OSA Production Volumes (q) - Duke University`s Fuqua School of

Learning in a Multi-Product Overseas
Manufacturing Environment
Carolyn Denomme 1, Erica Fuchs 1, Linda Argote 2, Dennis Epple 2
1 Dept.
of Engineering & Public Policy, Carnegie Mellon University
2 Tepper School of Business, Carnegie Mellon University
Organization Science Winter Conference
February 6, 2010
Learning & the Impact of Product Mix
• Organizational learning can contribute significantly to the performance
of firms. Yet, organizations vary significantly in their rate of learning
and knowledge transfer (Dutton & Thomas 1984; Argote and Epple 1990, Bohmer &
Edmonson 2001; Pisano, Bohmer & Edmondson 2001)
•
•
•
Majority organizational learning research on production has focused
on a small number of products with minor variations (e.g. aircraft,
ships, & trucks) (Alchain 1963; Rapping 1965; Argote & Epple 1990, Bernard,Redding & Schott
2006), 41% of U.S. firms, 90% of U.S. output: multi-product (Benkard 2000)
Theory: Organizations learn more from diverse experience than
homogenous experience (Haunschild and Sullivan 2002; Schilling, Vidal, Ployhart &
Marangoni 2003) However, lack of regular production can also lead to
organizational forgetting (Shtub 1993; Benkard 2000)
Empirical studies that exist, have looked only at learning across
product generations
-
Results mixed: in semiconductor production - spillovers weak across
generations (Irwin 1994); aircraft production - substantial spillovers from one
generation to the next (Benkard 2000)
2
Learning & the Impact of Turnover
•
Only 28% of manufacturing value added today is in the U.S.
(UNIDO 2007)
•
•
In industrializing nations, where production is moving, firms
suffer from high employee turnover (UNIDO 207, Hewitt 2008)
Past studies have shown conflicting stories on the effect of
turnover on learning:
-
New members can bring and stimulate the creation of new
knowledge (Gruenfeld, Martorana & Fan 2000; Choi & Thompson 2005)
-
Loss of members leads to a loss of experience and can disrupt
established processes (Bluedorn 1982; Argote & Epple 1990)
-
Theoretical ‘optimal’ rate of turnover
(Abelson & Baysinger 1984;
Glebbeek & Bax 2004)
-
Given high turnover, the impact can at times be mediated by
through detailed definition of tasks/processes (Ton & Huckman 2008)
3
Research Question
•
How do product mix and employee
turnover influence production learning
rates?
4
Learning Model
ln(qt) = β0 + β1ln(Qt-1) + β2ln(Lt) + β3ln(Ht)+ β4At + β5Dt + β6Mt + β7Pt + εt
Variable
Description
Expected Sign
ln(qt)
Productivity: Current Period Production
N/A
ln(Qt-1)
Experience: Cumulative Past Production Volume
+
ln(Lt)
Input: Labor Hours for Line Workers
+
Ht
Product Mix: Herfindahl Index
At
Fraction of New Arrivals
?
Dt
Fraction of Departures
?
Mt
Fraction of Moves between Workgroups
?
Pt
Fraction of Promotions
?
