Simulating hype cycle curves with mathematical functions some examples of high tech trends in japan

International Journal of Managing Information Technology (IJMIT) Vol.7, No.2, May 2015
DOI : 10.5121/ijmit.2015.7201 1
SIMULATING HYPE CYCLE CURVES WITH
MATHEMATICAL FUNCTIONS : SOME EXAMPLES OF
HIGH-TECH TRENDS IN JAPAN
Hiroshi Sasaki1
1
College of Business, Rikkyo University, Tokyo, Japan
ABSTRACT
In this study, a method to simulate Gartner’s hype cycle [1] is proposed. A search of the academic
literature on this topic provides no clear guidance on how to draw hype cycle curves with mathematical
functions. This article explores a new process for simulating the curve as a combination of bell-shaped
curves and S-shaped curves, and applies this process to some high-tech innovations in Japan. Trends in
technologies such as customer relationship management (CRM), supply chain management (SCM), and
cloud computing are analyzed by using a corpus of 4,772 newspaper articles. For these examples,
Gompertz functions show better fit than logistic functions. For the combined curve, polynomial functions of
degree 9 provide the best fit, with adjusted R-square values of more than 0.97.
KEYWORDS
Hype cycle, High-tech innovation, S-shaped curves, Diffusion of innovations
1. INTRODUCTION
Gartner’s hype cycle [1] is a popular method for visually showing an ongoing high-tech
innovation process. Fenn and Raskino [2] noted that “the hype cycle’s particular contribution is in
highlighting the challenge of adopting an innovation during the early stages of the innovation’s
life cycle.” Executives and managers check new hype cycle reports as a means of trying to find
new technological trends.
This study explores a new approach for simulating hype cycle curves with mathematical
functions. This paper is organized as follows. The next section reviews the literature related to the
generation of the hype cycle. After this, we propose a three-step process for simulating hype cycle
curves and then apply that process to some high-tech innovations, examining trends in areas such
as customer relationship management (CRM), supply chain management (SCM), and cloud
computing in Japan.
2. LITERATURE REVIEW
2.1. Five key phases of the hype cycle
Gartner’s hype cycle consists of five key phases [1]. The first phase is Innovation trigger
(Technology trigger), which begins when an announcement about a technological development
drives sudden interest [2]. In “Hype Cycle for Emerging Technologies, 2014” [3], “bio acoustic
sensing” appears in the first phase. The second phase, Peak of inflated expectations, begins when

2
publicized stories capture the excitement around the innovation and reinforce the need to become
a part of it [2][4]. In the same report [3], “data science” is shown entering into the second phase,
and the “Internet of Things” is positioned at the top of the peak of expectations, where it displaces
the trend on “big data.” The third phase, Trough of disillusionment, occurs when impatience for
results begins to replace the original excitement about potential value [2]. Fenn and& Raskino [2]
explains that “a number of less favorable stories start to emerge as most companies realize things
aren’t as easy as they first seemed”. In 2014, we see “cloud computing” reaching the bottom of
the trough. During the fourth phase, Slope of enlightenment, early adopters overcome the initial
hurdles, and understanding grows about where the innovation can be used[2]. Three-dimensional
(3D) technologies, such as “Enterprise 3D printing” and “3D scanners,” are in this phase. The last
phase, Plateau of productivity, begins when growing numbers of organizations feel comfortable
with the now greatly reduced levels of risk [2].
Thus, Gartner’s hype cycle [1] clarifies the position of each high-tech innovation. However, only
those in the Gartner organization can create the hype cycle, and researchers outside of Gartner
have no tools to generate it.
2.2. How to measure technology expectations
A critical issue for this study is to provide a measure for technology expectations. To do so, we
searched for empirical studies that meet the conditions below.
1. Source: The articles available in August 2014 in the Academic Source Premier and
Business Source Premier databases of EBSCO Information Services
2. Key word: The phrase “hype cycle” was used for the search.
3. Conditions: The search was restricted to academic journals and periodicals published
in English.
As a result, 25 articles were extracted. We extracted 66 additional articles (including 2 duplicates)
from the Science Direct database by searching for “Gartner’s hype cycle.” After eliminating the
duplicates and 22 articles from fields other than social sciences, 67 articles remained. These
articles were categorized into three types: qualitative analysis (53 articles), quantitative analysis
(9 articles), and other (5 articles; essays, editor’s comments, etc.).
(1) Articles with qualitative analysis
Figure 1 shows the technologies covered by 53 articles that focused on qualitative analysis. In
these studies, researchers try to apply the hype cycle model to education, cloud computing,
security, software, and energy and the environment, among other topics.

