Paper to be presented at the DRUID Summer Conference 2007 on APPROPRIABILITY, PROXIMITY, ROUTINES AND INNOVATION Copenhagen, CBS, Denmark, June 18 - 20, 2007 CHALLENGING THE S-CURVE: PATTERNS OF TECHNOLOGICAL SUBSTITUTION Brice Dattee Tanaka Business School, Imperial College London [email protected]Abstract: This paper revisits the relevance of the S-curve representation of technological substitution. I argue that the smooth S-curve does not properly account for the complexity of the phenomenon. First, I observe historical cases with patterns of substitution more complex than what the classical S-curve suggests. Second, I show that a broadened theoretical framework at the system level is required to better understand the underlying dynamics of technological substitutions. Third, I identify bifurcation points between generic substitution trajectories and show how they can be combined into longitudinal sequences. Finally, the results are discussed and strategic implications are drawn. JEL - codes: O33, O32, M00
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Paper to be presented at the DRUID Summer Conference 2007
on
APPROPRIABILITY, PROXIMITY, ROUTINES AND INNOVATIONCopenhagen, CBS, Denmark, June 18 - 20, 2007
CHALLENGING THE S-CURVE: PATTERNS OF TECHNOLOGICAL SUBSTITUTION
Brice DatteeTanaka Business School, Imperial College London
Abstract:This paper revisits the relevance of the S-curve representation of technological substitution. I argue that thesmooth S-curve does not properly account for the complexity of the phenomenon. First, I observe historicalcases with patterns of substitution more complex than what the classical S-curve suggests. Second, I show thata broadened theoretical framework at the system level is required to better understand the underlying dynamicsof technological substitutions. Third, I identify bifurcation points between generic substitution trajectories andshow how they can be combined into longitudinal sequences. Finally, the results are discussed and strategicimplications are drawn.
JEL - codes: O33, O32, M00
1
Challenging the S-curve: Patterns of
Technological Substitution
Abstract:
This paper revisits the relevance of the S-curve representation of technological
substitution. I argue that the smooth S-curve does not properly account for the complexity of
the phenomenon. First, I observe historical cases with patterns of substitution more complex
than what the classical S-curve suggests. Second, I show that a broadened theoretical
framework at the system level is required to better understand the underlying dynamics of
technological substitutions. Third, I identify bifurcation points between generic substitution
trajectories and show how they can be combined into longitudinal sequences. Finally, the
results are discussed and strategic implications are drawn.
Keywords: S-curve, technological substitution, trajectories, bifurcation, system dynamics
1. Introduction
The S-curve has been at the core of many concepts in management science for over
50 years. In fact, the logistic shape may be viewed as the quintessence of pattern recognition
in many social sciences. It results from the tension (and shifting dominance over time)
between two forces: a potential for growth and a saturation effect. When it comes to strategic
management, three phenomena are typically represented, discussed, and even modeled,
sometimes forcingly, through a logistic framework: the diffusion of innovations,
technological trajectories, and technological substitutions are all synoptically represented, as
shown in figure 1, by S-shape curves. Respectively, these are graphical representations over
time of the cumulative number of adopters of the innovation reaching market saturation, the
improvements in the performance of a technology reaching an upper limit, and the
2
substitution of a new technology for a former dominant technology. The S-curve’s ubiquity in
the literature may actually be misleading as these three processes tend to be undifferentiated,
and their interrelationships skimmed.
Phenomenon Underlying dynamics Graphical S-curve of the:
Diffusion
An innovation is
adopted through a
social system
Cumulative adopters
(reaching saturation)
Technology
Improvement in the
performance of a
technology
Performance
trajectory
(reaching upper limit)
Substitution Substitution of one for
the other
Relative market share
(reaching dominance)
Figure 1: The classical S-curves: diffusion, technological trajectories, and substitution
Christensen (1992; 1992) explored the limits of the technology S-curve, i.e. the
performance trajectory, and found it to be a firm specific rather than uniform industry
phenomenon. Similarly, I here set to explore the interrelationships between these three
phenomena and the limits of the substitution S-curve.
While Pistorius and Utterback have discussed other modes of interaction such as
predator-prey or symbiosis (Pistorius and Utterback, 1997), the focus of this paper is on the
substitution dynamics between two or more technologies which interacts on a purely
competitive mode. Along with the classical S-shape base case and other relatively well
understood patterns, I have also identified non-trivial and surprising patterns: the classical
base case (including the concatenation and overlapping generations cases), the long term
feedbacks, the sailing ship effect, the intermediate hybrid, the path finder, and the “double
shift”. I describe each of these generic patterns, show their normalized fractional rate of
substitution as a function of time, and detail a historical example.
I briefly discuss the underlying dynamics of these substitution patterns and present a
broad theoretical framework obtained by aggregating many literature streams on
3
technological change. Finally, by using the concept of substitution trajectories, I identify
bifurcation points between these generic patterns and draw strategic implications.
2. Patterns of technological substitution
Many famous classification of technological innovation have already been developed.
