A Dynamic Modei of Technology Diffusion
James M. Lyneis
Vice President, Pugh-Roberts Associates
Abstract
The diffusion of new technologies into the market is a critical factor in the ‘success of any technology-
based company. This paper describes a system dynamics model which integrates a number of key concepts
presently used to understand the diffusion process (e.g., technical progress functions, cost-experience curves). It
shows how these concepts, together with management decisions regarding R&D investment, marketing, and
Pricing, drive the evolution of diffusion between technologies. It then illustrates how simulation can be used to
understand the critical success factors in technology diffusion, and what this means for the management of
technology-based companies.
Introduction
Successfully introducing a product based on a new technology can give a company a significant advantage
over competitors, both in terms of the opportunity to define the standard for that generation of technology, and in
terms of the ability to drive down costs ahead of the competition. However, being market leader also comes with
significant risks: introducing a technology before the market is "ready" will yield slow sales growth, and
potentially allow competitors to leap-frog your technology with the next-generation when customers are finally
teady for improved technical performance; introducing too early can also prematurely reduce the sales and profits
on the last generation of technology, and trap the industry in a cycle where the costs of the last generation are not
recovered before the next generation is introduced. Determining how much to invest in developing the next
generation of technology vs. improving the current generation, when to introduce a new generation, how much to
spend on marketing, and how to price the product are all critical management decisions affecting success or
failure.
A number of techniques and tools have been developed to improve technology forecasting and assessment.
For example: com
1. Technological Progress or S-Curves -- As Figure 1 Technological Progress or S-
Curves
illustrated in Figure 1, these curves describe
the evolution of technical performance as a
function of cumulative R&D effort; for any
given technology, relative technical
performance follows an S-shaped pattern,
starting low and growing slowing as initial
R&D effort takes time to bear fruit, then
increasing rapidly with further effort, until
finally diminishing returns begin to set in; at
some point, the industry moves to a new
_generation technology, the performance of
which also follows an S-Curve; note that the
initial performance of the new technology can
and is often less than the ultimate performance
of the prior technology. Technological
progress curves help managers understand and
forecast the maturity of a technology as a guide
to R&D investment decisions.
‘Technology Progress Index
‘Cumulative R&D Effort
268 SYSTEM DYNAMICS '93
2. Cost-Experience Curves -- As illustrated in Figure 2,
these curves describe the behavior of costs as a
function of cumulative production volume; they
reflect the product, process, and organizational
learning that drives down costs as experience
accumulates. Cost-experience curves influence a
company's pricing strategy. A company must
balance the market share impact of reducing prices ©
as costs fall against the need to improve profit
margins in order to recover investments.
3. Price-Performance Curves -- Price/performance
curves depict the tradeoff that exists at any point in
time between price and technical performance. As
illustrated in Figure 3, over time the tradeoff curve
will move down and to the right as a result of cost-
experience effects (lower price for same
performance) and of technical progress (improved
performance for same price). Price-performance
curves evolve from management decisions regarding
R&D investment and pricing, and from the shapes of
the Technical Progress and Cost-Experience Curves.
Price-performance is one driver of market
acceptance of new technologies,
i iff As
illustrated in Figure 4, diffusion curves describe how
sales of a new product or service evolve over time.
Diffusion curves are also S-shaped, as at first only.
lead users are willing to adopt a new technology, ...
then as price drops and performance improves, more.
and more users switch to the new technology.
Eventually, sales slow as laggards are the only
remaining non-users. Diffusion curves are used to
guide the introduction of new technologies.
Figure 2 Cost-Experience Curve
LOG FRICECOST- Const
LOG ACCUMULATED VOLUME.
Figure 3 Price-Performance Curves
Technology Progress Effect
Experience/L earning Curve Effect
» Performance
Figure 4 Diffusion or Product Life Cycle
Curve
i Late
Maley
H
$ ey
Mabey
il.
et
de |
Time ———_—_—_—>
5. Substitution Curves -- These curves graph the diffusion curves for succeeding generations of products or services.
When sales of a new product will start, how fast the product gains market acceptance, and what its ultimate level
‘of acceptance is (before the next generation makes inroads) depends on many factors, including the price
performance tradeoffs and how management markets the new technology.
