
External (ppm) 105 105 No change - not addressed I I
Downtime

Assembly (min. per shift) 20min. 10min. Reduced waiting for material [J
Paint (min. per shift) 15min. 15min. Reduced waiting for material
Molding (min. per shift) 10min. 10min. Reduced waiting for material
Inventory turns
I!
Total 12 24 Estimated 2 weeks total on hand L_J
On-time delivery

To assembly 75% 98% With no overtime or expedites

D
To shipping 85% 100% With no overtime or expedites
To customer
Door-to-door lead time
100% 100% With no overtime or expedites
0
Processing time (min.)
Production lead time (days)
125.7
30
125.7
12
No change
Order throughput time of 2 days
D
Costs

Overtime costs per week $5,000 $0 100% reduction

D
Expedite costs per week $2.000 $0 100% reduction n
LJ

Productivity benefits now were beginning to surface as well due to level production and
n
LJ
adherence to a consistent takt time. With overtime eliminated, direct labor was averaging a
13.6% productivity improvement (pieces per person per hour) for the most recent week. n
Small gains in quality, although not a performance area specifically targeted during the
implementation, were made as well. Compared to the past when it sometimes took several l !
days or weeks for a defect to be detected, downstream problems were being surfaced much
faster than before. Although the root causes were not solved by the level pull system, the n
L_ __J
symptoms were at least being detected much earlier resulting in fewer problems.

74
 Indirect labor productivity, in the form of material handling, was improving as well because
the exterior-mirror value stream now was serviced by a single conveyance operator who handled
the production instruction loop and a single material handler running the parts withdrawal
loop. This was despite their underutilization (30-70%) while serving only the exterior-mirror
value stream, a problem Apogee will tackle next.

Question 10:
D How will you expand your level pull system across the facility?
These improvements, although impressive, had been accomplished only on one value stream
D in the facility. The Apogee team now was at a transition point in their implementation and
needed to think about expanding the improvements across all the value streams in the plant

0 to capture the full benefits of system kaizen.

Apogee Current-StateLayout
D
Legend

D ~ ~ ~ ~ ~ ~
Receiving
DDDOO AssomblyColl

~ r=i ~ ~ ~ ~ Ooclc
~

~
Injection Moldlng

OleCast
DDDDDDDDDDDDDD <, <, ~ Machining Center
<, <, ~ Inventory
,.,.,,,,, Central Market,
- velue-etreem

D ~~
DODD
DODD
ODDO
DODD
Painted
Patt,
Molded
Perts D
inventory

Value-stream eree

~EJ DODD DODD Purchased Empty

Pan, Con1aTner,
DD DD , . DODD DODD
DODD j
DODD:
DD DD i DODD -> ~-··········
i :
D DD DD :
DODD i
Open
Space Cell #3

B D 8
DO OD
:,
! Open
i~
i
: Space

DODD i
I I
DD DD ! i
/1 DD DD : . D t J
L...J DODD ......._
Paint
DODD
DODD
0 DODD DODD DODD DODD
Cell #6
DD D 0 D D 00 0D D 00 Coll #5 Coll #4
I
I
I
I
I
i• Open
I
I
I
I
I
i
DD DD Bo oa DD oo
0
Space I

DD DD DD D OD Do
I I
0 I I
I I

-
I I
I I

•----------------.
---
D D D I I

Shipping
Dock

DDDDDDDDDDDDDD•••••••• Finished·Goods
Stora.i&
DDDDDDDDDDDDDD••••••••

r----,
I I
L_j
Part 5: Expanding the System 75
 ~
i I
L______;

,~ I
L~
Project team members expressed varying opinions on how to move forward. One group
wanted to continue to expand the system by value stream-a value-stream rollout. Since []
the exterior-mirror value stream improvement effort had taken approximately eight weeks to
date and much of the elapsed time was learning the logic of the new system, they assumed
the remaining value streams could be accomplished in no more than four months (for a
maximum implementation time of six months).
0
Other members of the team instead proposed a departmental rollout, converting all the batch
processes, then attacking all the remaining assembly cells and finished-goods areas at once, n
L._1
and finally targeting the central markets. Both approaches have their merits, as summarized
in the table below.
D
Apogee Decision- Expansion Options
D
Option Advantages Disadvantages Comments D
Value-stream • Pattern already • Five more cells • Not right at
approach developed • this time
• Implementation timing
already known
Lengthy rollout
process 0
• Difficult to partially
• Proven implementation schedule shared
results assets and may
cause much
D
confusion during
transition period
D
Departmental • Solves scheduling • Revised plan needed • Best fit for Apogee's
approach

•
problem at shared
assets
Quicker resolution of
•
•
Timing uncertain
Less experience
current situation
n
main problem areas
• Capture of
cross-value-stream
D
efficiencies such
as material handling
D
Apogee realized that the value-stream approach normally worked best when all the assets
in the plant could be neatly divided and dedicated to individual value streams: decisions
D
and action items were clear because they did not affect other parts of the plant.
n
L___J
Unfortunately, many of the processes at Apogee used shared resources. All value streams
went through paint. Die cast supported two value streams. Many molding machines were
shared. Only the assembly cells and the four molding machines for exterior mirrors were
completely dedicated to specific value streams. Scheduling shared assets always had been !I
l_J
.---,
L_J
r--,

L_J
76
D
at the root of difficulties when push-scheduling the entire plant, and converting one area to
D level pull would set off a chain reaction that had implications for other areas.

As a consequence, Apogee decided to approach the rollout departmentally. The team would
I ! address all of molding together, all of paint and die-cast together, all purchased-parts together,
and all assembly cells together-in that order. Apogee chose this approach because it would:
D
• Prevent some machines, such as molding, from being pushed on some part numbers while

D being pulled on others.

• Allow the team to immediately address the big problem area of supplying material from
molding and paint to assembly and quickly help improve on-time delivery to customers.
• Make it possible to combine material handling across cells and value streams, capturing

D full benefits more easily.

D Expansion Approaches

How should you proceed with the systematic rollout of your level pull system?
D If your production assets are fairly light in nature, such as basic assembly operations,
and dedicated to specific value streams, it is fine to implement the overall system
one value stream at a time. As long as you can parcel out your factory into these
easy-to-manage discrete pieces and work on one area while not affecting others,
then it is logical to pursue the value-stream approach.

Unfortunately, most plants have shared assets, and it is tough to put only part of a
batch machine on pull while other part numbers for other value streams run on push
D schedules. If you have shared assets and many batch-producing pieces of equipment,
you will do better to take a whole "fix the department" approach. It may sound like
D more work, but the nature of the task is the same and it's easier to tackle it all at once.
On the positive side, benefits will accrue faster and there is less chance that you are
fixing one piece of the puzzle while causing damage elsewhere.
D In either case, you eventually will have to think across all departments and all assets.
Like Apogee, you will find that many tasks, such as material handling and scheduling,
must be shared across the facility to maximize their efficiency. This is a subtle feature
of system kaizen that is very powerful and can capture benefits that a vertical value-
[J stream rollout initially will miss.

In the end, as always, weigh your needs against your capabilities and resources, and
pick what works best for your specific situation.
D
D
D
D Part 5: Expanding the System 77
 D
D
To move forward with the departmental approach, Apogee held a meeting of the improvement
team and the entire management. This group created a future-state layout for the entire facility. D
The team realized that it would be easier to set each piece in place once they had a firm vision
of where all the pieces would go.

Future-State Overall Layout Concept

D
Legend

Q Raw Materiel
D
Receiv1ng
Dock
D
D
WI?· Die cast

Wl~·Molded
n
L__J
• WIP·Palnt

D
• Purchased Parts

Pl• ~·st
I
M,c:hln~_ Pirts
• Frnis;hed Goods

DD D AssemblyCell
cu
I I
········:
MoldedPart, ~

~
lnJectfon Molding

DiaCa&t
D
: §] Machtnlng Center

: [§] i'
~::;e f: L_J
Wesher
Cell #4 Coll #3 Coll #2 Cell #1
Aisle
Ov•rflaw D D D D
! D D D D D D D D
:
:: D D D D D D D D D
··-··-····

Open
:···-···--· -·-· Reserved Reserved Cell .fl6 Coll #5 D
Spece-
i:
.: Open
Space
D
D D
D D
D
D D
D D
D
D D
D D
D
D D
D D [J
:
I•••••••••••••
Shipping

D
Dock

Finished-Goods
Storage

Based on this vision, Apogee's improvement team decided to implement four large rollout
projects in the following order: D
Step 1: Complete the signal kanban implementation for batch processes.
Step 2: Organize the remaining inventory into central markets.
D
Step 3: Establish a plantwide parts withdrawal kanban loop.
D
Step 4: Establish a plantwide production instruction kanban loop.