These typologies attempt to reduce the complexity of the phenomenon to a few graspable
dimensions such as the type of innovation (product vs. process), the impact on organizational
competencies (enhancing vs. destroying), the link with market (established vs. new) or the
origin of the change (science based vs. supplier vs. clients, etc.). These typologies have been
fundamental for the management of innovation. However, their main focus is on the
industrial dynamics induced by technological change and especially on the survival of
incumbents versus new entrants. Their conclusions relate to the entry and exit rate, the
competitive advantage based on flexibility and know-how, and the effect of complementary
assets. While it is important for a firm to understand why and how its survival is threatened,
technological substitution is not a unified phenomenon. Thus, it is also important to know
how much time the firm may have before being possibly erased from the industrial landscape.
Tripsas highlighted that “understanding the origins and timing of discontinuous technological
change is extremely important for managers trying to better weather transitions” (Tripsas,
2005).
When it comes to technological change, the classical models of diffusion (Bass, 1969)
and substitution (Fisher and Pry, 1971) have been applied to a number of historical cases. The
normalized fractional rate as a function of time is the classical presentation of technological
substitutions. Despite its impressive statistical robustness, the smooth logistic shape of the
substitution S-curve must be challenged, in a Popperian sense. I thus provide
counterexamples, i.e. exceptions to the logistic generalization of technological substitutions. I
collected secondary historical data for a cases of technological change, many of which were
discussed in the literature. I show that the time-path of these substitutions did not follow the
classical S-curve.
2.1 Base Case Description:
The “base case” is a binary substitution that occurs when an emerging technology
N+1 substitutes for the current technology N which has reached maturity. This is where the
S-curve is at its best. The classical Fisher-Pry model states that the rate of substitution of the
4
new technology for the current one is proportional to the remaining amount of the old left to
be substituted (Fisher and Pry, 1971). The log of the ratio of the market share of the
succeeding technology to that of the first is a linear function of time. Fisher and Pry studied
the substitution rate for seventeen cases of technological change. They normalized the time
scale by use of the term 2(t-t0)/∆t, where ∆t is the time from 10% to 90% takeover and t0 is
the time of 50% takeover. This collapses all seventeen cases of substitution into the single
curve presented by figure 2.
Figure 2: Normalized substitution pattern of 17 cases (Fisher and Pry, 1971)
Generic pattern:
The generic pattern of a base case is presented by figure 3.
m1 m2 Figure 3: Generic pattern of base case substitutions
Historical example:
The transition from the Bessemer process to open-hearth in the steel making industry
is one of the earliest examples of binary substitution which the classical model has been
applied to (Fisher and Pry, 1971; Blackman, 1974). I collected historical data from the
American Iron and Steel Institute annual reports1 to present this classical example. The first
phase of technological change covers from 1880 to 1930. At the end of the 19th century, the
dominant method of steel-making was the Bessemer process, invented by Sir Henry
Bessemer in the late 1850’s. The rapidly expanding railroad industry provided a stimulus for 1 Sources : The American Iron and Steel Institute ; Annual Statistical Reports : (AISI, 1912), (AISI, 1965),
(AISI, 1979), (AISI, 1985), (AISI, 1993) and (Hendriksen, 1978).
5
heavy demand and the Bessemer converter was the foundation of the industry (Gold, Peirce
et al., 1984). Yet, the process had technical difficulties in part because the reactions involved
in a Bessemer blow were short and very violent. The open-hearth process, first proposed by
C.W. Siemens in 1861, overcame many of these difficulties and began substituting for the
Bessemer equipments.
The open-hearth uses the heat in the waste gases from the furnace itself to preheat air
and gas fuels and thus build up temperature. This enables the process to input scrap and other
cold metal in addition to the hot metal. By 1930 in the United States, the Bessemer process
accounted for only 12 percents of total output and was completely overshadowed by the
open-hearth process. The historical substitution pattern of this binary substitution in the U.S.
industry is shown on figure 4.
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
80.0%
90.0%
100.0%
1878
1888
1898
1908
1918
1928
1938
1948
1958
Bessemer Open Heart Figure 4: A base case substitution – Open-hearth for Bessemer (1878 – 1958)
2.2 Concatenation Description:
The base case relates to a binary technological substitution. But in an industry
successive generations of technologies replace each other over time. When considering a
sequence of technologies, the recurrence of the generic substitution pattern (emergence-
growth-dominance) is expected to look like a concatenation of “base cases”.
Generic pattern:
The generic pattern of a concatenation of base cases is presented by figure 5.
Since the early 1950’s the success of an aircraft was viewed as being heavily
dependent on the specifications of power output for its engines “independently of what was
precisely needed to fit the commercial and traffic requirements of the airline customers”
(Davies, 1964). Airline operations had steadily advanced towards commercial viability,
especially thanks the ‘incomparable’ DC-3 which probably introduced the dominant design
of modern aircrafts. In 1953, the de Havilland Comet 1, the first turbojet, started service. It
set the stage for a reappraisal of values in the industry. Despite being a dramatic
technological progress, several factors delayed the substitution of jet engines. They were
much louder and at landing required breaking distance much longer than propeller did by
inverting the angle of their blades. On the other hand, jet engine could not yet change the
direction of their air flow and the landing distances were still very important. Moreover, a
crash of a Comet 1 in April 1954 created a major crisis in the industry and turbo-jet services
were suspended.