SYSTEM. DYNAMICS '93 269
6. Fisher-Pry Technique -~- A means of estimating a substitution curve based on the assumption that the rate of
substitution is proportional to the fraction of the older one still in use. This technique is used to guide
investment decisions.
Management is interested in affecting the substitution curves to the company's best advantage through
decisions regarding: z
1. » How much to spend on developing the next generation technology vs. improving the performance of
the current generation;
2. When to introduce the next generation;
3. . How to price the product; and
4. How much to spend on marketing.
While the above concepts are all useful toward that end, they only describe possible behavior of pieces of the
system. How the pieces fit together to drive diffusion, and how alternative strategies can affect that process, has
been left to managers intuition and “mental models.”
Therefore, a system dynamics model was developed which integrates these common technology
forecasting and assessment concepts with key management decisions and with market characteristics and
responses. The model simulates the evolution of technologies (Diffusion and Substitution Curves), based on
assumed Technological Progress Curves and Cost-Experience Curves, and on management actions. The model
described herein was developed as a training device. We are in the process of using the model to develop heuristics
regarding the likely evolution of various technologies, and the best management actions depending on the
technology-market situation. The process for doing this is illustrated in the concluding section of this paper.
Description of the Model
‘The model represents the market for a given product or service (¢.g., television sets; computers), the
investment in technology which produces succeeding generations of products to serve this market, and
management actions regarding investment, pricing, and marketing. At this stage of development, the model
represents an industry as a whole rather than an individual company.
Figure 5 shows the major elements of the model (Figure 5 omitted because of space constraints; see Figure
6 instead). The key stocks and flows in the market are shown in the middle of the figure. For simplicity, there are
only three generations of technology: T1 represents the first/current generation; all users start with technology T1.
Over time, users can convert from T1 to T2 and ultimately T3, or they can bypass T2 completely and jump to T3.
“Users” is here defined as a fraction, so at any point in time the sum of Users of T1, Users of T2, and Users of T3
must equal 1.0. The critical area of the model determines the conversions between generations of technology. At
the top of Figure 5, users make purchases of units ("sales"). Sales by technology type depend on total demand and
users by technology type (which is a fraction). Total demand consists of a growth component and a replacement
component. These sales then enter a stock of Units in Use (by technology type), and are eventually retired. Sales
drive revenues to the industry.
Three key concepts, illustrated in Figure 6, are represented in the Technology Evolution and Diffusion
Model. The first concept, highlighted in Section A, is that users are ready to convert ("potential conversions”)
when they are ready to purchase a new unit. This occurs when a new user enters the market, or when a unit is
ready for retirement. Product lifetime depends on a normal lifetime, which can be reduced if user needs and the
performance of the new technology obsolete the old technology.
Actual conversions then depend on potential conversions and the second key concept,
switch” (see B of Figure 6). Users’ willingness to switch is first a function of relative price and relative
performance: the higher the relative performance of a new technology and/or the lower relative price, the more
willing are potential users to switch to the new technology. Price is described further below. Technical
270 SYSTEM DYNAMICS '93
performance depends on cumulative R&D spend on a given technology (refer back to Figure 5). The Technical
Performance Index for each technology as a function of Cumulative R&D is an input assumption in this model.
Willingness to witch based on price-performance is then modified by the effect of perceived risk: if the
risks are high, willingness to switch is less than that indicated by price and performance. Three factors affect ”
perceived risk: (1) users inherent risk aversion; (2) switching costs; and (3) lack of standards for the technology.
Inherent risk aversion reflects the fact that not many people are willing to try something new until they see lots of
other people using it. They are not sure it will work-as: promised; they do not want to be stuck with a product that
never gains broad acceptance; and so on. Hence, the more users of the new technology, the-lower the perceived
tisk. Marketing can also reduce inherent risk by rapidly spreading the word about existing users, and by
overcoming concerns that the technology does not work as promised, High switching costs also increase risk.