International Journal of Managing Information Technology (IJMIT) Vol.7, No.2
Figure 1. Technologies discussed in 53 papers
(2) Articles with quantitative analysis
Table 1 illustrates the measures and data sources employed in the 9 articles for
cycle curves[5][6][7][8][9][10][1
(technology expectations) from
number of items about the technology (news stories, papers, books, and so on); in contrast, patent
statistics are commonly used
productivity.
Table 1. Summary of quan
No Authors
1 Gray et al. (2014)[5]
Accounting
publications
2 Lente et al. (2013)[6]
Voice over internet protocol (VoIP),
gene therapy
superconductivity.
3 Budde et al. (2013)[7]
Hybrid
Fuel
4 Vahid (2012)[8] Unified Modeling Language
5 Jun (2012)[9] Hybrid cars
6 Konrad (2012)[10] Stationary fuel cells
7 Kim et al. (2012)[11] Approx. 500
8
Ruef& Markard
(2010)[12]
Stationary fuel cells
9 Konrad (2006)[13]
Electronic commerce and interactive
television
Technologies discussed in 53 papers with qualitative analysis
(2) Articles with quantitative analysis
Table 1 illustrates the measures and data sources employed in the 9 articles for simulating
11][12][13]. It is popular in these studies to measure the cycle
from Innovation trigger to Trough of disillusionment by counting the
when measuring from Slope of enlightenment to
Table 1. Summary of quantitative measures
Subject Method of Measurement
Accounting-related expert systems
publications
Yearly distribution of expert
systems research
Voice over internet protocol (VoIP),
gene therapy, and high-temperature
superconductivity.
Number of newspaper articles
Hybrid-electric vehicle (HEV) and
Fuel-cell vehicle (FCV) technology
Number of press releases
patent statistics
Unified Modeling Language (UML)
Number of books on
Hybrid cars
Search traffic on Google,
articles, and patent statistics
Stationary fuel cells Number of newspaper articles
Approx. 500 emerging technologies
Papers and patents
information, Decision tree
model
Stationary fuel cells Number of newspaper articles
Electronic commerce and interactive
television
, May 2015
3
simulating hype
to measure the cycle
by counting the
to Plateau of
Method of Measurement
early distribution of expert
umber of press releases and
books on UML
earch traffic on Google, news
patent statistics
umber of newspaper articles
Papers and patents
information, Decision tree