The first turbo-prop, the Vickers Viscount, was introduced the same year in 1954 and
piston-propellers started being pushed out of service. Later versions of the Viscount with
longer fuselage were developed and larger turbo-props like the Bristol Britannia were
introduced and operated quite profitably until… in October 1958, the jet services were flown
again on the Boeing 707, the first ‘big jet’ airliner. From 1959, jet airplanes started serving
the important longer routes, whilst the turbo-props were allocated to many of the routes of
secondary importance.
Figure 167 illustrates these three generations of aircraft technologies. Figure 17 shows
the evolution of cruising speed8 (Davies, 1964) and the substitution patterns for these three
technologies (Linstone and Sahal, 1976).
7 From photos 30, 62 and 69 of (Davies, 1964)
15
The ‘Incomparable’ DC-3 Viscount, the First Turbo-Prop Boeing 707, the First ‘Big Jet’The ‘Incomparable’ DC-3 Viscount, the First Turbo-Prop Boeing 707, the First ‘Big Jet’ Figure 16: Three generations of aircrafts – Piston DC-3 / Turboprop Viscount / Turbojet 707
m1 m2 m2* m3 Figure 23: Generic pattern of a double shift substitution
Historical example:
To substantiate this generic pattern of a double paradigmatic shift, I combine the
longitudinal study of the typesetter industry conducted by Mary Tripsas (Tripsas, 1996; 1997;
2005) with other references on the chronology, evolution of techniques and economical
aspects of this industry (Swann, 1969; Hutt, 1973; Solomon, 1986; Wallis, 1988).
“Typesetting is the process of arranging and outputting text and images. Text from a
manuscript is entered into a typesetter machine […] the output of the typesetter, either paper
or film is then used to create a printing plate that is used by a press for high-volume printing”
(Tripsas, 1997 p. 124). Typesetters’ customers include newspapers, commercial printers and
some corporate ‘in-house’ publishers.
Typesetting started manually back with Gutenberg’s invention of the movable type
around 1440. At first, each individual letter was cast into a body of type using a mixture of
lead, tin and antinomy (Solomon, 1986). All the foundry types were stored in large case
drawers and the letters were then composed by hand to form lines of types. The first
commercial typesetting system that automatically distributed letter types for reuse was
22
introduced in 1886 with the Mergenthaler Linotype. An operator typed out individual letters
on a keyboard. With each keystroke, a lever released an individual matrix (mold). After a line
of type was composed and justified, the matrices were moved and the machine tapped a
reservoir of molten lead to cast a slug from the matrix. This formed a “line of type” with
raised letters. All the separate bars were assembled by a compositor to form the complete
printing plate for the press. Each matrix had an individual code key was distributed back into
its proper channel in the magazine. Because of the use of molten lead this generation of
typesetter is referred to as ‘hot metal’. The speed of a typesetting technology can be measured
in characters per second (cps). Until 1930, the speed of hot metal followed a very clear S-
shape trajectory from around 1.5 cps to a limit of 3.5 cps already reached by 1910. When
1946, the first successful analogue phototypesetter was introduced, this induced a very
noticeable sailing ship effect in the performance of the hot metal technology. By 1965 it had
reached a new limit of 8 cps (Tripsas, 2005, p. 35), thus effectively doubling the old
performance limit!
In analogue phototypesetters, the metal matrices were replaced with a photographic
image of the character. Placed in front of a xenon light source, the image of each letter was
flashed and projected onto a step-moving photographic film to form the line. The film was
then developed and projected onto a metal plate chemically treated with light-sensitive
emulsion to create a printing plate for high-volume press. The characters width, size and
position were adjusted optically through a system of lenses. Among others, the introduction
of phototypesetting considerably reduced the composing time and the safety issues associated
with molten lead. By 1975, analogue phototypesetters had reached speeds of 80 cps.
In 1965, the first cathode ray tube (CRT) typesetter was announced. CRT systems
digitalized the previously analogue images of the types. Thus, the characters could be stored
magnetically and instead of a xenon flash, a CRT display was used to write the characters
onto the photographic film. The CRT generation eliminated most of the typesetters’ moving
parts as electronics substituted for electro-mechanical technology (Tripsas, 2005). Speed
from 500 to 2000 cps were commonly available, with particular models reaching more than
3000 cps. However, Tripsas notes that this technology had exceeded the speed requirements
of most users. It was only interesting to print large telephone directories. The real take off
occurred in 1977 with the introduction of Intel 8080 microprocessor that enabled greater
connectivity with large electronic database and better control of the typesetting unit (Wallis,
1988).