Switching costs here represent real dollar outlays, and implicit costs such as changes in procedures, required to use
the new technology. For example, switching from tapes to CD's requires purchase of a CD player; changing from
mainframe computers to PC's requires changes in procedures for control of software and access to information.
Again, as the new technology proliferates, switching costs will likely fall as supporting technologies and
procedures are developed and fall in price. Lack of standards occurs when several competing versions of the new
technology are developed and marketed (e.g., VHS and Beta). Many users are therefore unwilling to convert to the
“new technology because of the risk of picking the losing standard. As the technology matures and the number of
users grows, requirements for a standard force one to develop, and risk diminishes.
‘The third key concept in this model (see C of Figure 6) is that of “user need for improved performance."
User need depends on the users technical requirements relative to the technical performance offered by the old
technology. In this general model, technical performance and user requirements are measured by an index. In an
actual application, real concepts such as computer processing speed (MIPS), switching speed and reliability, and so
on, could be used to measure performance. When user requirements for technical performance exceed the
technical performance of the old technology, we have a "market pull” situation. In this situation, users become
more interested in relative performance than relative price, and are more willing to switch even in the face of high
prices. Further, users are more willing to prematurely discard their current products to get the improved
performance of the new technology. Hence, retirements and potential conversions increase. In a situation in
which user requirements are less than the performance of the old technology, we have a "technology push"
situation. Here relative price becomes more important, and the new technology must offer vastly superior
performance to the old in order to offset any price premium. Of course, especially for consumer products,
marketing can create a perceived increase in user requirements!
Finally, referring back to the highlighted variables in Figure 5, management actions tie the parts of the
system together. (Note also that each of the key concepts applies to conversions from T2 toT3, and 1 to T3 as
well.) Cumulative sales (as a proxy for production) drive unit costs through an assumed cost-experience curve.
Management must then set prices based on unit costs and profitability considerations (including earning a return”
on prior R&D investments). Pricing influences willingness to switch, and revenues and profits. The amount and
allocation of R&D spending depends on revenues, profits, and technical performance. A certain fraction of
industry revenues is spent on all R&D. The amount of this total spend allocated to Technology T1 depends on
trends in T1's technical performance and on cumulative profits on T1: as long as T1's technical performance
continues to improve, and if cumulative profits are below desired profitability, the industry will allocate most R&D
to T1; but as TI's technical performance improvement slows (no longer receiving benefits from continued R&D),
and as cumulative profits hit or surpass desired return, less and less will be spent on T1. As described previously,
marketing spend can reduce perceived risk, and increase perceived user requirements.
Base Model Behavior
The first step in defining a "Base Case" simulation is to specify the characteristics of the market and
technology being represented: :
272 SYSTEM DYNAMICS '93
Key Concepts in the Technology Evolution and Diffusion Models
Figure 6:
/ f
A: Potential Conversions
Willingness to Switch
‘User Need
C:
gue meee aa
i
i
-_
i EE
7, RB ON
Aa---- B-
v
f
i
N.
B
SYSTEM DYNAMICS '93
en
Rate of growth in demand
Average product lifetime
Initial user technical performance requirements and rate of growth in requirements
Normal risk sensitivity from "newness", switching costs, and lack of standards
eeee
rf istics --
© Technical performance index (for each technology, as a function of cumulative R&D)
‘© Cost-experience curve (for products from each technology, asa function of cumulative volume)
Relative production and other costs of products from succeeding generations of technology (before
the experience effect)
In this example, the market grows at 10% per year, with a 5-year product lifetime. User technical requirements are
set such that the performance of the technologies slightly exceeds need. Hence this is somewhere between pure
market pull and technology push situations. The market's risk sensitivities are set in the middle range between a
very high sensitivity and_a low sensitivity. Technical performance indicies are illustrated in the output below
(Figure 7). The required R&D spend to get this performance is assumed to be: for T2, twice that for T1; for T3,
three times that for T2. Costs are assumed to fall by 20% for every doubling of volume, with the normal cost of
products from each succeeding generation assumed to be the same.