1, no product mix
0, high -product mix
Level of analysis (t): by week
5
Study Context: U.S.-Owned Multinational
•
•
•
One of the leading revenue
earners in the $3.3B
optoelectronic component
industry (Ovum-RHK 2007)
Following the telecom bubble
burst, they began the move
manufacturing offshore to
Southeast Asia in 2001
Since opening, their offshore
facility has produced over
2,000 variations of
subassemblies and over
4,500 variations of optical
transceiver modules
Site Visit & Data Collection Overview
Preliminary
Site Visit
2 days
Data Collection
Site Visit
7 days
Production Floor
Observation
7 hours observation
2 hours participation
Semi-structured
Interviews
3 with VP
4 with engineers
4 with managers/trainers
Field Notes
55 pages
Archival Data
Records
15GB of data from
production and human
resources databases from as
early as 2000
6
Product and Process Selection:
Optoelectronic Transceiver
A. Optical Subassembly
(OSA)
B. Module Components
PC Board Assembly
(PCBA)
Process
Steps
Process Steps
Optical Sub Assembly (OSA)
Optical Alignment
Light Cure Epoxy & UV Curing
Backfilling & Oven Cure
Final Optical Test
Electrical Test
Visual Mechanical Inspection
Module Assembly & Testing
Transmitter OSA
(TOSA or Laser)
OSA Lead Trimming & Solder
Inspection
Receiver OSA
(ROSA or Photodiode)
C. Assembled Module
Module Housing
Shell Assembly & PC Board
Solder
Enclosure
Initial Assembly
Test
Temperature
Cycling
ER Setup
Cold Temp Test
Shipment
Hot Temp Test
Final Verification Test
7
Quantitative, Archival Data Sources
Production
Hard Copies of
Data Source Floor Tracking Weekly Hours
System
Report
Data Details
Start and
output
production
volumes
broken down
by:
shift, product,
process step,
test station,
and operator
Data
Availability
2004 - 2009
Calculated
Variable(s)
Per-step
production
volumes &
product mix
(qt, Qt-1, Ht)
Human
Resources
Employment
Database
Date that each
of the factory’s
Weekly hours
11,742
worked by rate employees was
and division hired, resigned,
(OSA, Module, promoted, or
Engineering)
changed
shifts/superviso
rs
2004 - 2008
Labor hours
(Lt)
2001 - 2008
Arrival rate,
departure rate
& experience
(At, Dt)
Warehouse
Database
Sales &
Shipment
Reports
Weekly reports
of the volume
of each product
type which
finished
production and
was sent to the
warehouse
Volume of each
order and date
the order was
placed,
requested,
scheduled for
shipment, and
actually
shipped
2006 - 2008
2000 - 2009
Indicator of
production
backlogs
Shipment
dates,
instrument:
sales order
dates, past
production vol.
(Qt-1)
8
Human Resource Database: Track
positions, experience, quality over time
Employee
Characteristic
Metric
Line Workers 6 positions including operators, line leaders
Line
Supervisors
Position
We group 118
Engineers
positions into 5
categories.
Managers
Other
29 positions including trainers, technicians, specialists
20 positions including industrial engineers, test engineers
47 positions including general managers, accountants,
directors, VP
16 positions including assistants, security guards, nurses
Experience
Length of time at firm
Supervisor, promotion, product, and process history within the firm
Quality
Specific position (i.e. Engineer vs. Senior Engineer)
Training quality ranking, Defect statistics
•
… Equally detailed data on product and process
characteristics
Organizational Learning: Module
Experience: ln(Qt-1)
Labor Hours: ln(Lt)
1
0.153*
(0.081)
0.481*
(0.259)
Product Mix: Ht
2
0.155**
(0.071)
0.691**
(0.245)
0.187
(0.125)
Arrival Fraction: At
Departures Fraction: Dt
Promotion Fraction: Pt
Move Out Fraction: Mout,t
Move In Fraction: Min,t
AR(1)
Observations
R2
0.408***
(0.0001)
182
0.521
0.379***
(0.0001)
182
0.541
3
0.209***
(0.069)
0.405*
(0.240)
1.417**
(0.645)
1.263
(1.638)
47.535**
(26.524)
3.684***
(1.198)
1.817**
(0.735)
0.351***
(0.0001)
182
0.563
4
0.249***
(0.075)
0.326
(0.243)
0.167
(0.112)
1.504**
(0.665)
1.665
(1.647)
44.869*
(23.371)
3.996***
(1.090)
1.770**
(0.815)
0.323***
(0.001)
182
0.564
Model is a two stage least squares with Newey West Standard Errors. Instruments include a 7 week
lagged average shipment volume, 2-week lagged labor hours, and 2-weeks lags of all employeerelated
Note: *,**,***
denote variables
statistical significance at the 10%, 5%, and 1% levels
respectively.