23
The third technological shift occurred with the laser technology. In 1976, Monotype
International revealed the Lasercomp. The laser technology writes out text in a raster fashion
by a spinning polygonal mirror across the breadth of a page at thousands of sweeps per inch.
This raster stroke approach was a significant development for the imaging of pages complete
with text and graphics. However, it requires a page description language. The first language,
InterPress, was developed by John Warnock while at Xerox PARC, but Xerox did not
commercialize it. John Warnock and Charles Geschke left Xerox and in 1982 they formed
Adobe Systems. They then developed a simpler and high-level raster image processing
software called PostScript which went on the market in 1984. PostScript specifies the curves
that define the outline of a typeface in terms of straight lines and Bézier curves. By filling the
outline it allows the typefaces to retain smooth contours when rotated or scaled to any size.
PostScript offered flexibility, high-quality, and on-the-fly rasterizing.
The inclusion of the PostScript language in 1985 in the Apple LaserWrite effectively
sparked the desktop publishing revolution! It induced tremendous externalities and sudden
improvement of utility for the laser technology which became the best option for the novel
user needs of setting text and graphics in an integrated manner. From this point, laser
imagesetters started dominating the market.
Since the early 2000’s, yet another technology, computer-to-plate (CTP), has
revolutionized the printing industry because instead of striking a film (which must be
developed and then projected on a plate), the laser beam is used directly on a special printing
plate covered with light-sensitive emulsion (McCourt, 2002; Candille and François, 2004).
Figure 24 illustrates these successive typesetting technologies from 1886 to 2006.
FilmFilm
Pre 1886Hand-set type cases
1886 Hot-Metal Linotype
1946Analog Phototypesetter
1965Cathode Ray Tube
1976Laser Imagesetter
1984PostScript Outline Font
2000’sComputer-To-Plate
FilmFilm
Pre 1886Hand-set type cases
1886 Hot-Metal Linotype
1946Analog Phototypesetter
1965Cathode Ray Tube
1976Laser Imagesetter
1984PostScript Outline Font
2000’sComputer-To-Plate
Figure 24: Successive generations of typesetting technologies (1886-2006)
24
Since the introduction in 1977 of the Intel 8080 microprocessors, the CRT technology
had really took off and by 1985, CRT had reached more than 65% market share. Incumbent
firms were probably confident that their technological choice was strong and that they did not
have anything to fear yet from the 15% share of the emerging laser technology. But the
introduction of PostScript resulted in an explosive substitution and by 1988, laser
imagesetters had themselves reached 65% of market share. In an industry which had so far
experienced long technology cycles, such a double shift in less than three years was
shattering.
Figure 25 gives a longitudinal view of the technological substitutions in the U.S.
typesetter industry (Tripsas, 1997). We can clearly see the double shift whereby the
substitution of the CRT technology for the analog phototypesetters is cut short by the
emergence and rapid diffusion of the Laser technology enabled by PostScript.
Finally, figure 26 offers a synoptic view of all these generic patterns of technological
substitution. It demonstrates that substitution is not a unified phenomenon in the shape of a
smooth S-curve; rather there are various patterns induced by complex underlying dynamics.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
Hot Metal Analog Photosetter Digital CRT Laser Imagesetter Computer To Plate Figure 25: Double shift: Typesetters Hot Metal – Analog Photo – CRT – Laser – CTP (1949-2006)
3. Underlying dynamics
As shown by an immense body of literature, many technological substitutions occur
on a basic binary mode. Nevertheless, the above examples illustrate that substitution is
neither a unified logistic phenomenon nor a passive process. As Christensen puts it, many
authors “simply report observations of S-curve phenomena”, but “a few examine the
processes […] in considerable depth” (Christensen, 1992). While being a very well plough
academic ground, technological change has, according to Sahal, “turned out to be one of the
25
most vexing of all problems in the social sciences […] in particular, there remain all too
many missing links in our knowledge of the subject” (Sahal, 1981). More than a quarter of a
century later, I believe her comment still holds.
These generic patterns of substitution result from broad and complex underlying
dynamics. The technological burst and path finder patterns include a combination of long
term systemic interactions and social dynamics that greatly influence the creation of a mass
market. The long term feedback illustrate how broad system changes can trigger a
substitution. The defensive surge of the threatened technology, as with the sailing ship, and
the intermediate hybrid technology can both induce a delay in the substitution trajectory.
As shown in figure 1, the innovation and technology management literature
classically represents technology trajectories with a new technology taking over when the
existing technology has reached its technological limits (Linstone and Sahal, 1976; Sahal,
1981; Christensen, 2003; Durand, Granstrand et al., 2004 p. 108). The double shift, as an
extreme case of overlapping, challenges this view of the disruption timing. The technology
burst also illustrates the difficulty of strategic planning for long-term high technology
projects that are embedded in highly dynamic contexts. Indeed, when Concorde was finally
launched in 1976, it entered an aviation market that had changed drastically since the initial
decision back in 1956. Similarly, by the time the Iridium satellite system was launched in
1998, GSM had really changed the dynamics and growth of the mobile telecommunication
industry since the initial decision in the late 1980’s.