D
D
I'
L __j
78
r-·1
Step 1: Signal Kanban
Complete the signal kanban implementation for batch processes. Special Conditions
Apogee chose to first complete the implementation of the triangle
Keep the following points
D kanban in all the injection molding and die-cast areas. The team had
learned from the initial pilot that upstream batch processes often were
in mind when using signal
kanban:
the problem in terms of getting the right part to the right place at the
right time. Curing the big problems here first would have tremendous Signals for raw material:
spillover benefits for the entire facility and the remaining value streams. You will need to develop
either a time-based or
The process was straightforward and identical to their work with the quantity-based method to
exterior-mirror value stream (the methods described in Part 4). Machine signal the need for raw
material for machines
by machine and part number by part number, the improvement team
using the signal kanban.
calculated the available time, the run time, the lot sizes, and trigger
0 points for each part number in regular production.
Otherwise, you'll lose
valuable production time
getting raw materials to
Triangle kanban were created, inventory quantities were determined,
the machines in time for
and inventories for molded and cast parts were moved to a central product changeovers.
market. On average there were three part numbers per molding machine
in the plant. With 12 injection-molding machines, the molded parts Nonproduction parts:
.----, You may need to run
' section of the market held 36 part numbers with an average of one day
LJ ·' of inventory.
non-production parts in
conjunction with the signal
kanban replenishment
Die-cast processes were similar, but with a slight twist. After the parts
system (e.g., engineering
were cast they needed a small hole drilled and tapped. In this case, samples or infrequently
material could flow directly from the die-cast machine to a special run service parts that are
machining center and washer unit on a first-in/first-out (FIFO) basis. not stored in inventory).
r- There was no need to create a market between the two process steps. For these items, you can
i i
L_i
Instead a FIFO lane was created and a standard amount of work-in- make a generic temporary
kanban that is painted
process was established: two totes, or about one hour of material,
a different color to call
between the processes to absorb minor downtime and fluctuation.
attention to the fact it will
The finished items then were stored in the central market. run only one time. If you
need to run it periodically,
Over a four-week period, Apogee was able to convert the entire molding
you'll need to adjust
and die-cast portion of the facility (approximately 10% of the parts) your trigger points for
to pull scheduling with triangle kanban as the control mechanism. replenishment of other
After some time trials it was found that one person needed to be parts made on that machine
dedicated to moving produced items from the machines co the market because the response times
area for all of these machines. While in the market area, the same will be longer.

person was responsible for scanning the markets for inventory levels
that had hit the trigger point for replenishment. When this occurred,
r··-·1
LJ

79
 D

the individual took the triangle kanban and the appropriate amount of empty containers
from the market to the designated machine. This cycle of activity dramatically lowered L~_J
inventory across the plant and improved the linkage between the molding departments
and downstream processes. n
1.______,

Step 2:
Organize the remaining inventoryinto central markets.
The team now moved to establish markets for all remaining parts, including those coming
from suppliers (purchased parts). D
Because lot sizes and trigger points for all the part numbers in injection molding and die cast n
L.__.J

had already been defined, the two remaining challenges for chis seep of the rollout were the
interfaces with the paint department and purchase-pares storage area. The paint department
was governed by the weekly production pattern (described at the end of Part 4) that dictated
n
'-----J

the order of the parts through the paint system and, hence, the replenishment time. The team
merely had to identify all the regular part numbers and calculate the average demand for each
[]
of them over the past several months. This average demand multiplied by the one-week
replenishment cycle and the addition of a small safety factor to cover for possible rework losses
determined the amount of inventory in the central market for all SO painted part numbers.

Each day at the shift's start, Production Control checked the inventory of painted items in the
central market to see how it compared to target levels and, consequently, the amount to run
that day. Production Control then issued basic quantity information to the paint supervisor
n
LJ
on how much to run of each color. Because of the need to paint nonregular items-such as
samples, engineering requests, and rare spare-part orders-the paint team reserved a small D
amount of open time at the end of every shift for this type of work. The open time also was
used for preventive maintenance. D
For the purchased-parts section of the market, the task was similarly straightforward. For
each purchased part, a review was conducted to determine the appropriate replenishment
and delivery interval for external suppliers. Approximately 35% of pares were from local
suppliers that would deliver on a daily basis provided they received one-day notice on the
exact quantity. An additional 50% came from in-state sources that would deliver only on a
weekly basis. The final 15% of the suppliers were located far out of the region and some i \
were in other countries. These suppliers shipped in bulk quantities on either a two-week or
monthly delivery schedule. On average, however, about five days of inventory was held for
each purchased part in the plant.
D
Replenishment of the market was conducted by Production Control using a simple check- n
L_I
and-order process. At predefined trigger points (e.g., the two-day level for weekly items) a
1.------,
I \
L .t

r---i
L_l
80
[J
Production Control analyst reviewed the items that were due to be ordered and fine-tuned
D the actual order amount as necessary, adjusting for scrap losses and other problems. Although
not a pure kanban-replenishment system with the supplier, this basic method worked well
[J enough for monitoring and controlling inventory at this point in Apogee's lean transformation.
(For detailed instructions on setting up purchased-parts markets and calculating replenishment

D cycles with suppliers, see the LEI workbook Making Materials Flow.)

While establishing the central market the improvement team also came up with ideas to
n handle materials more efficiently.

• Flow racks for parts were marked and color-coded with signage to show when areas were
D full, at what point to reorder, and at what point-dangerously low levels-to expedite.
• The market was addressed by aisle, row, rack, and shelf, and this information was placed
on the parts withdrawal kanban for each part.
• Heavy parts were put in easy-to-access locations.
• Frequently used parts were stored in locations near the end of the aisle where they were
easy to pick.
• Parts of similar description were stored together for ease of locating by the material
handler (storing parts by supplier was tried initially, but deemed less helpful).
D • Signs and aisles were posted to show how traffic should flow through the markets.

Step 3:
Establish a plantwideparts withdrawal loop.

D The current parts withdrawal loop in the plant covered only the exterior-mirror final assembly
cell. To expand the withdrawal loop across all six cells, Apogee needed to first organize and
r---i
LJ
address all of the cells and flow racks for ease of delivery. The improvement team would then
create parts withdrawal kanban for all items at the assembly cells, and determine the right
pitch for the withdrawal loop and the resources needed to run the loop.
D
For the parts withdrawal loop to function effectively, the addressing scheme was rolled out

D across all of assembly. Since there were only six final assembly cells and all were located in
the same area, the simple method used earlier for addressing the exterior-mirror cells was
continued (see page 50).
D
Each item that needed to be stored at assembly cells also needed a parts withdrawal kanban
D established. The improvement team created these kanban and filled in the needed information,
including the address locations for pick up and delivery of product.

D The team then had to determine the total number of material handlers to run the route, pick
up the kanban, and deliver the parts.

D Part 5: Expanding the System 81

 D
D
The team estimated that across the six assembly cells, an average of 180 cards per hour
would be triggered. The team earlier had estimated that a material handler could handle 10 D
to 12 cards on a 15-minute route, which implied that just over four material handlers should
be needed to handle all the cells. This number was much less than the 10 material handlers
previously used for each shift-and that was just for assembly-related material handling!
D
The challenge was how to best organize material handling for total efficiency. Dedicating D
material handlers to zones made the workload hard to balance because some areas triggered
II
I I
more parts than others due to varying pack quantities and take times. Dedicating material LJ
handlers to individual cells could clarify responsibility, but this approach would require six
material handlers, meaning that each would be underutilized.
D
Finally Apogee struck upon a winning solution: To balance work and total efficiency, Apogee
would have all four material handlers serve all the cells in an integrated fashion. The team D
would work in a staggered, clockwise fashion. Imagine the overall parts withdrawal loop as a
circle; the material handlers would rotate around the loop 90 degrees apart from each other. D
After several attempts, this method worked well (see graphic below). Each material handler
consistently picked up 10 to 12 kanban on what was now a full 15-minute route across all cells. D
The variable work component that changed in the equation was mainly the additional walk
time to circle all six assembly cells for pick up and delivery. The rest of the work elements D
remained the same.
n
LJ
Four-Person Material-HandlingTeam

Central Market Area Assembly Cells D

3-4 min. gap
(9~

D u-@ D
Painted Psrte Molded Parts
D D D D D
D D D D
D D
D
D D D D
Purchased Parts Empty Containers
D D D D D
D D
D
D D
D D
D D ©~DD
3-4 min. gap
D
Ii
L.J
[]
82
 r----,
L.J
r----,
I I
L.J

Updated Work Flow and Times for Withdrawal Material Handling

D
Conveyance
time
Step Standardwork elements estimate
Cell operator withdraws kanban card from the container upon removing NIA
the first part out of the container. The withdrawal kanban cards collect
inside a kanban post at the station.

