Professional Documents
Culture Documents
TECHNOLOGICAL CHANGE
William J. Abernathy, Harvard Graduate School of Business Administration
ABSTRACT
A descriptive life cycle model of process development and innovation, as
previously reported, is used as a framework to analyze historical trends in the
technological development of U.S. automotive engine plants. Using the plant as
the unit of analysis, the author examines trends in product diversity, equipment
development, backward integration, and productivity improvement. Results sup-
port the original research framework and help to clarify the important impact of
process changes on the sources of productivity improvement, the capability for
product innovation, and the cost of change.
This paper is based on a forthcoming book, The Prodiicrivity Dilemma: Roadblock to 111-
mvarion in the Automobile hdustry, by William J. Abernathy. The underlying research was sup-
ported by the Division of Research, Harvard University, Graduate School of Business and by a
grant from the National Science Foundation, Office of National R & D Assessment DA-2057/.
Appreciation is expressed for the cooperation of the Henry Ford Museum, Greenfield Village,
and of individual managers of the Ford Motor Company, although they are in no way responsible
for the observations and conclusions of this paper.
607
608 DECISION SCIENCE [Vol. 7
nature since they must consider many ramifications. Change comes more slowly but
may be spurred by the development of a new technology, a sudden or cumulative shift
in the inputs, the requirements of the market or external regulation [l, pp. 378-3961.
Some of the elements of this model have been tested through research that
the model itself stimulated [ 5 ] . Other elements are indirectly supported by find-
ings from research that was conducted for other purposes [l]. The present
historical research on the development of engine plant technology at Ford was
undertaken to explore a few general hypotheses that could best be studied by
focusing on technological change in one process over time. A few of these initial
hypotheses were:
1. The technological development of a plant will be evolutionary in nature, and it will
proceed from conditions that are described as fluid toward a highly developed spe-
cific state.
2. A certain degree of evenness in development is required among the major aspects
that determine the plant’s technology (as given in Figure 1). For example, change in
important product line characteristics like product standardization, will not be inde-
pendent of other aspects such as equipment development.
3. Innovation and productivity advances will depend upon changes in the stage of de-
velopment.
A focus on engine plant development in the Ford Motor Company is par-
ticularly useful for several reasons. The U.S. automobile industry has been of
vital economic importance since its inception, and the engine plant is at the
heart of this industry. Ford has been a major factor in the industry since its
inception, and it has operated with management continuity for about seven
decades. Data about Ford is also accessible because many books, articles, and
reports have been written about the company and because a corporate archive
is maintained.
PROCESS DEVELOPMENT IN ENGINE PLANTS
The history of engine plant technology at Ford reveals a complex course
of development that involves change in many different aspects. By examining
changes in just six of the many different aspects, however, it is possible to cut
through some of the complexity, explore the questions that were posed for
research, and grasp the underlying pattern of process change that has determined
Ford’s competitive capabilities. Table 1 defines the five aspects that will be ex-
plicitly examined in this section. They are: 1) the diversity in product line that
is accommodated in one plant; 2) the degree of specialization in process equip-
ment; 3) the degree of process integration among different production opera-
tions; 4) the extent of backward integration in inputs that support engine pro-
duction; and, 5) the concentration and decentralization of engine plant opera-
tions. The following section considers these five aspects in aggregate and relates
the overall development of the plant to a sixth aspect-productivity improvement.
Figure 2A, 2B, and 2C present time series trends for the first five of these
six aspects. Data underlying these indices were obtained from open trade litera-
ture published in various years and from the Ford Archives. The particular
measures that are used to represent each of these four aspects are developed
below.
m
w
FIGURE 1 0
Summary of Hypothesized Relationships Among Aspecrs of a P l a n t
--.?
Frequent b novel p r o d u c c lllgh prodinat line d l v c r - Small s e a l c , l o c n t e l l ncnr
innova t i a n - m a r k e t s t imu- s i t y prilduccd ca c u s c me i c e c l l n o l o g y riourcc or "Dfr.
lnted. order. Lou l c v c l o [ bockvnrd
verrlcnl integracion.
Damlnnnt design a c h i e v e d
F a c i l i t l c s locatcd t o
C o n r r a l nchluvcd by nean8 achieve l a w f a c t o r I n p u t
of g o a l s c t t l n g . Iiicr- COSCE, to mlnimirc disrup-
llf glily f i ~ a n d o r d i z e d n r c h y and rulrs YS t h e c l o n and f n c i l l t s t c
product v i t h f c v u J o r f r c q u c n c y oi ch.tngc distribution.
opcions decr'uxs.