Discussion
•
•
The module production line learns from past experience
at a learning rate of 84% (Dutton & Thomas 1984, p=0.55-1.07)
-
Fraction of arrivals, promotions, and internal moves in
a given week all increase productivity
-
Product mix has a non-significant effect, and requires
more fine-grained analysis
-
Notably, we did not find statistical significance on the
fraction of departures.
More work needs to be done to investigate the
underlying phenomenon behind these personnel
movements and potential interactions with
product/process characteristics.
13
Upcoming Work
•
Does product complexity affect impact of turnover?
-
•
•
Is process mix a better indicator than product mix?
How is the firm allocating labor? Which people are being
moved? Lost?
-
•
Complexity measure: number of tests per product,
laser speed & wavelength, level of customization
Labor experience: type of employee, days with firm,
quality of work
How does simultaneous production in U.S. affect
learning rate in offshore facility? (Galbraith 1990: co-production)
14
Long-Term Research:
NSF SciSIP/IOS Grant
•
Organizational learning and knowledge
transfer across product, workgroup, and
geographic boundaries
15
Geographic Distance and Knowledge
Transfer
•
Over the past three decades firms have been
increasingly globalizing (I.e. “offshoring”) their productive
activities -- both manufacturing and white collar work (Oviatt
& McDougall 1997, Lewin & Couto 2007)
•
While such global expansions can involve outsourcing
(or moving production activities outside firm boundaries),
they also can occur within a firm’s boundaries (Anderson et al
2007)
•
This second type of offshoring has led to a growing
number of geographically distributed technical teams
need to coordinate across spatial, temporal, and
configurational dimensions (Espinosa et al. 2003, Oleary & Mortensen 2010)
16
Geographically Distributed Teams
•
•
A long history of work has demonstrated that
geographic distance can limit the flow of
knowledge (Allen 1977, Gibson & Gibbs 2006)
In addition, factors other than distance including
language and cultural differences (Hofstede 1980, Kogut &
Zander 1992)), lack of shared context , electronic
dependence, change in team structure (Gibson & Gibbs
2006), limited communication due to time zone
differences (Espinosa et al. 2007), and limited observation
of distant processes (Tyre & vonHippel 1997, Hinds and Mortensen 2005)
may be equally or more significant
17
New Product Development
•
The challenges associated with knowledge
discovery and transfer in geographically
distributed teams come particularly to the
forefront in the process of NPD
-
Greater degree of process, marketing, creative, and
technical uncertainty than found in many settings
(Anderson et al. 2008)
-
Difficult to design norms, practices, and procedures
that employees can uniformly apply to all situations
(Anderson et al. 2008)
18
Research Question 2
•
How is…
-
the changing percentage of New Product
Development (NPD) activities in the U.S.
versus Malaysia
-
and associated management policies enacted
on the globally distributed NPD team during
that period (end co-production, sold directly off NPD line)
-
… affecting learning rates in NPD and on the
main production floor once the new products
enter mass production?
19
Learning in NPD:
Three Performance Measures
•
•
•
Time to complete NPD processes
-
Expect relationship between 5 NPD steps
Will relationship differ with distribution of team?
Learning rate for the product while it is on the
dedicated NPD line
Learning rate for the product on the main product
floor once introduced into mass production
21
NPD: An exciting context
•
•
•
What types of problem-solving are most difficult
across distance?
What management policies are most effective?
-
Co-production beneficial? (Galbraith 1990)
Learning vs. performing orientation? (Bunderson & Sutcliffe 2003)
Given the fact that distributed environments will
hinder knowledge creation and transfer, which
activities should be grouped together?