On a fixed-time route the material handler travels from the central 2min.
1.
market and arrives at the assembly cells.

The material handler visits and picks up all the withdrawal kanban at each 6min.
[l 2. of the assembly cells as well as any empty containers, delivers parts to
the cells, and returns to the market.

11 Once back at the store, the material handler sorts the cards for the best 15sec.
1 j 3. pick order.
L.........i

Empty containers picked up on the route are dropped off in the correct 1 min.
4. location in the stores.
D Any necessary bins for small items, such as nuts, bolts, screws, 1min.
5. washers, etc., are obtained.
[J Parts are picked from the store and placed on the material-handling cart 4min.
6.
(assume average of 12 withdrawal items each at 20 sec. per pick).

The withdrawal kanban are placed into the container for the 15sec.
7. appropriate item.

D Total time observed for plantwidewithdrawal kanban rollout

14mln.
30sec.

11
Once Apogee was able to standardize the system, there were several benefits to using a
D conveyance team:

• Multiple people were trained in the material-handling job and ochers could quickly be
D trained as well. The risk of lost skills due to retirement, vacation, or sickness was reduced.
• The facility now had a standard for determining material-handling work in assembly.
When additional cells or pares came online the team could determine exactly how much
extra work or resources this would involve.
• The focus for kaizen became clearer. The long walk time involved with the cares and the
current route were not the optimal solution. Incremental improvement of 25% would
allow the team to reduce the number of material handlers in the loop from four to three.
If material handlers had been dedicated to cells, cross-value-stream improvements
D could not lead to gains in productivity.

r ··-,
LJ
Part 5: Expanding the System 83
 n
LJ

Step 4:
Establish a plantwide productioninstructionkanban loop. D
Much like the initial parts withdrawal loop in the plant, the production instruction loop currently
covered only the two exterior-mirror cells in the facility. The current loop pitch for the pilot D
cell was still 18 minutes, with only 10 minutes of actual work for the material handler. To
expand the instruction kanban loop across all cells, Apogee needed to determine the total
workload for the production instruction kanban loop as well as establish the overall pitch
D
for the withdrawal loop and the resources necessary to run the loop.
D
To determine the workload for the production instruction kanban loop, Apogee's improvement
team reviewed the pack quantity and takt times for each of the six assembly cells in the plant D
as well as their respective pitch intervals.

Cell Workloadfor InstructionKanban Loop

Area Takttime Finished-g<>odspack quantity Pitch interval D

Cell #1 54 sec. 10 540 sec. (9 min.)
Cell #2 54 sec. 10 540 sec. (9 min.) D
Cell #3 60 sec. 12 720 sec. (12 min.)
Cell#4 60 sec. 12 720 sec. (12 min.) D
Cell#5 45 sec. 8 360 sec. (6 min.)

Cell#6 45 sec. 8 360 sec. (6 min.)

n
L._J

I !
As with most plants, there was no common pitch interval across the facility. Each product
family had its own take time and pack quantity, and Apogee needed to develop a loop co accom-
modate these differences. D
Apogee wanted to follow its previous practice of creating standardized work and balancing
the material handler's workload to the pitch intervals. Fortunately the basic work sequence
D
of delivering the instruction kanban and empty containers, picking up the full containers from n
L ..J
the cells, and delivering the product to the finished-goods area was standard across the cells.

Apogee calculated the total number of instruction kanban needed per hour per cell by dividing
a typical hour by the pitch interval.
D
D
D
D
84
D
Instruction Kanban per Hour per Cell
D Area Time + Pitch interval = Instruction kanban per hour
.,_
Cell #1 60 min. 9 min. 6.7
D
;;;;;

Cell #2 60 min. + 9 min. = 6.7

Cell #3 60 min. 12 min. 5

D T ;;;;;

.,_
Cell #4 60 min. 12 min. ;;;;;
5
Cell #5 60 min. T 6min. = 10
D Cell #6 60 min. .,_
6 min. = 10

Total 43.4
D
D Like the parts withdrawal kanban loop, Apogee's improvement team realized that much
of the work in the instruction kanban delivery loop was fixed (e.g., walk time) and that the
r·-, variable portion was directly related to how many containers were transported per trip by
LJ the person handling the material.

Regardless of how many containers were picked up, the total transportation time using a
material-handling cart for finished goods was measured at roughly three minutes. Each
finished-goods item picked up added about one minute of additional work. The team created
D several options for finished-goods material handling based upon this insight.

Apogee Decision-Options for Finished-Goods Material Handling

D Potential Fixed
Trips
per Kanban Variable Total
Options Time hour per trip* time** time Comment

D 18-min. loop 3 min. 3.3 13.2 13.2 min. 16.2 min.

OK.
Current pilot pitch
OK.
15-min. loop 3 min. 4.0 10.9 10.9 min. 13.9 min.
D Even 4 trips/hr.

OK.
12-min. loop 3 min. 5.0 8.7 8.7 min. 11.7 min.
r~1 9-min. loop 3 min. 6.7 6.5 6.5 min. 9.5 min
Even 5 trips/hr.
Overburdened
6-min. loop 3 min. 10.0 4.3 4.3 min. 7.3 min Overburdened

D * 43.4 kanban per hr . .,. trips per hr.

** 1 minute incremental work per kanban

D
LJ
D
D Part 5: Expanding the System 85
 D
D
Apogee knew that the shorter the loop the more frequent the feedback to the cell regarding
production status. However, short loops required more skill and the ability to respond quickly D
to problems in collaboration with line supervision. Having finally demonstrated the ability to
maintain production to the 18-minute pitch in the trial phase, the Apogee team was ready to l_]
move to a shorter pitch interval.

Apogee realized that pitch intervals of either six or nine minutes were not attainable. Both D
of these options were rejected as next steps, but in the future they would be considered after
additional kaizen. D
Both the 12-minute loop and the 15-minute loop remained as options. With a 12-minute loop,
instruction cards would be picked up five times per hour and approximately nine containers D
per trip would be handled. Although this was an aggressive target, with just 18 seconds of
slack between each run, the conveyance supervisor felt confident of handling this interval D
with current methods and equipment.

Apogee decided to make the 12-minute loop the plantwide instruction kanban pitch. But
D
there was one problem. The pitch interval for cells #1 and cell #2 would not match evenly
with the 12-minute overall material-handling interval (e.g., nine minutes does not evenly
[]
divide into 12). Without some alteration there would be times when cells #1 and #2 would
be left without an instruction kanban and, hence, the signal to build. D
Pitch Interval Delivery D

Minutes
I
0 3
Pitch 1

6 9
Ir
Pitch 2

12151821
I
Pitch 3

24 27 30 33 36
I Instructionsfor Material Handler r
Drop off 2 kanban at start of interval.

Interval D
Pick up 1 finished-goods container at
end of interval and drop off 1 more D
D kanban. Repeat next interval. Pick up
Cells 1 and 2 9 min. CT D D D D 2 finished-goods containers at end of
next interval and drop off 2 kanban.
Cycle repeats.
D
Drop off 1 kanban at start of interval.

Cells 3 and 4 12 min. D D D D D D Pick up 1 finished-goods container at

end of interval and drop off next D
kanban. Cycle repeats.

D D D
Drop off 2 kanban at start of interval.
Pick up 2 finished-goods containers D
Cells 5 and 6 6 min.
[5:) D 0 D 0 D at end of interval and drop off next 2
kanban. Cycle repeats.

Legend
D
O
D
Instruction kanban
Finished-goods container
D

[l
86
D

To deal with cell #1 and cell #2, Apogee decided to stagger the delivery of information at these
cells (see Pitch Interval Delivery). Instead of dropping off only one production instruction
kanban at cells #1 and #2, the conveyance operator dropped off two instruction kanban on the
LJ first running of the route. This kept the cells running with 18 minutes' worth of production,
until the route operator came around again 12 minutes later.