Large scale f e c l l i t i e s
Direction -- specialired to p a r t i c u l a r
Comodity-like product Burcaucratic, v e r t i c a l l y
of trchnologics. capacity
s p e c i f l e d by rechnical Integrated production i n t o g r . t e d . nnd h i r r -
Tans 1t i o n
Cost s t i m u l a t e d incremen- pacumc.t e n . p r o c e s s d c b i g n c d ns a nrchically u r e a n i r o d i n c r e a s e s a c h i e v e d only
t a l innovation predomin- "syatcn>." r r l t h f v n c t i o n n l cmplrasis. by d C s 1 ~ u i n gnew f a c i l l t i e a
s t e r . Novel chnnges in-
v o l v e s i m u l t a n e o u s prod- Labor t ~ i b k sp r c d o m i n a n t l y
L C C and proccss ad;?ta- ones <if syotcms nonitor-
ing.
SPECIFIC BOUNDARY
TABLE 1
1. Product Diversity of the Engine Plant-The number of different automotive
passenger engines that were produced by one engine plant in various years.
Engines are counted as distinct units if they mechanically differ in respect
to bore, stroke, cylinder alignment, or valve configuration.
2. Process Equipment Specialization-The maximum number of different
operations that are performed either simultaneously or automatically at one
station by one machine in engine production.
3. Process Zntegration-The number of different stations in the production
process that are linked and integrated together with transfer equipment so
that they function as a continuous automatic process unit.
4. Backward Integration-The degree to which corporate backward integration
extends toward raw material sources for the major materials and components
that are used in engine production. This is approximated as the average
number of major process stages leading up to engine assembly (e.g., ore
extraction, iron and steel making, part forming or foundry, rough machin-
ing, finish machining) that Ford owned and operated at a particular time
for a sample of 12 engine parts.
5. Process Concenlration-An indication of the extent to which major engine
manufacturing processes are concentrated in one facility under common
operating control, as opposed to being decentralized in separate specialized
plants. The average number of backward integration steps (from 4 above)
that are managed as a common unit with the engine plant are taken as
an indication of concentration.
FIGURE 2
Trends in Five Plant Indices
A. Product Divsraity
New Model Introduction Introduction of
8.
i!
Bquipunt Specialization k Procssa Intr&mtion
l50. 150
p‘ - - - - -- - - --
+r
.rl
d 125. :2Procese
+. Integrntion
Proceas Concentration
2.
The most plausible explanation for the trends in diversity would seem to
originate from developments in engine plant technology that have acted to re-
duce the plant’s capability for diversity. The narrow range of variation in actual
diversity and the presence of consistent long term trends toward lower diversity
since the 1950’s would seem to suggest that changes in product diversity may
be an important parameter of technological development. The only known
research that deals generally with product diversity is the work of sociologists
on technology and organizational structure [3] [4] [6]. The trends observed at
Ford are consistent with this research; but, given the importance of the impli-
cations for capacity planning, it is surprising that more research has not ad-
dressed the basic considerations involved.
Equipment Development and Process Integration
The two curves in Figure 2B reveal several underlying tendencies in the
development of engine plant process equipment and at the same time provide
insight into the forces that reduce product line diversity, increase the cost of
change, and cause product and process innovation to be linked. The solid curve
traces progressive improvement in the number of different process operations
that are performed by a single machine at one work station. For example by
1912, a single machine was in use that performed 24 operations simultaneously
(e.g., drill 5 holes from 3 different directions, ream, mill, etc.). Although the
reported number is for the one machine in the process that performed the most
operations, this maximum number is also a general indicator of capability levels
for there were typically many other machine tools in the process that approached
this maximum level. This number can also be interpreted as a general index of
specialization in machining capability. As more functions are added to machines,
their use becomes increasingly limited to a particular product and a particular
operation in the overall process.
The build-up in equipment specialization as reflected by this index nicely
mirrors the broad range of changes that accompanied the development of process
equipment technology in Ford’s engine plants. Before 1910, the manufacturing
process was essentially a job shop. Engine design changed frequently; parts
were routed over long indirect paths to machine groups that were arranged by
type of equipment. General purpose equipment was utilized even though ma-
chine tools designed particularly for engine manufacture were available from
U.S. machine tool suppliers.
As demand for the Model T grew, there was intense pressure to expand
capacity through process innovations. Under this pressure, the production
process evolved toward a very strict line flow configuration through a continu-
ing series of changes that improved balance, reduced cycle times, and introduced
mechanical transport and moving assembly in a variety of applications. Ford’s
own machine tool design group and outside suppliers responded by incorporat-
ing more and more functions in individual machine tools. Individual machine
tools became “islands of growing mechanization” in the line flow process. By
the time demand for the model T peaked, a maximum of 48 distinct functions
were performed by one machine. Many other machine tools approached this
614 DECISION SCIENCE [VOI. 7
limit. In a very real way the pattern of equipment development was linked with
the organization of the production process. An orderly and highly regular pro-
duction process was essential to machine development, but equipment develop-
ment in turn reinforced and enabled this pattern.