First or second-order technological externalities, i.e. changes induced by links with
other technologies, greatly influence the substitution trajectory. Externalities have been
discussed in the literature to occur in two forms. On one hand, network externalities increase
the expected utility as the number of adopters increase. The underlying dynamics are
economics. On the other hand, bandwagon effects result from strong social dynamics which
generate a boom and burst behaviour. I argue that there is a third type of externalities, which I
call technological externalities. By creating links between industries or practices, some
innovations act as catalysts, and sometimes even triggers, to explosive technological change.
26
Figure 26: Generic Patterns of Technological Substitutions
27
The case of the sailing ship shows that the substitution of steam boats for sailing ships
resumed thanks to improvements in steelmaking brought by open-hearth furnaces in the late
1870’s. Their diffusion of allowed the production of better steel, which in turn enabled boiler
plates and boiler tubes to withstand higher pressures; through a second-order feedback more
efficient steam boats could then be operated profitably (see figure 4 and figure 13). The
introduction of Intel 8080 microprocessors into the design of the digital CRT typesetter offers
another example. It enabled greater connectivity with large electronic databases and greater
control of the typesetting unit (Wallis, 1988); hence creating a step discontinuity in the utility
of this generation of technology. Finally, the PostScript is certainly a radical example of such
catalyst innovations. It created externalities with the growing installed base of desktop
computers which led to the desktop publishing revolution and a double shift in the typesetter
industry. Macromedia Flash and the USB port can also be thoughts of as catalysts
innovations that led to explosive change in the multimedia and consumer electronics.
These generic patterns show that we need to broaden the scope of our analysis in
order to better understand the underlying dynamics of technological substitution. A system
approach to technological change should account for classical industrial dynamics
(Utterback, 1994), but also regulatory changes, spillovers from science and academia
(Henderson and Cockburn, 1996; Murmann, 2003), the availability of financing and
technological development and externalities. A broader model should also recognize the
critical role of social factors (Dattee and Weil, 2005). Without detailing its structure, figure
27 shows an aggregated theoretical framework (Dattee, 2006) which offers a synoptic view of
the major concepts of technological change and the research traditions that have discussed
them.
SocialDynamics Market Diffusion
TechnologicalEvolution
IndustrialDynamics
Science &Acdemia
Heterogeneity
Socio-Political Co-Evolution
Socio-Technical Co-Evolution
Lobbying
Opportunity
Driver of Growth
Knowledge Spillover
Perceived Risks & Opportunities
Technology Development
Offering
Ethical Issues
Regulation &Policies
FinancialSector
Taxes and Innovation Programs
Political Environment
Investment
Technological Paradigm
Research Programs
Discursive Actions
Figure 27: A broad theoretical framework of technological change.
28
4. Bifurcation analysis
The substitution time-paths, patterns, or trajectories are influenced by the dynamics
taking place in the broad technical system described in figure 27. In this section, I identify
bifurcation points between these trajectories and how the generic patterns can be combined
into sequence to replicate the longitudinal view of technological substitution in an industry.
Based on the life cycle theory, an emerging technology generation must go through a
growth phase before reaching dominance. Figure 28 shows the three phases over time of the
classical logistic pattern of technological substitution. Every technological change starts with
a spark that ignites the substitution dynamics. Then, for a base case, the new technology
smoothly enters a growth phase which Moore refers to as “crossing the chasm” (Moore,
2002) before reaching market dominance.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Spark
Growth
Dominance
Figure 28: Three time phases of a base case substitution
Using a simulation model developed with the system dynamics methodology (Dattee,
2006), it is then possible to plot substitution trajectories under various scenarios. A base case
scenario can be altered by changing the dynamics at the system level, as described in figure
27 – e.g. changing the timing of emergence of the new technology, or accounting for specific
social dynamics, etc. This is illustrated by figure 29 which shows that there exist important
bifurcation12 points in the substitution trajectory13 of a technology N+1. Indeed, while the
12 Pasquet in his study of technological transition defines points of technological bifurcation by analogy with
the physicochemical theory of dissipative structure (Pasquet, 2002). Around bifurcation points, macroscopic
qualitative changes in the structure can be produced by the amplification of an infinitesimal internal fluctuation
or by a small external perturbation, while the system is in an instable state. Nevertheless, Pasquet refers to
bifurcations between two technological paradigms, i.e. moving from N to N+1. In my work, the bifurcations
points are between types of substitution trajectories already started (after the paradigmatic bifurcation point in
Pasquet’s meaning). 13 For clarity, the complementary fractions for technology N were omitted; i.e. fn+fn+1=1
29
substitution is taking place along a given trajectory, changes at the system level may create a
bifurcation towards another substitution trajectory.
René Thom developed the catastrophe theory in order to understand sudden
phenomena. In a system, these abrupt changes occur at points of tension between two
variables. At a particular moment, there is a conflict between two attractors and the system is
constrained to suddenly decide for one of them. The catastrophe theory emphasizes
phenomenological discontinuities but also relate them to an underlying slow evolution
(Thom, 1984).