On the second trip, the conveyance operator dropped off one kanban card and picked up only
one finished-goods container. This same basic pattern repeated on the third cycle, except that
r·1
LJ there were two finished-goods containers to pick up. The cycle then repeated. Because of their
pitch interval and finished-goods container size, staggering was not necessary for the other
r 1
I
L___J
I assembly cells.

Dealing with Unmatched Pitch Intervals

Frequently you will find that pitch intervals do not align across cells. One option is
LJ to determine the lowest common integer applicable to each pitch. For Apogee, six,
nine, and 12 all evenly divide into 36. Thus every 36 minutes cells #1 and #2 get four
instruction kanban each; cells #3 and #4 get three instruction kanban each; and cells
LJ #5 and #6 get six instruction kanban each.

For some facilities in the initial stages of level scheduling and pull systems this
D option is fine. In Apogee's case, this interval was too long to pace the system and
signal problems.
LJ A second option is to dedicate material handling by value stream and product-family
pitch interval. In this case, one material handler would handle cells #1 and #2, a
·- · · 1
L sec-ond person would handle cells #3 and #4, and a third person would handle cells #5
and #6. If the workload for the material handlers matches the pitch interval, then this
is an acceptable solution. However, this option also did not match up well for Apogee.

In some situations, finished goods can be moved away piece by piece from the cells
and put in a shipping area for combined packing efficiency. But this is risky because
parts might be mixed into the wrong pack containers, and proper steps must be taken
to avoid errors. Other alternatives exist, such as changing the pack quantity of con-
tainers, but this, too, is risky. Takt time can change and disrupt your efforts to make
pitch intervals match.

r·._J
· I When the pitch intervals simply don't align, as in Apogee's case, you can always have
some cells that get instructions slightly ahead of time to ensure that they will keep
building, and then pick up finished goods on the next or following trip. This is what
Apogee did for cells #1 and #2.

r: ""]
I_!

[]
Part 5: Expanding the System 87
 D
n
LJ
Expansion Complete
Apogee completed the rollout of the four major expansion steps over a four-month period as
D
outlined. During this period it was necessary to build up some inventories and to make minor
layout changes, which required some overtime and weekend work.

The information and material-flow map (below) for cells #1 through #6 shows Apogee's
completed system kaizen to create a level pull system.

Informationand Material Flow for Entire Facility

I l
Parts Instruction
by signal hnban
Ports Instruction by
pattern production
Parts withdrawal
materlal-handllng
Production instruction
loop for finished goods D
(triangle]

;------\}·-; :---- --- ~

loop

-- --: ,.- ----- --: D ---- ._

IO X OX! Heljunka box

'
·ui~~
: ...'
Empty-ContilinerArea

! IM-1 ! G
!IM-2 !
tJ! DD
DD
DD
DD
DD
DD
DD
DD
DD
n
Ll
Cell#T
~ D DD DD
! IM-3 ! 27 parts 0
1 day
- - - - - - - - - - - - _.,.. 5days
0Ql0©0 Finished-GoodsStoreArea D
! IM-5 [ • [i]Q) ©0
DD
DD
DD
DD
DD
DD
! IM-6 !
! IM-4 [
Ce/1#2
DD
DD
DD
OD
DD
DD D
w&Raw material
!IM-11! ~G 0 DD
DD
! IM-7 ! 0G>0©0 DD DD
5-7days !IM-12[ QJG> © m DD DD
! IM-8 !
.
'
' ________
\ ....
9parts
1day
DD
DD
DD
DD
DD
! IM-9 ! Ce/1#:3
DD DD
DD
DD
[J
IIM-10! 0 DD DD DD
Purchased-Parts 0Ql0©0 DD

------------l
Market [iJ Q) ©m DD
DD
DD
DD
DD
D
20parts DD DD
5days DD DD
Celi#4
.---------------y::-z--------------------.~ finishedqoods 1-2 days
•' v "' ' 0
0Ql0©0
I:_;,':(' jcc-1 I -mo- !Mc-21 -mo- !WA-1! ~ G 0Ql ©0 ["I
~~rial....__ __. ........
1_J
5days
Ce/1#5 .----,
0
I I
L.....__ _,
4parts
lday [IJQ) ,Q, © 0
' ©0
[i]Q)
n
1 I

[ __ _j
Central Markets
Cel/#6
,------,

[I]QJ,Q,©0
0 t•
[iJ G> © m
----------------- . n
,--,
IL __._JI
88
D
[]

r
L__j

D
[]
r·,
' I
i t
l-....J Expanding the System - Keys to Success
• Expansion via value-stream rollouts across the facility
D works best when equipment is dedicated and few
assets are shared.

[] • Batch processes often need to be addressed

collectively and not one process at a time.

LJ • Capturing efficiency in scheduling and material-

handling operations often means sharing resources

,--i
c_J •
across value streams.

Establish plantwide parts withdrawal kanban loops

and instruction kanban loops-do not establish

c •
individual loops.

Sequence implementation steps carefully and tackle

the biggest problem areas first.
LJ
r-1
L..........!

,-·-)
I ,
L.J

D
,·--1
I
c__;
I

r-,.1
LJ

\J
I I
'---'

r--1
I .
~

r-~
I ;
L-...1

f~l
L.J
89
'[J

D
Ji
L....J

D
,-·---,
LJ
[J
11
LJ

D
r: -·1

LJ
r~ -i
L_J

,--1

[J

r ·1
L

,.--·-~- ·-- .,
i\.._J ;
 []

D
D
Sustainingand Improving
r-,
, I
The benefits of Apogee's plantwide step-by-step implementation of a level pull system were
"------' far-reaching, with improvements on many dimensions. Nothing changed with regard to the
actual processing activities-molding, paint, and assembly-yet direct labor productivity went
up substantially and indirect labor productivity went up dramatically because information
and material flows now support rather than thwart the workflow. At the same time, operating
expenses for inventories, space, overtime, and expediting were slashed, in some cases to zero,
and Apogee became much more responsive to changes in customer requirements, particularly
for product mix.

r·---, Despite their success over a six-month period (as seen on the next page), Apogee's improvement
. I
f I
team knew that the complete implementation of the level schedule and pull system was not
.____!

the end of their challenge. This event simply marked the need to shift focus from the lean
D conversion to creating a management system that could sustain and improve the new
system over time.

r··· 1
-------

0
r--i
F ··1
'---...)

, :··-·--,
..
I i
~

Part 6: Sustaining and Improving 91

 ,----..
' (
L____J

Box Score-All Value Streams

n
D
Start of Current
level pull Target state Improvement Comments
Productivity
D
Direct labor 10.2 12.5 12.5 22.5% No overtime
(pieces per person per hr.) D
Material handlers 25 12 15 40% Full utilization
(all shifts)
Quality
D
Scrap 2% <1% 1.5% 25% Not
addressed
D
Rework 15% <5% 12.5% 16% Rapid
detection
External (ppm) 105 <50 105 0% Not
.----,
addressed : I
LJ
Downtime
Assembly (min. per shift) 30 <5 10 66% Reduced wait
for material
Paint 15 <10 10 33% Reduced wait
(min. per shift)
Molding (min. per shift) 20 <10 10 50%
for material

Reduced wait
LJ
for material
Inventory turns
[]
Total
On-time delivery
10 30 24 140%
n
·-_J

To assembly 75% 98% 98% 23points No overtime

or expedites
To shipping 85% 100% 100% 15points No overtime
or expedites
To customer 100% 100% 100% Oposite No overtime
or expedites
Door-to-door lead time L l
Processing time (min.) 125.7 125.7 125.7 0% No change

Production lead time (days) 30 12 12 60% Order lead

D
time of 2 days
Other
Overtime costs per week $25,000 $0 $0 100% No overtime
Expedite costs per week $9,000 $0 $0 100% No expedites

n
L~

n
l_._._.'..
92
 D
D
Question 11:
[] How will you sustain your level pull system?
Sustaining a level pull system is not an easy or short-term activity. Apogee needed to
change its fundamental management practices for Production Control and Operations.