This same mode of equipment development was continued and extended
after the Model A was introduced and again after the V-8 was introduced in
1932. A peak level of specialization equal to 83 was reached during the depres-
sion, but by this time a new trend in mechanization was beginning to ernerge-
the transfer machine.
Shortly after the new and highly complex V-8 engine block was placed in
production, Ford introduced the first automatic transfer machine. It only moved
the work piece a few inches, but it represented a change in the thrust of process
innovation from unit equipment advancement to systems integration. With this
innovation, the principal mode of mechanization changed from a growth of
individual “islands of mechanization” to a joining of highly mechanized units.
The dotted curve in Figure 2B traces progress in this second mode. This graph
indicates the maximum number of separate machine tools that are integrated to
form a single process unit that performs automatically.
As the graph indicates, progress in transfer line integration was rapid.
When Ford’s innovative Cleveland Engine plant was put in operation in 1951,
most major machinery operations on large parts were controlled automatically
by transfer lines. The equivalent of 150 separate machine tools were integrated
into one line. Since that date, the mode of equipment development has been to
extend integration: transfer lines have been extended further into manual as-
sembly operations, and feedback and control have been added to correct for
tool wear, locate failures automatically, inspect parts, etc.
These developments in equipment capability have acted to make product
change and process change interdependent and to increase the cost of change.
In the early years before the Model T when the production process was fluid,
almost any engine could be produced without significant impact on the pro-
duction process. Five or six engines for models A through S were designed
and produced before the Model T became a mass produced car, and there was
little impact on the production process. This flexibility decreased as equipment
capability advanced. For example, the introduction of both the Model A and
the V-8 engines required that more than half of the machine tools in engine pro-
duction be scrapped or redesigned. The change in ability t o accommodate new
product design is also reflected in the specialization of the process equipment.
The introduction of both the Model A and V-8 engines is associated with an
abrupt dip in the specialization index. Today, with the introduction of a new
engine, the entire transfer line is typically replaced and often stored so the engine
can be put into production again. A more advanced line i s introduced with the
new engine at a cost of more than $20 million.
To summarize, the course of development in process equipment has been
associated with and perhaps instrumental in changing basic capabilities. Equip-
ment development is tied closely to the organization of the production process,
the level of interdependence between product and process, and the cost of
change. It would seem that these fundamental changes help to explain why the
19761 PRODUCTION PROCESS STRUCTURE 615
capability to accommodate diversity has decreased and why there has been a
reduced rate of product change. Although the total cumulative impact of these
developmental trends has been great, they have come about largely through
incremental and evolutionary change.
Backward Integration
The final graph, Figure 2C, is concerned with changes in Ford’s backward
vertical integration in the streams of material and components that support
engine production. To form a rough indication of backward integration, a
generalized set of eight process steps has been defined to represent the major
groups of production stages in the chain that converts raw material into final
engine parts (ore extractions, smelting or iron production, part forming, etc.).
The differences between engine parts (carburator, cylinder block, etc.) in
the level of backward integration at any time can be roughly represented by
the number of steps through which backward integration extends. The average
number of backward integration steps for a sample of 12 engine parts has been
used to construct the index of backward integration. Relative changes in this
index provide a basis for making rough comparisons of period to period changes
in backward integration.
Figure 2C graphs changes from period to period in two slightly different
versions of this index. The solid line labeled Backward Integration reflects the
full integration level of the corporation, irrespective of whether the included
process operations are centralized or decentralized. The dashed curve titled
Process Concentration includes only those process operation steps that are lo-
cated with and managed as part of the engine plant. This latter index helps to
show how much of the overall manufacturing process for engine parts was cen-
tralized and managed as part of an engine plant. It reflects the degree to which
process operations are concentrated in one plant.
The profile of change in these two indices as illustrated in Figure 2C calls
attention to two important characteristics of engine plant development. The first
is a dependency between product innovation and backward integration. The
second is the plant concentration-decentralization hump that occurs over time.
The shallow deflections in the indices that parallel the introduction of the
Model A and V-8 engine are a surface indication of very dramatic and abrupt
changes in the composition and extent of backward integration. The consistent
effect of major model changes has been to destroy channels of backward inte-
gration. The extent of the impact is masked in the present data because rapid
re-integration moves accompanied major model changes. For example, the
Model A incorporated new and advanced engine components (aluminum pistons,
carburators, starters, and many alloy steels) that were purchased from innovative
suppliers even though Ford had previously integrated backward to produce
corresponding parts for the Model T. The same impact on backward integration
is also evident for the first mass produced V-8 engine, in later post World War
11 engine introductions, and to a much greater extent in major model changes
for bodies in all periods. As the technology has matured, however, the impact
of model changes on backward integration levels seems to have decreased.