Figure 29: Bifurcation graph of technological substitution trajectories
The initial spark is common to every cases of technological change because it is the
initial disruption that ignites the substitution dynamics. However, a first bifurcation point is
evident after this initial takeoff. In the classical S-curve view, the substitution continues on
the left of this point as it is assumed that the technology smoothly enters the growth phase.
The system is on a base case trajectory (1). If this substitution reach completion, the next
spark (N+2) will generate a concatenated pattern. Nevertheless, in many cases, the next spark
will create an overlapping pattern (2). These are the classical views of technological change
between successive generations of technology. However, as I have discussed earlier when the
generation N+1 is on its way to complete substitution, there is another potential bifurcation
point because the system could suddenly bifurcate towards a double shift (3). The catastrophe
theory states that at bifurcation points there is a tension between two attractors, a slower
underlying dynamics and a quicker one (Thom, 1984). Figure 29 shows that a double shift
can be considered as a particular case of overlapping, but the catastrophe theory also
30
highlights that the sudden bifurcation that can be triggered by a specific N+2 spark (e.g.
PostScript).
These trajectories (1,2, and 3) are from an initial bifurcation towards mass market.
But often the proponents of the previous technology react and respond either with a defensive
surge or a hybrid intermediate. In both case the resulting pattern for N+1 is a delayed
substitution; the substitution bifurcate towards the right. As in the case of steam boats or CRT
typesetters, technological externalities can create a new point of bifurcation whereby the
substitution dynamics eventually resume. The technology N+1 is back on track and enter the
growth phase (4). The rest of the substitution trajectory will be determined by the emergence
of N+2 (i.e. concatenation, overlapping, etc.). As an example, figure 29 actually indicates an
overlapping case occurring after the system had followed a sailing ship pattern (5).
At the initial bifurcation point, generation N+1 can actually become stuck in the burst
scenario. This can happen because of the defensive surge of technology N was sufficient –
but this seems to be a rare case – or broader dynamics (cf. Concord or Iridium). The new
generation N+1 only appeals to a small elite. From this point, the perspective of entering the
phase of rapid growth (i.e. crossing the chasm) is greatly compromised and the system will
most probably follow the very strong attractor of a burst pattern (6).
However, the path finder pattern shows us that in some cases a “last chance”
bifurcation is possible because the broader system change and the growth phase is finally
reached (7). Nevertheless, either creating this point through institutional entrepreneurship or
guessing the right timing to enter will be extremely difficult. It will demand a deep
understanding of the emergence of bifurcation point. Munir and Philips show how Kodak
fought for many decades using discursive strategies to make its roll-film – an initial burst –
bifurcate towards a mass-market success (Munir and Phillips, 2005). However, figure 29 has
us wondering how long can a “sleeping beauty” technology wait before it becomes a
mummy?
This bifurcation analysis shows that strategic actions may be undertaken by change
agents to influence the dynamics of substitution, increase the strength of an attractor and thus
favor the occurrence of a preferred pattern. As an example, if a company is stuck in a burst it
probably has four alternatives:
1. First, wait for the right system conditions to happen,
2. Second, undertake strategic actions to influence the discursive dynamics and change
the evaluation criteria of adopters in order to create those right conditions,
3. Third, create an alternative use for the technology,
31
4. Fourth, withdraw from the business and admit failure.
The institutional entrepreneurship of Kodak offers great lessons in changing the dynamics of
a burst and make the system bifurcate towards a path finder trajectory. As many authors
emphasize, the dynamics of substitution can be socially constructed through discursive
actions that influence the decision criteria and market preferences (Van de Ven and Das,
2000; Maguire, 2003; Schilling, 2003; Munir and Phillips, 2005).
Sometimes the entire technical system has so much inertia that it’s just too big to
influence its trajectory. Nevertheless, Yoffie and Cusumano (1999) explain that like in judo
whose strategy is based on rapid movement, flexibility, and leverage, there are strategic
actions that one can take to turn these larger dynamics to one’s advantage. Indeed, the
essence of strategy is timing. Hence, identifying the emergence of a double shift might for
example offer the opportunity to leapfrog the sandwiched generation without wasting time.
This would also allow profiting from the momentum of change already initiated. By
definition, a double shift occurs during the growth phase of the technology N+1 when major
investments have just been made to increase volume, etc. Therefore, these commitments and
limited financial capacity will make it extremely difficult for those engaged in the
sandwiched generation N+1 to follow and switch to N+2.
Finally, this approach shows that the generic patterns of substitution that I have
described can actually be combined to replicate more precisely the longitudinal view of
technological change in an industry. For example, instead of a concatenation of base case S-
shape substitution the typesetter industry, as discussed in section 2, went through a sailing
ship from hot metal which delayed analogue, the CRT were stuck in a niche market until the
introduction of the Intel 8080, but when the substitution resumed like for a path finder it was
suddenly curt short by a double shift from a combination of laser and PostScript!