4
The contents of Table 1 are consistent with the findings of Jun[9], who notes that the number of
news stories and patents can well explain consumer behavior along the hype cycle. More
importantly, in the same article, Jun divides the hype cycle into two separate curves, and states
that a) the first curve is a bell curve representing the initial cycle of enthusiasm and
disappointment, and b) the second curve is an S-shaped curve showing how an innovation's
performance improves slowly at first and then accelerates steadily before finally yielding
diminishing returns [9].
We adopt this idea of treating hype cycle curves as comprising two stages. We call them as “the
hype stage” and “the implementation stage”.
(A) The hype stage: This stage covers the period from Innovation trigger to Trough of
disillusionment. The curve for this stage can be constructed as a bell-shaped curve, with
time along one axis and the instantaneous (non-cumulative) number of articles along the
other. One popular way to measure this stage is to use the number of items (newspaper
articles, academic papers, books) mentioning the technology or the volume of search
traffic about the technology as the non-time axis.
(B) The implementation stage: This stage covers Slope of enlightenment and Plateau of
productivity. The curves for this stage can be simulated by S-shaped curves with time
along one axis and cumulative number of articles along the other. In some of the
literature, patent statistics are used for the non-time axis.
3. A PROCESS FOR SIMULATINGHYPE CYCLE CURVES WITH
MATHEMATICAL FUNCTIONS
To position ongoing high-tech innovations along a hype cycle curve, mathematical functions are
needed. This section proposes a three-step process for doing so, with mathematical functions.
(1) Data collection
Similar to previous studies, this paper uses newspaper articles. After collecting newspaper data
for each high-tech innovation, we divide the articles into two stages, (A) the hype stage and (B)
the implementation stage, according to the content of the article. The key issue at this point is
how to determine which stage should be used for each article. Among the titles of the articles, a
substantial number mention either organizational changes or the appointment of managers as
innovation proceeds. Such articles state, for example, “Company X appointed Mr. Y as a new
SCM leader” or “Company X forms a new SCM division.” This type of article indicates that the
mentioned company is in the implementation stage. We can partition articles into one of the two
stages on the basis of this type of content.
(2) Curve fitting
There are several cumulative time series that form an S-shaped curve. To seek the best S-shaped
curve for each stage, two sigmoid functions (Gompertz and logistic) were examined. It should be
noted that, in our previous study [14], we found that Gompertz functions fit better than logistic
functions for some IT innovations. The forms of these functions are given by the following.
1.Logistic function: y = a / (1 + b exp(-k x))
2.Gompertz function: y =a exp ( -exp(–k (x-‫ݔ‬௖)))

These two functions are distinguished by differences in their waveforms. The logistic function
provides a curve that is symmetrical
function forms a curve that is not symmetrical around the point of inflection.
the two functions to the two stages
To form a curve for the hype stage, S
transformed to bell-shaped curves that use non
after data standardization, we obtain an initial hype cycle curve (see the dotted curve in Fig
Figure 3. A sample hype
(3) Polynomial fitting
We conduct polynomial fitting to
Polynomial functions of degrees
1. Polynomial functions of degree 5:
provides a curve that is symmetrical around the inflection point; in contrast, the Gompertz
function forms a curve that is not symmetrical around the point of inflection. This
the two stages separately(Figure 2).
Figure 2. A sample curve fitting
stage, S-shaped curves (formed by using cumulative data) will be
shaped curves that use non-cumulative data. By combining the two curves
after data standardization, we obtain an initial hype cycle curve (see the dotted curve in Fig
Figure 3. A sample hype cycle curve
We conduct polynomial fitting to express the dotted curve with mathematical functions
5, 7, and 9 are tested.
Polynomial functions of degree 5: y ൌ ݅݊‫ݐ݌݁ܿݎ݁ݐ‬ ൅ ∑ ሺ‫ܤ‬௜‫ݔ‬௜ሻହ
௜ୀଵ
Polynomial functions of degree 7:y ൌ ݅݊‫ݐ݌݁ܿݎ݁ݐ‬ ൅ ∑ ሺ‫ܤ‬௜‫ݔ‬௜ሻ଻
௜ୀଵ
Polynomial functions of degree 9:y ൌ ݅݊‫ݐ݌݁ܿݎ݁ݐ‬ ൅ ∑ ሺ‫ܤ‬௜‫ݔ‬௜ሻଽ
௜ୀଵ
, May 2015
5
point; in contrast, the Gompertz
study applies
shaped curves (formed by using cumulative data) will be
cumulative data. By combining the two curves
after data standardization, we obtain an initial hype cycle curve (see the dotted curve in Figure 3).
with mathematical functions.