The degree of backward integration that a company achieves is often dis-
616 DECISION SCIENCE [Vol. 7
FIGURE 3
Stage of Development and Labor Hours per Engine
/
Labor H a w s per Engine
\
\ Korean War
\
---
L
Truck
P
‘+- ’I-
-v”r
5.
sented by the inverted scale along the left margin. For this scale the origin repre-
sents the most advanced state. The aggregate level of engine plant development,
as depicted by the curve in various years, reflects varying historical conditions.
It is a weighted index of equipment development, integration, process flow con-
figuration, product standardization, labor task characteristics, etc. Although
there is room for disagreement about the exact stage of development in any year,
there should be little question about the overall shape of the curve. It is deter-
mined by major structural changes in the important aspects that shape the
technology of engine plants. The early years were marked by rapid change, and
this was followed by a plateau. In the early 1950’s, the introduction of extensive
integration through the use of transfer lines caused a shift in conditions.
The second curve in Figure 3, the dotted upper line, plots actual changes
in total labor hours per engine. Changes in this curve reflect variations in labor
productivity. By comparing changes in the two curves, one can gain insights
about the relationship of productivity and development.
Labor hours per engine were reduced rapidly during the early years of de-
velopment when only one relatively standardized engine was produced (Model
T). There was a corresponding rapid advance in the engine plant’s level of de-
velopment. The minimum was reached at least two years before the Model T
was discontinued. The rate of engine design changes then increased, beginning
with the changes that were made to incorporate a starter on the Model T. With
618 DECISION SCIENCE [Vol. 7
increased product change rates and diversity, the rate of labor utilization also
increased. For the next two decades, the base line trend in the number of labor
hours per engine followed a steady upward movement. Labor inputs also are
seen to react sharply to new model introductions.
The rate of labor utilizations was decreased substantially in the early 1950’s
through the introduction of engine plants that incorporated the new transfer
line technology. Since this time, new engine plants have been designed specifi-
cally for each new major engine design. Thus, the process technology is exactly
matched to the product. The cost involved in this mode of change is reflected
more heavily in investment than in increased labor hours.
Although the rate of change has slowed since the 1950’s, the overall trend
would still seem to be moving toward higher labor productivity rates and a
higher degree of process development.
These data show that throughout the entire period of seven decades, there
has been a strong relationship between labor productivity improvement, process
development, and product innovation. There is a tendency in the literature to
assume that marketing strategy can be divorced from productivity issues and
production policy. The present data show that they are all intimately related.
A high rate of product innovation seems to have induced a reversal in the level
of process development, thereby increasing labor input rates. Conversely, a high
rate of productivity improvement would seem to constrain product innovation
by making change more costly and by restricting the range of product line
diversity.
The principal mode of rapid productivity improvement has changed dramat-
ically as the engine plant evolved from the fluid stage of development toward a
highly advanced state. I n the fluid stage, a learning curve phenomenon appears
to be principally important. In this case the long term learning curve is a period
of rapid transition in the plant’s stage of development, Important skills that were
initially labor and management embodied are transferred to capital embodied
capabilities. During this transition, major process innovations in equipment
typically take the form of new machines that are a mechanical or electronic
analogy to human behavior.
I n the more advanced stage of development, productivity gains are achieved
principally through major investments in entire systems or plants. Advances
are realized through effectively integrated or optimally designed systems for a
particular product. These productivity gains are sometimes referred to as static
economies of scale since they depend upon investment, can be purchased, and
need not be achieved through the dynamic or learning curve processes.
CONCLUSIONS
Present observations pertain only to one case, but they are considered im-
portant because they establish the feasibility of a set of hypotheses about the
way that a plant and its technology evolve over time.
1. The use of the plant and its technology as a unit of analysis appears to
provide fruitful focus for research. This level of abstraction seems to represent
capabilities that endure for long periods and have a force of their own in in-
19761 PRODUCTION PROCESS STRUCTURE 619
[2] Abernathy, William J. and James Utterback. “Innovation and the Evolving Structure of
the Firm.” Harvard Business School Working Paper 75-18, June, 1975.
(31 Harvey, Edward. “Technology and the Structure of Organizations.” Americaii Sociological
Review, Vol. 3, No. 2 (April, 1968), pp. 247-259.
[4] Hickson, D. J., et at. “Operations Technology and Organization Structure: An Empirical
Reappraisal.” Admbiistrative Sciences Quarterly, Vol. 14, No. 3 (Sept., 1969), pp, 378-397.
[5] Utterback, J. and W. J. Abernathy. “A Dynamic Model of Process and Product Innova-
tion.” Omega, Vol. 3, No. 6 (1975), pp. 639-656.
[6] Woodward, Joan. Zitdustrial Orgaiiizariori: Theory arid Pracfice. London: Oxford Univer-
sity Press, 1950.