5. Conclusion
In this paper I started by challenging, in a Popperian sense, the smooth logistic shape
of the substitution S-curve. I provided counterexamples, i.e. exceptions to the logistic
generalization of technological substitutions by collecting secondary historical data for a
series of examples used in the literature on technological change. I showed that the time-path
of these substitutions did not follow the classical uniform S-curve but that rather more
complex substitution trajectories. These were summarized in figure 26. This variety of
patterns requires us to broaden the scope of our analyses and account for the dynamics
32
occurring at the system level; I proposed an aggregated theoretical framework of
technological change. Using the catastrophe theory, I then conducted a bifurcation analysis.
This resulted in figure 29 which presents the bifurcation points between the generic patterns
of substitution.
Contrary to the classical view of a concatenation of smooth logistic base cases where
each successive generation reaches dominance, these generic patterns of substitution can
actually be combined to replicate more precisely the longitudinal view of technological
change in an industry. The combination of these analyses shows that a better understanding
of the underlying dynamics of substitution could help identify the conditions of emergence of
particular patterns. Hence, a company could for example undertake strategic actions to
influence the bifurcation towards preferred patterns (e.g. engage in institutional
entrepreneurship to change a technological burst into a path finder), or try to identify a double
shift and to leapfrog the crushed generation.
33
References
AISI (1912). Annual statistical report, American iron and steel institute. AISI (1965). Annual statistical report, American iron and steel institute. AISI (1979). Annual statistical report, American iron and steel institute. AISI (1985). Annual statistical report, American iron and steel institute. AISI (1993). Annual statistical report, American iron and steel institute. Bass, F. M. (1969). "A New Product Growth Model for Consumer Durables." Management
Science 15: 215-227. Bijker, W. and J. Law, Eds. (1994). Shaping Technology / Building Society: Studies in
Sociotechnical Change (Inside Technology). Cambridge, MA, The MIT Press. Blackman, A. W. (1974). "Market dynamics of substitutions." Technological Forecasting and
Social Change: 41-63. Candille, A. and D. François (2004). Le classement annuel de l'impression. Caractere. 604:
52-110. Carter, S. B., S. S. Gartner, et al. (2004). Historical Statistics of the United States: Colonial
Times to 1970. U. S. B. o. t. Census. Christensen, C. M. (1992). "Exploring the Limits of the Technology S-Curve, Part I:
Component Technologies." Production and operations management 1(4): 334-357. Christensen, C. M. (1992). "Exploring the Limits of the Technology S-Curve, Part II:
Component Technologies." Production and operations management 1(4): 358-366. Christensen, C. M. (2003). The Innovator's Dilemma. New York, HarperCollins. Dattee, B. (2006). The dynamics of technological substitutions and successful innovations.
Paris / Dublin, Ecole Centrale Paris / University College Dublin. PhD: 315. Dattee, B. and H. B. Weil (2005). "Dynamics of social factors in technological substitutions."
Massachusetts Institute of Technology, Sloan School of Management(Working Paper 4599-05).
Davies, R. E. G. (1964). A history of the world’s airlines. London, New York, Oxford University Press.
Durand, T., O. Granstrand, et al., Eds. (2004). Bringing technology and innovation into the boardroom: strategy, innovation and competences for business value. Basingstoke, Palgrave MacMillan.
Durand, T. and B. Stymne (1991). "Technology and Strategy in a High-Tech Industry: Reflections on the past and future of two European telecom companies." Corporate and Industry Strategies for Europe: 193-213.
Finkelstein, S. and S. Sanford (2000). "Learning from corporate mistakes: the rise and fall of Iridium." Organizational Dynamics 29(2 ): 138-148.
Fisher, J. C. and R. H. Pry (1971). "A simple substitution model of technological change." Technological Forecasting and Social Change 3: 75-88.
Foster, R. (1986). Innovation: the attacker's advantage. New York, Summit Books. Foucault, M. (1966). Les mots et les choses: Une archéologie des sciences humaines,
Gallimard. Gar, N. (2005). "Flying Blind." Strategy & Business 41(Winter): 26-29. Gold, B., W. S. Peirce, et al. (1984). Technological change in primary steel making.
Technological Progress and Industrial Leadership: The growth of the U.S. steel industry, 1900-1970. Lexington, MA, LexingtonBooks: 529-558.
Graham, G. S. (1956). "The ascendancy of the sailing ship 1850-1885." The Economic History Review 9(1): 74-88.
Henderson, R. M. (1995). "Of life cycles real and imaginary: The unexpectedly long old age of optical lithography." Research Policy 24(631-643).
34
Henderson, R. M. and I. Cockburn (1996). "Scale, Scope, and Spillovers: the determinants of research productivity in drug discovery." RAND Journal of Economics 27(1): 32-59.
Hendriksen, E. S. (1978). Capital expenditures in the steel industry 1900 to 1953. New york, Arno Press.
Hutt, A. (1973). The changing newspaper; typographic trends in Britain and America 1622-1972. London, Gordon Fraser.