22
Acknowledgments
•
•
•
•
Sloan Industry Studies Site Visit Grant
Berkman Faculty Development Grant
Innovation & Organization Sciences NSF Grant #0622863
Science of Science & Innovation and Innovation &
Organization Science NSF Joint Grant #0965442
23
Learning Model
ln(qt) = β0 + β1ln(Qt-1) + β2ln(Lt) + β3ln(Ht) + β4At + β5Dt + β6Mt + β7Pt + εt
Variable
Description
Expected Sign
ln(qt)
Productivity: Current Period Production
ln(Qt-1)
Experience: Cumulative Past Production Volume
+
ln(Lt)
Input: Labor Hours for Line Workers
+
Ht
Product Mix: Herfindahl Index
-
At
Fraction of New Arrivals
?
Dt
Fraction of Departures
?
Mt
Fraction of Moves between Workgroups
?
Pt
Fraction of Promotions
?
Level of analysis (t): by week
N/A
24
Product Characteristics: Complexity
Product Characteristic
Increasing Degrees of Complexity
Form Factor
Laser Type
Laser Speed (Gbit/sec)
SFF
LED
FP
SFP
SFP+
VCSEL DFB CWDM DWDM
0.100 0.622 1.25 2.488 4.25 10.7
Laser Reach (km)
0.150
Laser Wavelength (nm)
850 1310 1470 1550 1610
Mode
Laser Cooling
0.550
10
Single
Multi
Uncooled
Cooled
Process Count
10 - 20
Part Count
20 - 30
Level of Customization
•
80
150
# parts not common to broader product family
… Process (test) commonalities even more insightful?
25
Annual & Monthly Turnover Over Time
Annual plant turnover
of all workers declines
from 35% to 15% from
the opening of the plant.
Module
Plant
OSA
Line Workers
Monthly module line worker
turnover declines significantly
after mid-2004
Line Departments
Engineers
26
Organizational Learning & Product Mix
Experience: ln(Qt-1)
Expected: +
Labor Hours: ln(Lt)
Expected: +
2SLS
2SLS + Herf
2SLS w/ NW
2SLS + Herf w/ NW
0.118*
0.155**
0.118*
0.155**
(0.069)
(0.067)
(0.063)
(0.071)
0.669**
0.601*
0.669***
0.601**
(0.304)
(0.308)
(0.239)
(0.245)
Product Mix: Ht
Expected: +
AR(1)
Observations
R2
0.187**
0.187
(0.089)
(0.124)
0.395***
0.379***
0.395***
0.379***
(0.0000)
(0.0000)
(0.0000)
(0.001)
184
184
184
184
0.529
0.541
0.529
0.541
Two stage least squares model instruments include a 7 week lagged average
shipment volume, and 2-week lagged labor hours.
Note: *,**,*** denote statistical significance at the 10%, 5%, and 1% levels
respectively.
27
References
Abelson, M. A. and Baysinger, B. D. T., 1984. "Optimal and Dysfunctional Turnover: Toward an
Organizational Level Model." The Academy of Management Review 9(2): 331-341. Alchain, A., 1963.
"Reliability of progress curves in airframe production." Econometrica 31: 679-693.
Almedia, P. and Kogut, B., 1999. "Localization of knowledge and the mobility of engineers in regional
networks." Organization Science 45(7): 905-917. Argote, L. and Epple, D., 1990. "Learning curves in
manufacturing." Science 247: 920-924.
Benkard, C. L., 2000. "Learning and forgetting: The dynamics of aircraft production." American
Economic Review 90(4): 1034-1054. Bernard, A., Redding, S. and Schott, P. K., forthcoming. "MultiProduct Firms and Product Switching." American Economic Review. Bluedorn, A. (1982). The theories
of turnover: Causes, effects, and meaning. Res. Sociol. Organ., 1, 75-128. Bohmer, R. and
Edmonson, A., 2001. "Organizational Learning in Health Care." Health Forum: 32-35.
28
References
Choi, H. S. and Thompson, L., 2005. "Old wine in a new bottle: Impact of membership change on
group creativity." Organizational Behavior and Human Decision Processes 98: 121-132.