In any level pull system three management activities are critical:

A. Continuous monitoringof customer demand

B. Continuous assessment of performance metrics and process stability
C. Daily supervision of productioncontroland operationalprocessesto ensure that
standard work is being followed

Someone must perform each of these tasks, but every organization probably will have a different
D allocation of responsibilities. Apogee assigned responsibility for its ongoing activities as described
here, but you will need to have an explicit discussion and make precise decisions about the
D best allocation of responsibilities in your organization.

D A. Continuous monitoringof customerdemand

,-- - , Customer demand (which is then divided by the available production time and converted to
LJ takt time) is the critical foundation of any level pull system. The amount of finished goods
r~-, to hold, the amount of goods in the central market, and the hours of production to schedule
. I
L__J are interlinked and tied directly to a calculation of average demand over a given period.

r--·1 As average demand changes, Apogee must react by adjusting inventory levels throughout the
facility. Apogee therefore established a monthly demand review led by Production Control
and also attended by all of the Operations departments. The review assesses demand for
D each product.

For example, in Apogee's case we saw how one small improvement dramatically affected the
level of finished-goods inventory (see Part 3). Before the schedule was leveled, finished goods
for part #14509 were replenished on a weekly cycle, and this required a total of 1,200 parts to
be held in the finished-goods market. After the schedule was leveled and the A items were
put on a daily replenishment, the amount of finished-goods inventory required for this part
[J number was significantly lowered to 240 parts (see next page). The reason was simply that
one of the key drivers of finished-goods inventory (the lead time required to replenish that
.--~
LJ inventory) had changed from five days to one.

D
,-··,
J i
L--1

Part 6: Sustaining and Improving 93

 []

D
Cycle Stock and Buffer Stock Variables
D
Finished-goods Key drivers

n
Part#14509
inventory levels
Cycle stock*
Amount required

J
160 to cover normal
Lead time to replenish
D
I
demand
Buffer stock Demand volatility
Amount required
240
L_]
J
to cover customer-
,___..___..__---1 induced variation Forecast/MRP errors

Safety stock Downtime tosses

~~
40 ~
Amount required
to cover internal
D
losses Quality losses
Time c:,
a,
'<
N
0
a,
-c
w
D
~
* Assumes an average customer demand and draw down during each day. I
,_J I

A change in other drivers also can change finished-goods inventory levels. For example, a
reduction in average demand for this part number would further reduce the amount of cycle n
~_J
stock required. If daily demand falls from 160 items to 100 items, then the amount of finished-
goods inventory needs to be reduced proportionally. Similarly, if downtime significantly
increased, then inventory would need to increase in proportion.
D
A second dimension that should concern Apogee in addition to total demand for each product
family and total inventory is mix variations. Often changes in mix are more important than
ii
changes in total demand, particularly in an operation such as Apogee's with a large number LJ
of paint colors and configurations.
,------.,
For example, the paint market for the exterior mirrors holds five days' worth of inventory
L__ J
for each part number. The five days' worth of inventory is a relative number that depends
upon average demand. Total demand for the value stream might easily stay constant, but
the mix between two colors in the market might shift up by 50% on one item and down by
50% on another. The inventory in the market needs to accommodate chis change. Otherwise n
L_J
when the parts withdrawal material handler goes to pick product for the assembly cell there
will be too few of the items needed.
n
The final but critical component of Apogee's monthly demand review concerns demand
volatility. Although Apogee has gone to great lengths to level demand within its walls (to the []
great benefit of both Apogee operations and its suppliers), there is no reason to think Apogee's
customers have leveled their orders. Indeed, these may be getting more volatile even though n
L.J
their average demand remains constant.
!\
L.J

n
L._j
94
 As volatility of customer demand changes, Apogee needs to change the amount of finished
goods held in its buffer inventory to protect the facility from the erratic demand waves. Doing
this is a great use for statistical analysis with standard deviation calculations. For example, two
D standard deviations of demand for part #14509 equaled roughly 25% of demand or 40 parts
daily or 200 parts weekly. An increase in volatility leads to a proportional increase in buffer

D stock. Apogee will need co adjust the amount of this product held in finished goods even
though average daily demand is unchanged at 160. If Apogee managers fail to adjust finished

c goods in a period of rising volatility, waves of expedite orders will flood through the facility
and risk capsizing the ship they have carefully transformed. And if volatility falls without
anyone noticing, the facility can carry unneeded finished-goods inventory indefinitely.

The key questions related to customer-demand information for Apogee management co

D consider on a monthly basis include the following:

• Has average demand, mix, or volatility changed for any reason? If so, how will this affect:
o Finished-goods inventory (cycle, buffer, and safety)?
o Layout of finished-goods stores or shipping lanes or markets?
o Inventory in markets?
o Trigger points for batch processes?
r1
• Has takt time changed for final assembly? If so, how will this affect:
o Staffing levels in each assembly cell?
o Pitch interval for instruction kanban?
o Number of withdrawal kanban per hour triggered and required number of material
handlers?
D o Withdrawal and instruction kanban loops?

D B. Continuousassessment of performance metricsand process stability

At the individual manufacturing process level (assembly, paint, and molding) and in production
control, it is important for Apogee to measure capability and performance over time. Without
measurement, performance will backslide and the causes of the slippage will be a mystery.

At each process step it is essential to measure basic items such as scrap rates, changeover times,
and downtime. Knowing the precise performance in each activity and charting it over time
will show what improvement is needed.
[_]
Performance is not just important for direct operations but also for inventory levels because
i' the safety stocks in the system are directly related to the likelihood that the right amount of
LJ

Part 6: Sustaining and Improving 95

 D
D
products can be produced at the right time. Failure to monitor performance will result in
ineffective markets chat fail to absorb process instabilities and protect the customer. However, D
as you calculate safety stocks please remember that no market can solve an instability problem.
The only solution is targeted kaizen as problems are prioritized, root causes are determined, D
and permanent fixes are put in place.

Key questions related to process stability for Apogee to consider include: LJ

• What metrics do you need in the value stream to monitor the stability of performance
(on-time delivery, scrap, rework, productivity, safety, downtime, etc.)?
D
• How often will you measure and update these items for the value stream?
D
• Who is responsible for tracking, reporting, and improving these items?
•
•
Where exactly is the instability in each value stream (e.g., paint, molding, die cast)?
How large is the instability and what amount of inventory and lead time is it driving?
r
• What point kaizen events will you need to solve these problems? D
• What do you need to improve first?
D
C. Daily supervision of productioncontroland operational processesto ensure that \t
standard work is being followed I
L__J
I

The final activity necessary for Apogee to sustain level pull scheduling is active supervision
by Production Control and Operations. There is no scheduling system that works perfectly
according to plan, day in and day out, so these two departments will need to make adjustments.
The techniques described in chis workbook are ways to set a plan in place, buc they are
intended to be flexible enough to enable you to react to the inevitable changes that occur.

In one sense, Apogee's level pull system is the ultimate tool for uncovering sets of problems
D
that existed in the facility but were previously unnoticed-defects, overproduction, lace
deliveries from suppliers, lengthy changeover events, and material stored in wrong locations. D
However, it still is the role of Apogee supervisors to detect these systematic problems and r--,
' I
make sure they are addressed and solved over time. LJ
With that in mind, Apogee made anything important to the success of the overall system Ii
(__J
highly visual. This will increase the odds for sustaining improvement because everyone will
catch abnormal conditions faster and react to them quicker. Talking with people or getting
reports from subordinates is one way for supervisors to obtain information, but it is passive
and may miss or filter out critical items. Active management is required among front-line
supervisors for lean manufacturing to succeed.

n
L_J

96
D
n
L__J

D
In implementing a more active and visual style of management for each value stream, Apogee
managers will need to seek answers daily or even more frequently to several questions:

• Is production ahead or behind schedule?

D • Are inventory levels above or below normal?

D • Do we have the right number of resources in place?

• Are machines producing to cycle time?

D • Is assembly producing to takt time?

• Are defects occurring and escaping downstream?
• Are suppliers delivering on time?

n Making the answers to these questions-which seek to identify abnormal conditions-obvious

LJ
and visual will go a long way toward helping Apogee sustain the gains from its level pull
implementation.
D
D Question 12:
How will you improveyour level pull system?
D By applying standard management co its processes, Apogee was able to sustain the gains it had
made in its level pull system. The next and ongoing challenge for Apogee and for every
facility is to improve the system.

In Apogee's case there still is a considerable amount to improve. By design, several short-term
compromises and trade offs were made during the implementation that will gradually need to
be resolved. This is a typical pattern and almost certainly will apply to your facility.