ITU (2002). Yearbook of statistics: Telecommunication services 1993-2002, International Telecommunication Union.
Lin, B.-H., M. Padgitt, et al. (1995). Pesticide and fertilizer use and trends in U.S. agriculture. E. R. Service, U.S. Department of Agriculture.
Linstone, H. A. and D. Sahal, Eds. (1976). Technological substitution : forecasting techniques and applications. New York, Elsevier.
Maguire, S. (2003). "The Co-Evolution of Technology and Discourse: A study of substitution processes for insecticides DDT." Organization Studies 25(1): 113-134.
Mahajan, V. and E. Muller (1996). "Timing, Diffusion, and Substitution of Successive Generations of Technological Innovations: The IBM Mainframe Case." Technological Forecasting and Social Change 51(2): 109-132.
McCourt, A. (2002). Bringing CTP out of the dark ages. Graphic Repro & Print. 18. Moore, G. A. (2002). Crossing the chasm: marketing and selling high-tech products to
mainstream customers. New York, HarperCollins. Munir, K. A. and N. Phillips (2005). "The birth of the 'Kodak moment': institutional
entrepreneurship and the adoption of new technologies." Organization Science 26(11): 1665-1687.
Murmann, J. P. (2003). Knowledge and Competitive Advantage: The Co-Evolution of Firms, Technology, and National Institutions, Cambridge University Press.
Norton, J. A. and F. M. Bass (1987). "A Diffusion Theory Model of Adoption and Substitution for Successive Generations of High-Technology Products." Management Science 33(9): 1069-1086.
Osteen, C. D. and P. I. Szmedra. (1989). Agricultural pesticide use trends and policy issues. E. R. Service, U.S. Dept. of Agriculture.
Pasquet, N. (2002). Pour une compréhension complexe des processus de bifurcation technologique : le temps-devenir - Le cas de l'énergie solaire photovoltaique. Laboratoire Strategie & Technologie. Paris, Ecole Centrale Paris. PhD.
Phister, M. (1979). Data processing technology and economics. Bedford, MA, Digital Press / Santa Monica Publishing Company.
Pistorius, C. W. I. and J. M. Utterback (1997). "Multi-Mode Interaction Among Technologies." Research Policy 26: 67-84.
Rosenberg, N. (1976). "On technological expectations." The Economic Journal 86(34): 523-535.
Sahal, D. (1981). Patterns of Technological Innovation. Reading, MA, Addison-Wesley Publishing Company.
Schilling, M. A. (2003). "Technological Leapfrogging :Lessons from the U.S. Video Game Console Industry." California Management Review 45(3): 6-32.
Smith, C. (1992). "Understanding technological substitution: Generic types, substitution dynamics, and influence strategies." Journal of Business Research: 279-302.
Snow, D. (2003). Extraordinary Efficiency Growth in Threatened Technologies: Explaining the Carburetor’s “Last Gasp” in the 1980s, Haas School of Business, UC Berkeley.
Sohn, S. Y. and B. J. Ahn (2003). "Multigeneration diffusion model for economic assessment of new technology." Technological Forecasting and Social Change 70: 251–264.
35
Solomon, M. (1986). The art of typography : an introduction to typo-icon-ography. New York, Watson-Guptill Publications.
Sull, D. N. (1999). "The dynamics of standing still: Firestone Tire and Rubber and the radial revolution." Business History Review 73(Autumn): 460-464.
Swann, C. (1969). Techniques of typography. London, Lund Humphries. Taylor, J. (1998) "LaserDisc FAQ." Widescreen Review Laser Magic Special Edition,
http://www.access-one.com/rjn/laser/ldfaq_30.pdf Thom, R. (1984). Stabilité structurelle et morphogénèse. Paris, Dunod Tripsas, M. (1996). Surviving radical technological change: an empirical study of the
typesetter industry. Sloan School of Management. Cambridge, MA, Massachusetts Institute of Technology. PhD.
Tripsas, M. (1997). "Unravelling the Process of Creative Destruction: Complementary Assets and Incumbent Survival in the Typesetter Industry." Strategic Management Journal 18(Summer Special Issue): 119-142.
Tripsas, M. (2005). Customer Preference Discontinuities: A Trigger for Radical Technological Change, Harvard Business School.
Tushman, M. and P. Anderson (1986). "Technological Discontinuities and Organizational Environments." Administrative Science Quarterly 31(3): 439-465.
Utterback, J. M. (1994). Mastering the Dynamics of Innovation. Boston, MA, Harvard Business School Press.
Van de Ven, A. H. and S. S. Das (2000). "Competing with New Product Technologies: a process model of strategy." Management Science 46(10): 1300-1316.
Wallis, L. W. (1988). A concise chronology of typesetting developments, 1886-1986. London, Wynkyn de Worde Society in association with Lund Humphries.
Yoffie, D. B. and M. A. Cusumano (1999). "Judo Strategy: The Competitive Dynamics of Internet Time." Harvard Business Review January-February: 71-81.