Dutton, J. M. and Thomas, A., 1984. "Treating progress functions as a managerial opportunity."
Academy of Management Review 9: 235-247.
Glebbeek, A. C. and Bax, E. H., 2009. "Is high employee turnover really harmful? An empirical test
using company records." Academy of Management Journal 47(2): 277-286.
Gruenfeld, D. H., Martorana, P. V. and Fan, E., 2000. "What do groups learn from their worldiest
members? Direct and indirect influence in dynamic teams." Organizational Behavior and Human
Decision Processes 82(1): 45-59. Haunschild, P. and Sullivan, B., 2002. "Learning from complexity:
Effects of airline accident/incident heterogeneity on subsequent accident/incident rates."
Administrative Science Quarterly 47: 609- 643.
Hewitt Associates. (2008, Oct. 24). Hewitt's Shenzhen City Compensation and Benefits Study
Provides New Market Data and Insights. Retrieved Dec. 15, 2008, from
http://www.hewittassociates.com/Intl/AP/enCN/AboutHewitt/Newsroom/PressReleaseDetail.aspx?cid=5759
29
References
Kane, A. A., Argote, L. and Levine, J. M., 2005. "Knowledge transfer between groups via personnel
rotation: Effects of social identity and knowledge quality." Organizational Behavior and Human
Decision Processes 96: 56-71.
Ovum-RHK, 2007. Optical Components Market Share Data. Liu, K., Inniss, D., Harris, T. and Ichida, T.
Pisano, G. P., Bohmer, R. and Edmondson, A. C., 2001. "Organizational differences in rates of
learning: Evidence from the adoption of minimally invasive cardiac surgery." Management Science
47(752). Rapping, L., 1965. "Learning and World War II production functions." Review of Economics
and Statistics 47: 81-86. Schilling, M. A., Vidal, P., Ployhart, R. E. and Marangoni, A., 2003. "Learning
by doing something else: Variation, relatedness, and organizational learning." Management Science
49: 39-56. Shtub, A., Levin, N. and Globerson, S., 1993. "Learning and Forgetting Industrial Skils: An
Experimental Model." The International Journal of Human Factors in Manufacturing 3(3): 293-305.
Ton, A. and Huckman, R. S., 2008. "Managing the impact of employee turnover on performance: The
role of process conformance." Organization Science 19(1): 56-68. UNIDO. (2007). The International
Yearbook of Industrial Statistics. Edward Elgar.
30
Our Context: ‘Corporation X’
•
•
‘Corporation X’: A U.S. firm
-
Leading revenue earner in the optoelectronics industry: NSF hightech industry, $17B market in optical communications, ‘canary in
the coal mine’ for growth in ICT
-
Offshored manufacturing in 2001, began transferring New Product
Development (NPD) in 2003
-
High product mix - 4154 product variants, High turnover
Our focus: Organizational learning and knowledge transfer
across (1) product, (2) workgroup, and (3) geographic
boundaries
-
Remove traditional culprit: firm boundaries
31
Learning & the Impact of Product Mix
•
Most research focuses on production of a small number of
products with minor variations (e.g. aircraft, ships, & trucks)
(Alchain 1963; Rapping 1965; Argote & Epple 1990, Bernard,Redding & Schott 2006), but
41% of U.S. firms, 90% of U.S. output: multi-product (Benkard
2000)
•
Theory: Organizations learn more from diverse experience than
homogenous experience (Haunschild and Sullivan 2002; Schilling, Vidal,
Ployhart & Marangoni 2003) However, lack of regular production can
also lead to organizational forgetting (Shtub 1993; Benkard 2000)
-
Larger the lot size, the more learning during production of each
product, but also the longer the breaks between production periods
(Shtub 1993)
•
Empirical evidence: semiconductor production - intergenerational learning spillovers weak (Irwin 1994); aircraft
production - substantial spillovers from one generation to the
next (Benkard 2000)
32