4. HYPE CYCLE CURVE
INNOVATIONS IN JAPAN
Articles printed in the Nikkei newspaper (Japan’s leading economic newspaper) are used as data
for simulating hype cycle curves. All articles printed in the Nikkei morning edition from 1990 to
the end of March 2014 were searched, and articles containing an
selected: SCM, CRM, and cloud computing. From among all articles,
extracted: 616 articles for CRM; 1,550 for SCM;
4.1. CRM
Figure 4 shows the diffusion process for CRM in Ja
stage represents the non-cumulative number of articles about CRM, and the line graph for the
implementation stage represents the cumulative number of articles on the same topic.
Figure 4. Time series of Nikkei articles about CRM
We fit Gompertz and logistic functions to the two line graphs. As a result, the Gompertz functions
showed better fit than the logistic functions for both stages (see Table
squared values).
Table 2. S
Logistic function
Number of points
Degrees of freedom
Reduced Chi-squared
Residual sum of squares
Adj. R-squared
Gompertz function
Number of points
Degrees of freedom
Reduced Chi-squared
Adj. R-squared
CURVE SIMULATION FOR HIGH-TECH
JAPAN
s. All articles printed in the Nikkei morning edition from 1990 to
the end of March 2014 were searched, and articles containing any of the following terms were
cloud computing. From among all articles, 4,772
extracted: 616 articles for CRM; 1,550 for SCM; and 2,606 for cloud computing.
Figure 4 shows the diffusion process for CRM in Japan. In this figure, the line graph for the hype
cumulative number of articles about CRM, and the line graph for the
Figure 4. Time series of Nikkei articles about CRM
showed better fit than the logistic functions for both stages (see Table 2 for the adjusted R
Table 2. S-shaped curve fitting for CRM
Logistic function Hype stage Implementation stage
oints 16 16
reedom 13 13
uared 161.88294 41.47205
quares 2104.47818 539.13661
0.97206 0.97748
Gompertz function Hype stage Implementation stage
oints 16 16
reedom 13 13
squared 74.1598 18.12732
quares 964.0774 235.6552
0.9872 0.99016
(B) Implementation stage
(A)Hype stage
, May 2015
6
s. All articles printed in the Nikkei morning edition from 1990 to
y of the following terms were
articles were
pan. In this figure, the line graph for the hype
cumulative number of articles about CRM, and the line graph for the
for the adjusted R-
Implementation stage

Table 3 shows the best-fit parameter values for the Gompertz function. After data standardization,
the initial hype cycle with a bell
implementation stage can be obtained (Fig
Table 3. Best
Best fit parameters
Hype stage
Figure 5.Bell-shaped curve and
Next, we conducted polynomial fitting to seek the best
results. The best-fit function was a polynomial of degree 9
By using the parameter values (B
CRM (Figure 6; the circle indicates the position of 2014).
Table 4. Polynomial fitting for CRM
Polynomial,degree
Number of p
Degrees of freedom
Residual sum of
Adj. R-Square
Polynomial,degree
Number of p
Degrees of freedom
Residual sum of
Adj. R-square
Polynomial,degree
Number of p
Degrees of freedom
Residual sum of
Adj. R-square
fit parameter values for the Gompertz function. After data standardization,
the initial hype cycle with a bell-shaped curve for the hype stage and an S-shaped curve for the
implementation stage can be obtained (Figure 5).
Table 3. Best-fit parameter values of Gompertz function
est fit parameters Value
Standard
error
Hype stage a 219.807 5.36602
xc 4.43241 0.16187
k 0.35532 0.03165
Implementation stage a 128.6113 3.96502
xc 5.31204 0.18904
k 0.28213 0.02388
shaped curve and S-shaped curve for CRM (after data standardization)
Next, we conducted polynomial fitting to seek the best-fit functions. Table 4 summarizes the
fit function was a polynomial of degree 9 (adjusted R-squared value of 0.99229).
By using the parameter values (B1to B9 and the intercept), we can draw a hype cycle
6; the circle indicates the position of 2014).
Table 4. Polynomial fitting for CRM
,degree5
points 33
reedom 27
um of squares 1.65385
Squared 0.83371
,degree7
points 33
reedom 25
quared 0.95247
,degree9
points 33
reedom 23
quared 0.99229
, May 2015
7
fit parameter values for the Gompertz function. After data standardization,
shaped curve for the
(after data standardization)
Table 4 summarizes the
value of 0.99229).
and the intercept), we can draw a hype cycle curve for