D
Process Stability
For each of the production processes in Apogee, there is some remaining waste that manifests
itself in the form of availability, speed, or quality losses. All of these types of losses must be
compensated for in your inventory calculations in the form of safety stock. Once safety stock
is created, however, it does not mean that Apogee has license to leave it there forever. Unlike
customer demand and demand variation, which is out of Apogee's control, safety stock can
be reduced over time as causes of instability are eliminated.

n
L__j

n
\...__..]

Part 6: Sustaining and Improving 97

 LJ
D
Stability ImprovementLevers

Finished-goods Part#14509 Key drivers

inventory levels
Cycle stock
Amount required
Average demand D
!
160 to cover normal

I
Leadtime to replenisl
demand
Bufferstock Demand volatility
D
Amount required
240
to cover customer- []
~®
! induced variation
Safety stock
Amount required ~
Forecast/MAPerrors

- Downtime losses
D
40 - to cover internal ~
losses ~ Quality losses
Time c,
DJ
c,
DJ
c,
DJ
n
LJ
-< -< -c
N w

D
At each seep of the system, Apogee still has 10 to 20 minutes per shift of downtime and 2% to
15% quality instability in the form of rework and scrap. These problems all result in safety-
stock inventory (central markets and finished goods) that must be held to buffer the system
D
against internal sources of volatility. For part #14509, approximately 17% of the total inventory
is being held co buffer against these internal losses, and this is typical.
D
Apogee will need to target and reduce safety-stock inventory through a series of targeted kaizen D
efforts. Apogee should focus on:

• Reducing scrap and rework in the paint department, specifically the problem of
[]
inclusions (small particles adhering to the paint).
• Eliminating equipment downtime in molding, paint, and final assembly, specifically D
minor mechanical downtime related to limit switches and sensors.
• Eliminating delays caused by material-handling problems between processes,
specifically between the producing departments and the central markets as well as
between the central markets and final assembly. n
L.....-·---'

Other point kaizen activities, such as set-up and changeover reduction, are important as well. ii
LJ
Reducing changeover time will shorten the lead time to replenish cycles and help reduce cycle
stock, which subsequently reduces a fraction of the safety stock as well. 11
L.J

,--,
I I
LJ

98
D
LJ
Customer-Induced Variation
Another important objective is decreasing inventory buffers set in place because of
LJ customer-induced variation. Ultimately the root cause of customer variation may be
entirely out of manufacturing's control, but a closer look always is warranted.

ln Apogee's case there were a few part numbers in the door-handle value stream that
suffered tremendous amounts of volatility. For example, while the average demand
for molded part #13901 was 1,000 units per week, the variation was up to five times
D this amount in a three-month period.

Investigating this specific case revealed that the part number in question was used
by a sister plant located two hours away. Since the demand for this item was fairly
low, the customer-service department at headquarters held orders for several weeks
before releasing them to the Apogee mirror plant. While demand was fairly constant,
the practice of lumping orders created demand spikes and volatility that did not
really exist.

When confronting high variation, you will need to peel back the onion several layers
to find the root cause. Often, in larger companies where the customer-service and
r·· ·1
L__)
order-entry functions are located external to the plant, information is not shared in
a timely fashion.

D Other customer-induced variation problems actually may be caused by internal pricing

decisions and sales behavior. Often customers are trained to expect end-of-the-quarter
price breaks or special volume-purchase deals. These types of variation are outside
D of the control of plant operations. Operations can and should, though, question whether
they provide value to the organization that offsets the variation and disruption they cause.

,-1 Systematic improvementpotential

A major source of system improvement can be found in nonproduction time, specifically waits
[J in inventory markets that constitute 99% of the nonvalue-creating time for products as they
proceed through the facility.
[J
Cycle stock can be reduced only by shortening the door-to-door lead time to manufacture.
Apogee had done an excellent job of reducing the required order lead time for any one unit
from several weeks to two days (the time from when the instruction kanban was sent to the
assembly cell to the time of customer shipment), but total manufacturing lead time was still
D approximately two weeks for the longest lead time item (molding in this case). For overall
system improvement, the greatest opportunity still lies from the assembly cell backward to

D raw material.

D
D Part 6: Sustaining and Improving 99
 D
System Lead lime vs. Order Throughput lime

Parts instructron by P•rt• lnstructronby Parts withdrawal Production Heijunka box

n
slgnal kanban sequence pattern material handling loop instruc1ion loop

lox ox I 11
L_J

Empey-Conralner Area
D
DD DD DD
DD
DD
DD
DD
DD
DD
D
D DD DD

WIPThr.
Flnl•hed·Goods Store Area · n
L_

1------------- I
DD DD DD
3parts
Tday
DD
DD
DD
DD
DD
DD
D
'' DD DD DD
DD
D
DD
DD I i
Purchase,H'an;s
Market finished goods 1-2 days
Parts Instruction by
slgnatkanban
20/tems
D
:--~~~--·~---------------------------::
f .
D
w&(~~~~~fjG
~WITU~rt-/
5-7days D
4parts
1day
n
LJ
2weeks 2days

In the case of the longest lead-time item (painted injection moldings) there are three locations
where inventory is kept upstream of the assembly cell: the market after paint, the market
D
after molding, and the raw-materials storage area. Each of these can be examined more
closely to find system improvement potential.
D
.---,
: I
Improvementsat Markets LJ

Market Quantity lead-time driver

5 days 1-week
·.

.. Possibilities n
Painted parts • Decrease paint batch size
replenishment cycle • Add second shift in paint
• Send instruction from .
D
assembly to paint
Molded parts 1 day 1-hr.
changeover time
• Reduce changeover time []
• Add third shift
Raw material 5 days Supplier delivery • Increase frequency of
• 50% weekly supplier delivery of current
• 35% daily supplier
• Identify new local suppliers

100
• 15% 2-4 weeks
CJ
r-1
D
1. Painted-parts market
D The painted-parts market is driven by a one-week replenishment cycle of the paint department
and the need to paint 50 colors across five production days. On average 10 colors are being
D painted per day during the one shift the paint department operates. In order to reduce the
replenishment lead time there are three main options for Apogee to weigh in terms of cost
D and benefit.

The first option would be co reduce the batch size in paint and paint more colors per day.
D Increasing to 20 colors per day would equate to a 50% reduction in painted material in the
market. Painting all 50 colors in one day would mean that replenishment time could be taken
D down to one day. The cost of extra paint purging would have to be weighed against the value
of the reduced inventory, but it often is the case that clearly identifying a large opportunity
D inspires technologists to rethink a whole process.

The second option for paint, assuming other factors are constant, would be to decrease the
D line speed and run a second shift of paint production. This would cut the replenishment lead
time in half and reduce the market inventory by 50% as well. Obviously the effects on quality
D and operating cost must be factored against the value of the reduced inventory.

If these options are feasible, especially the first, then Apogee should consider whether it needs
D the painted-parts market. If the paint process is stable enough and lead time to replenish is
short enough, then production instruction kanban can be sent from the assembly cells directly
D to the paint department to signal replenishment (see below). However, the cells will need
to increase the amount of painted inventory held at the line co cover any increase in
D replenishment time.

D Direct Replenishment Instruction from Assembly to Paint

D Parts instruction by
signal kanban
Replenishment*
instructions

D
D
.
:- -- -- -'\]-- -- -- -- --- :
Molding
:

G
.
'
'
---------------------------------- '

i-------.
Paint
-FIFO--
Assembly

0
0Ql0©0
D Rsw material
5-7days [QQ) ©0
12parts

D I day

" In this instance, the withdrawal kanban functions as an instruction kanban for the paint department.

D
D Part 6: Sustaining and Improving 101
 []

D
2. Molded parts market
The key lever for reducing inventory in the molding market is further reduction of the current D
changeover time of 55 minutes. Each minute that can be converted to production time reduces
replenishment time. This in turn enables Apogee to lower lot sizes and trigger points for D
replenishment and reduce lead time.