As itlustrated i in Figure 7, at the start of the simulation in 1990 the technical performance of T1 is
approaching its maximum level. However, TI's performance is somewhat above user requirements until nearly |
2000, In 1996, as progress on T1 slows and because investment has been recovered, the industry begins investing
in T2, but it is not until the year 2002 that the performance of T2 surpasses that of T1 (and user requirements).
The industry starts investing in T3 in the year 2008, but it is not until about 2014 that T3's performance surpasses
that of T2. For the most part, the technical performance of the available technologies slightly exceeds user
performance.
Figure 7 Technological Progress Index
oss ST], —- 13
a 12 = User Requirements,
6.
A.
ei
2. =
ee reny Se Pes (es eps pee tee,
0.
1990. 1996. 2000. 2004. 2008. 2012. 2016. 2020. 2024, 2028.
TIME~
Figure 8 shows the diffusion of the three technologies over time (substitution curves). All of the users stay
with technology T1 until 2004. At that time, the technical performance of T2 surpasses that of T1, and users begin
to switch. The substitution occurs at an accelerating pace until around 2012, at which point many users have
switched from TI and the rate of gain for T2 progressively slows. By 2017, the fraction of users with T2 peaks and
begins to decline as users begin to switch to T3 (because the technical performance of T3 surpasses that of T2 at
this time). Substitution from T2 to T3 follows the expected S-curve, reaching completion in 2030.
SYSTEM DYNAMICS '93 273
Figure 8 Users by Technology Type (Fraction)
ul — a oan 99
ae JA > 4 r ae
yy }
75) 7 :
/ Nek
Ne
5
/ ’
25 L 2
Y iN ee
: &
0. Late ire
1990. 1996. 2000. 2004. 2008. 2012. 2016. 20:
TIME
Willingness to switch from T1 to T2, shown in Figure 9, increases from zero starting in 2004. The
increase is initially driven by price-performance. As shown in Figure 7, the performance of T2 surpasses that of
Tat this time, and while T2 is substantially more expensive than T1 (see Figure 10), user requirements also
exceed T's performance and some early risk-takers are convinced to switch technologies. Then, a number of
feedback loops accelerate’ the improvement in willingness to switch: (1) as T2 volumes increase, unit costs and
prices decrease, thereby improving price-performance; (2) continued investments in T2 R&D drive up T2 technical
performance, further improving the price-performance tradeoff; and (3) as more and more users switch
technologies, perceived risk falls and more and more users become willing to switch (see Figure 11; perceived risk
is the product of the three risk components). In the end, T2 risk increases as users begin to switch to T3.
Figure 9 Willingness to Switch to T2
—— Willingness to Switch
— —Willingness Based on Price-Performance
eae Effect of Perceived Risk
N\
.75) \
\
5]
.25
qT ~ 2000. 2004. 2008. 2012, 2016. 2020. 2024. 2028.
TIME
274 SYSTEM DYNAMICS '93
Figure 10 Price.
=—=—Price, T1 — —Price. T2 +77 Price, 13
78) \
~~~ a al -
.25
1990, 1996. 2000. 2004. 2008. 2012. 2016. 2020. 2024, 2028.
TIME
Figure 11 Risk of Switching to T2 (1=Low, 12=High)
Perceived Risk(0.,12.) -- - From Lack of Usage(0.,6.)
12.— — From Switching Costs(0..6.) — - From Lack of Standard@.,6.)
é —
9. _
45) /
6.
jee at ache lal rl are
15 =_
3 -
0. 1990. 1996, 2000. 2004. 2008. 2012. 2016. 2020." 2024. * 2028.