In some situations and without capital investment, companies have been able to reduce set-up D
and changeover time by 80%. Over longer periods of time and with cooperation from the
maintenance staff and equipment manufacturers, changeover times on some processes have D
been reduced to the single-minute level. If this can be achieved, the need for holding molded
parts in markets should be questioned. D
If changeover times cannot be reduced, Apogee could increase the time available for production
by adding more shifts and using this time to conduct more changeover events. The additional D
cost would need to be carefully measured against the benefits of reducing inventory from the
already low level of one day. D
3. Raw material D
The last area for reducing lead time is in the raw materials portion of Apogee's value streams.
These were not a significant problem for on-time delivery or quality, but because there is D
an average of five days' inventory on hand and a mix of supplier delivery frequencies, these
elements of Apogee's value streams eventually will be a target for system improvement.
D
Apogee can investigate the possibility of more frequent delivery direccly from suppliers, switch
to local suppliers, and change the logistics pattern for pick up of supplier material. Of course, D
these options have to be evaluated in terms of cost, quality, and inventory reduction to
determine the benefit to the overall system. D
The objective in every case must be to look at benefits vs. costs for the total system rather
than at individual areas or process steps. D
D
[]

D
D
D
102
D
D
D
D
D
D
D Sustaining and Improving-Keys to Success
• Monitor changes in customer demand and make
D adjustments to inventory as needed.

• Measure key process activities and monitor them

D for improvement.

• Practice visual control-if it is important, make it

visual.
D
• Improve process stability over time.

D • Reduce changeover time on internal processes.

• Investigate system design options that eliminate

D the need for markets between processes.

D
D
D
D
D
D
D
0
D
D 103
 r---,i Conclusion
L.1 This workbook illustrates a company taking a major leap in lean manufacturing by implementing

rl
.t.__J
a level and pull-based production-control system across all value streams within an entire
facility. While previous Lean Enterprise Institute workbooks have focused on point and flow
kaizen, Creating Level Pull aims to help manufacturers tackle system kaizen to introduce lean
D production scheduling.

Implementing level and pull production across several value streams and shared production
assets is a difficult task in any environment. In pursuing this type of improvement, you will
,-·1 face many decisions on which path to take and a series of options. There are no simple answers
u applying to every situation. Basic issues such as your current level of internal process stability,
length of manufacturing lead time, and the nature and frequency of your customer orders
D (As, Bs, and Cs) will affect many of your decisions.

The Apogee example shows that a level pull system can have tremendous benefits for any
company, especially one with discrete processing and repetitive demand. Often the step of
implementing level and pull methods for scheduling is an important enabler for stabilizing
the pace and consistency of production within a facility. Once level and pull systems are
implemented the tasks of operating cells to take time and making material flow between
locations become much simpler. Apogee's case shows that the measurable benefits of a level
pull implementation are large.

As you approach your implementation, I encourage you to focus on the following activities
and thought processes:
r··:
L.J • Treat your lean implementation as the systematic improvement of interconnected process
steps across the value stream. Don't just fix isolated areas or apply random improvement
tools. You'11 have greater impact and sustainability if you structure your implementation
as outlined in this workbook.
• Use value-stream mapping in conjunction with level and pull-based scheduling to highlight
your next set of improvement opportunities. Don't merely assume the location and cause
of problem areas. The implementation of level and pull production will aid you in seeing
more clearly where waste, overburden, instability, and variation exist.

• Remember to think and improve across value streams as well as within them. Often there
are huge improvement opportunities to be realized by attacking the wastes inherent in
C· shared production assets that are batch in nature, as well as addressing shared tasks such
as scheduling and material handling. Focusing on single value screams will miss large
r··--1 opportunities for improvement.
I I
t__J

I wish you great success when implementing your level pull system, and I look forward to
r:
---
,• i hearing about your progress.

105
 D
0
0
D
About the Author D
Art Smalley
Art Smalley learned about lean manufacturing while living, studying, and working in Japan for
D
10 years. Art was one of the first foreign national employees to work for Toyota Motor Corp.
in Japan and spent the majority of his Toyota career from the late 1980s to the mid-1990s D
helping Toyota transfer its production, engineering, and management system to facilities
around the world. D
As part of his education at Toyota, Art was assigned to the Kamigo Engine Plant. This is
Toyota's largest engine plant and the facility where Taiichi Ohno made many important
D
discoveries while serving as the founding plant manager.
D
After leaving Toyota, Art served as director of lean production at Donnelly Corp., a global
Tier One automotive supplier with plants in 15 countries. (Donnelly is now part of Magna
Inc.) More recently, Art helped start the lean manufacturing practice of McKinsey & Co.,
D
where he is still an advisor. Art currently divides his professional time between aiding
companies implementing lean systems in the U.S. and acting as managing director of SEI
D
Inc., an engineering and maintenance services company in Japan.
D
D

D
D
D
D
D
106
D
r--·-, Introduction to Kanban
LJ In lean manufacturing, the kanban is the specific tool for controlling information and regulating
materials conveyance between production processes. (The term is Japanese, meaning "sign"
D or "signboard.") Kanban coupled with takt time, flow processing, pull production, and level
,··----, scheduling is what enables just-in-time production to be achieved in a value stream. Typically
I
L_..J
, a kanban is used to signal when product is consumed by a downstream process. In the simplest
case this event then generates a signal to replenish the product at the upstream process.
[_] Kanban differs from traditional production control methods in several important ways. In
traditional manufacturing, the production schedule is provided to each individual process
and each process produces in accordance with that schedule (without timely feedback from
downstream processes regarding exact needs). Kanban instead functions as a physical schedule
tool that tightly links and synchronizes production activity between upstream and downstream
processes. Furthermore, in traditional manufacturing the movement of material between

D processes occurs when the upstream process completes the product. This results in a push
of material to the next station, independent of the exact needs of the downstream process.
Kanban instead combines control over movement of material with respect to both time and
D quantity dependant upon signals from the downstream process. Thus, kanban controls
production in a value stream by controlling material and information flow.

Traditionally within a single facility, kanban is a simple paper card, sometimes protected in
a clear vinyl envelope. The card contains basic information such as part name, part number,
D external or internal supplying process, lot size, pack quantity, storage address, and consuming-
process location. A bar code may be printed on the card for tracking or automatic invoicing.
D When communicating over long distances, electronic signals often are used in place of simple
kanban cards.
[J
Purpose of Kanban
There are four major purposes for kanban:

D • Prevent overproduction (and overconveyance) of material between production processes.

• Provide specific production instructions between processes based upon replenishment
principles. Kanban achieves this by governing both the timing of material movement
and the quantity of material conveyed.
[J • Serve as a visual control tool for production supervisors to determine whether production
is ahead or behind schedule. A quick look at the devices that hold kanban in the system

D (kanban accumulation posts) will show if material and information are flowing in timely
accordance to plan or if abnormalities have occurred.
r ··--,
t__J

r~--1
LJ
Appendix 107
 r--1
i I
L__J

LJ
• Establish a tool for continuous improvement. Each kanban represents a container
of inventory in the value stream. Over time, the planned reduction of the number of D
kanban in a system equates directly to a reduction in inventory and a proportional
decrease in lead time to the customer.
D
Types of Kanban
There are two main types of kanban: production instruction kanban (also known as make
D
kanban) and parts withdrawal kanban (also known as move kanban).
D
Types of Kanban
D
n
Production
instruction [ Signal kanban for scheduling

Kanban
batch processes D
Interprocess kanban
for internal purposes D
Parts
withdrawal [ Supplier kanban for
external purposes [l
The distinguishing feature between the instruction kanban and the withdrawal kanban is that
! l
the former is a signal to make something, while the latter signals that something needs to be
removed from inventory (which then triggers replenishment) and conveyed to a downstream
process. Each type of kanban has two further subdivisions (see Types of Kanban graphic). D
In-Process Kanban D
The in-process kanban is used to convey make instructions for small amounts (ideally one-
by-one production or at least one pitch corresponding to one pack quantity) to an upstream D
process. Typical uses include scheduling final production areas based upon withdrawal of
inventory from a market or a direct replenishment signal from a customer. An example of D
the type used at Apogee is shown on the next page.
D

D
n
LJ

[]
108
r- ·1
In-Process Kanban

Finished-goods Product Production

storage information information area area

Finished-goods Part#: Line address:

market
Assembly Cell #1
J 145091
Supplier: Part description:

Internal LH Exterior Mirror

[J Heated Black Location:

Final Assembly
Market location: Quantity: 10 Department
A4
R2

[]
Signal Kanban
The signal kanban is used to convey make instructions for large quantities to upstream batch
processes such as stamping presses and molding machines. In-process kanban would be less
effective in these applications due to the large number of cards required and the associated
time to handle them. Instead, signal kanban utilizes lot size in conjunction with markets to feed
D downstream processes while still allowing time for changeover work to occur at upstream
processes. There are three variations on the signal kanban: pattern production, lot-making,
and triangle kanban.

Pattern production is an effective method for scheduling processes when there is an optimal
D order or sequence for production to follow because of either the types of materials in use or
a changeover sequence that must be accommodated. For example, there often is an optimal
changeover sequence when processing different gauges of steel, chemical formulas, paint
colors, etc. In pattern production a basic sequence is established and adhered to, but the lot
i: size produced each time may vary. In this fashion a fairly stable production lead time and
I
..._____;!
time interval for producing every part can be established. The lead time for replenishment
for the pattern is then used to set the inventory level in the market.
D Another form of signal kanban, called lot-making, uses batch boards in conjunction with
ii inventory markets. Each item in the market has a kanban that is detached and returned to the
It__l '
upstream producing process as inventory is consumed. Once the kanban cards accumulate
to an established amount (trigger point), replenishment begins in accordance with the number
of cards. This form of signal kanban differs from an in-process kanban in that cards are grouped
into a production lot, rather than production occurring one card at a time.
[J

Appendix 109
 D
D
The most widely used form of signal kanban is known as the triangle kanban (the name comes
from its distinct triangular shape). Triangle kanban are used to schedule a batch process that
has substantial changeover times and machine cycle time significantly faster than the takt time
of production downstream. This kanban uses a lot size for production in conjunction with a
trigger point to replenish inventory and is frequently used for stamping, injection molding,
and similar processes. A key benefit of the triangle kanban is that only one kanban per part
number is created-multiple cards do not need to be managed.

An exampled of the type used at Apogee co regulate the molding process is shown: r
Triangle Kanban D
Part# Part description Market location []
15897 Internal bracket A2 R6 S2
)I

Date triggered* Lot size

LJ
Nov.23 3p.m. 400

Trigger point
200
D
Tool#
14 D
D
D
* The point in time at which the kanban is placed on the signal kanban rail. This aids production management in assessing I~

status of the process and reduces the possibility of getting kanban out of sequence. LJ

Interprocess Kanban

The interprocess kanban is used to signal the need to withdraw (move) parts from a storage
D
area and convey them co a downstream process within a facility. This type of kanban normally
is employed in conjunction with continuous-flow assembly cells that use a large number of
n
L i

components from either internal or external sources. A prerequisite for use of the withdrawal
kanban is the creation of a material market as well as the determination of storage quantities D
at line side. The intent of this kanban is to enable the storage of small quantities of material
at the final assembly area in order to maximize the space available for production. This in D
turn requires that the production assembly cell be supplied with frequent and regular delivery
n
c _J

[]
110
D
D
of small quantities of material. In order for this system to work, each location must have
.D dedicated positions and addressing mechanisms in place for ease of material movement.
An example of interprocess kanban used at Apogee to withdraw parts from the central
D market and convey them to the point of use in assembly cell #1 is shown:

D Interprocess Withdrawal Kanban

Supplier Product Point-of-use

D information area information area area

Supplier code: Part#: Line address:

ABC Station #1
Flow Rack#2

D Supplier: Part description:

Ajax Springs Spring

Line location:

C1
Market location: Quantity: 40 S2
D A2
R7
R2

S1
D
D Supplier Kanban
The supplier kanban is used to signal the need to withdraw parts from an external supplier for
D conveyance to a purchased-parts market or central market at the downstream customer. The
supplier kanban differs from the interprocess kanban in that it is used with external suppliers.

D In advanced applications the supplier kanban also contains information that is known as a
kanban cycle. An example of a kanban cycle is the notation 1 :4:2, which indicates that for
D a particular part number the supplier in one day will deliver that part and pick up kanban
four times, and the kanban picked up on any given cycle will be returned with the requested
[l parts two trips later. In this case, since there are four deliveries each day (at six-hour intervals),
the requested material will return in 12 hours. (Note that the supplier often is delivering other

D materials throughout a plant, adhering to additional kanban cycles, and delivering parts for
some locations while picking up kanban for others throughout the day.) Various kanban cycles
can be created depending upon the nature of the supplier's production process, the amount
D of supplier finished-goods inventory, and the supplier's distance from the customer.

D
Appendix 111
 D
D
An example of a supplier kanban used by Apogee to withdraw parts from Ajax Springs is shown:
D
SupplierKanban

Supplier Product Point-of-use

D
information area in_f_o_rm_a_ti_o_n_a_re_a _.. .. __ a_r_e_a__ ._
1.
Supplier code: Part#: Receiving
dock:
D
ABC

Supplier:
I X2174 I
Part description:
n
Ajax Springs Spring
Market location:
D
A2
Supplier location: Quantity: 40 R7
S1
D
123 Main Street
Acme, OH 1 :4:2
Kanban cycle:
[l

D
Temporary Kanban

Not depicted on the kanban classification chart is what is often referred to as the D
"temporary kanban." Typically the number of kanban cards are calculated and regulated
monthly by Production Control, and the quantity of cards is only subject to change D
monthly based upon changes in demand and lead time. Often, however, there are short-
term events that require additional kanban to be injected into the system for smooth []
production. Reasons to do this include the build up of inventory to adjust for the
differences in working days between customer and suppliers, or to make up for time
spent on die maintenance or machine repairs. The temporary kanban, as the name
D
suggests, is for one-time use only and should clearly show an expiration date. It is good
practice to color code or identify these kanban in some special manner so chat they are D
not accidentally kept in the system after their intended use is completed.
D
Kanban in Combination

In general, production and withdrawal kanban are used in combination to control pull D
production between processes where some work-in-process muse be stored in markets.
This arrangement works at Apogee as shown for regulating pull production between D
final assembly, the central market, and paint:

D
112
D
 r-,

LJ
D
Production and Withdrawal Kanban
D
D r··-®------~ :--··----~---------~
D
r-, LJ··E?;t. ~ LJ Every10
pieces
----------------~
Assembly

I \
L_J

5days
D
A similar use of production and withdrawal kanban in combination is shown among final
assembly, the central market, and molding:

Signal and Withdrawal Kanban

:-· -- -- --- --V·- --- -- -- ·: :··: -- -- -- -~------ -- - ~

: :
' I '

LJ-~ ~'.~;~. ~ !.~~~~-- -· I

t ! l

A,58m·~ I
D 1day

Finally, as a word of caution, kanban usage must be taken into consideration in conjunction
with conveyance routes in the facility. There must be a standard conveyance mechanism in
D place if kanban cards are to control the flow of production. The principle of conveyance in lean
manufacturing is to deliver mixed loads of products to downstream processes as requested in a
D timely, frequent manner. Normally a fixed-time variable-quantity style of conveyance is used
within a facility or between facilities to supply product in conjunction with kanban. This requires
that the conveyance route be on a precisely timed route with a repeatable standard pattern.
In certain instances between processes, a fixed-quantity variable-time method also may be
used to signal for certain type of material, particularly items chat are used less frequently.
In this latter case the key factor is to decide how to signal for the item requested and when
the signal needs to be made in order to prevent operations from running out of material.
D

D Appendix 113
 ,'I
r --,
I I
_l • .1

D
n
References []
Harris, Rick, Chris Harris, and Earl Wilson. 2003. Making Materials Flow.
Brookline: Lean Enterprise Institute.
. ..

Jones, Dan, and Jim Womack. 2002. Seeing The Whole.

Brookline: Lean Enterprise Institute.

Marchwinski, Chet, and John Shook, compilers. 2003. Lean Lexicon.

Brookline: Lean Enterprise Institute.

Rother, Mike, and John Shook. 1998. Learning To See.

Brookline: Lean Enterprise Institute.

Rother, Mike, and Rick Harris. 2001. Creating Continuous Flow.

Brookline: Lean Enterprise Institute.

iJ
L ..

D
D
,---,
I _!I
c, _

D
[]
rl
I I
! . __ J

r,
L.J

I'
L .. J
114
 -...
'
I
'

D
D

With gratitude to Ron Sacco, Bryan Shipway, John Shook, George Taninecz, Helen Zak,
and Design Continuum for their roles in the development of this workbook. D
D

D
D
D
D
0
D
© Copyright 2004 The Lean Enterprise Institute, Inc.
P.O. Box 9, Brookline, MA 02446 USA D
www.lean.org
Version 1.0, April 2004
ISBN 0-9743225-0-4 D
All rights to the text and illustrations reserved by The Lean Enterprise Institute.
0