TIME —
Understanding the Critical Factors in Technology Diffusion
As at this stage the generic model is primarily a learning device. Its principal purpose is in determining
heuristics regarding the most critical factors affecting the successful diffusion of downstream technologies, and in
particular how management actions can improve performance. It seems likely that different factors may be more
or less important in different market situations. The matrix illustrated in Figure 12 captures a possible range of
environments, Three factors are assumed to vary: (1) user need (i.., technology push vs. market pull); (2) market
growth rate (the higher the growth rate, the more new users there are relative to replacement users, and therefore
the higher the willingness to switch to a new technology); and (3) product lifetime vs. technology lifetime ( the
shorter the technology lifetime relative to the product lifetime, the more critical the timing of technology
introduction relative to the replacement cycles becomes). As an example, we might imagine TV's being in the
upper left box (technology push, relatively low market growth rate, ad short product life relative to technology
life). In contrast, personal computers may be in the back right corner (market pull, high market growth, and long
product life relative to technology life).
SYSTEM DYNAMICS '93
Figure 12 Scenario Matrix
xa
<€
&
Tech Push VV A
User Computer
Need Base
Market B
Pull
Low High
Market
Growth
Rate
The Base Case assumptions would put the product in the middle of the front matrix. For the sake of
illustration, we have conducted a range of sensitivity tests on the two user need alternatives illustrated by A (Tech
Push) and B (Market Pull). In these tests, we have varied 5 factors by plus or minus 25% (3 technology
assumptions and 2 management policies):
ak all od 5 aad
Inherent risk level of users
Maximum technical performance of the technology
Effort, and therefore the time, required to bring the new technology to peak performance.
Marketing effort during technology introduction (i.e., before revenue-based spend is practical)
Initial product price (j.e., before enough investment has been recovered to switch to cost-based pricing).
The results of these experiments are summarized in Figure 13, which compares the differences in volume for
Technology T2 between the +25% case and the -25% case for each of the five factors. The higher the line, the
More sensitive volume is to changes in either the input assumption or to the management policy. These results
indicate that:
L
Technology diffusion is almost always less sensitive to technical uncertainties, risks, and management
actions in a market pull environment than in a technology push environment.
2. Marketing and pricing are much more critical factors in successful diffusion i ina technology push
environment than in a market pull environment.
3. Technical performance is the’ most important factor in a market pull environment (ability to satisfy user
need is the primary factor determining which technology succeeds).
4. In general, achieving higher technical performance is somewhat more important than the effort (cost and
time required) of doing so.
276 SYSTEM DYNAMICS '93
Figure 13 Variations in Cumulative Volume, Technology T2
i Tech Push
12,000 0 Mit Pull
Risk Technical Technical Marketing Price
Performance Effort Spend
‘While these observations are based on a limited set of experiments, should they hold up under more rigorous
testing they point to several guidelines for managers:
1. The technology push situation requires the greatest balance among technical, marketing, and pricing
policies, but marketing and pricing are critical determinants of success: marketing because it can create a
perceived need for the product and because it can reduce perceived risk; pricing, because given the relative
lack of need for a “better mousetrap," price becomes the dominant factor in the price-performance.
tradeoff. Therefore, prices should be set relatively low, and investments recouped over a longer period
(and because the next generation technology is not needed, this generation should have a longer life to
earn back the investment).
2. Ina market pull environment, success is enhanced by getting the best technical performance, as fast a
possible, Once in the market, the product will sell itself. Prices can and should be set relatively high in
order to recoup investments before the next generation of technology is ready (the next generation is likely
to diffuse rapidly, and therefore limit the period over which investments can be recovered).
Further Work
Several parallel activities are in progress:
1. Further sensitivity and policy analyses to more rigorously identify the key factors in the diffusion of new
technologies and to develop guidelines for management policies;
2. Development of case studies and a numerical data base to illustrate the diffusion concepts and policy
guidelines; and *
3. Applications of the model to current technology decisions.
As this work progresses, a body of empirical results together with the simulation model will provide managers with
valuable tools for determining strategies for investment in and marketing of new technologies. By drawing.on the
case studies and heuristics, it should be possible to quickly classify a new technology along critical performance
dimerisions (e.g., user need, rate of market growth, product vs. technology lifetime), develop management
guidelines, and then set up the model to adapt these guidelines to the new situation. Further, the model can be
modified to represent alternative situations. For example, technologies T2 and T3 could represent competing
solutions for the next generation technology, and more technologies added downstream (T4).
SYSTEM DYNAMICS '93 277