Figure 6.
4.2. SCM
Similarly, Figures7, 8, and Table 5 show the process to
We obtain Figure 9 (the best-fit function was a polynomial of degree 9
0.99656) as the result of that process.
Figure 7. Time series of Nikkei articles about SCM
Figure 6. Simulated hype cycle curve for CRM
and Table 5 show the process to simulate the hype cycle curve for SCM.
fit function was a polynomial of degree 9 with Adj. R
as the result of that process.
Figure 7. Time series of Nikkei articles about SCM
, May 2015
8
the hype cycle curve for SCM.
Adj. R-squared:

Figure 8. Bell-shaped curve and
Table 5. S
Logistic function
Number of points
Degrees of freedom
Reduced Chi-squared
Adj. R-squared
Gompertz function
Number of points
Degrees of freedom
Reduced Chi-squared
Adj. R-squared
Polynomial
Number of
Degrees of
Residual
Figure 9.
shaped curve and S-shaped curve for SCM (after data standardization)
Table 5. S-shaped curve fitting for SCM
17 17
14 14
uared 380.89083 117.77175
quares 5332.47167 1648.80457
0.98341 0.98166
Gompertz function Hype stage Implementation stage
17 17
14 14
ed 131.17589 45.66755
quares 1836.4625 639.34571
0.99429 0.99289
Polynomial of degree 9
Number of points 34
Degrees of freedom 24
Residual sum of squares 0.03615
Adj. R-squared 0.99656
Figure 9. Simulated hype cycle curve for SCM
, May 2015
9

4.3. Cloud computing
Cloud computing is still in the hype stage and experiencing growth. Figures 10
show the process, and Figure 12 show
function was a polynomial of degree 9 with
proposed method.
Figure 10. Time series of Nikkei articles about cloud computing
Figure 11. Bell-shaped curve and
Table 6. S-shaped curve fitting
Logistic function
Number of points
Degrees of freedom
Reduced Chi-squared
Adj. R-squared
Gompertz function
Number of points
Degrees of freedom
in the hype stage and experiencing growth. Figures 10 and11 and Table 6
12 shows the hype cycle curve for cloud computing
function was a polynomial of degree 9 with Adj. R-squared: 0.97438), as calculated by the
Figure 10. Time series of Nikkei articles about cloud computing
shaped curve and S-shaped curve for cloud computing (after data standardization)
shaped curve fitting and polynomial fitting for cloud computing
oints 8 6
reedom 5 3
uared 2860.65605 27.91975
quares 14303.28026 83.75924
d 0.99464 0.96986
function Hype stage Implementation stage
oints 8 6
reedom 5 3
, May 2015
10
11 and Table 6
the hype cycle curve for cloud computing (the best-fit
calculated by the

Reduced Chi-squared
Adj. R-squared
Number of
Degrees of
Residual
Figure 12. Simulated
5. SUMMARY AND CONCLUSIONS
This study demonstrated a method of
functions. By applying polynomial functions, the current position along the curve
identified. Because of the simplicity and operational
complex simulation technologies,
capture the state of high-tech innovations.
Through the process, this study found some common features among the diffusion patterns of
different high-tech innovations. First,
both the hype stage and the implementation stage
functions in all examined high-tech innovations.
previous study [14], means that the high
point of inflection for both stages
best fit for the combined curve,
for SCM, and 0.97 for cloud computing
roughly by polynomial functions.
In conclusion, it is reasonable to
proposes a process for simulating
that can be used to understand the position along the hype cycle.
noteworthy limitations to this study. First, when examining newspaper articles
implementation stage by using the titles (specifically, titles mentioning organizational changes or
announcing the appointment of managers were taken as indicating the implementation stage).
uared 552.55357 11.79637
quares 2762.76784 35.3891
d 0.99896 0.98726
Number of points 25
Degrees of freedom 15
Residual sum of squares 0.20605
Adj. R-squared 0.97438
Simulated hype cycle curve for cloud computing
CONCLUSIONS
a method of simulating Gartner’s hype cycle[1] with
. By applying polynomial functions, the current position along the curve
simplicity and operational ease of this method in comparison with
complex simulation technologies, the proposed method is suggested for use when trying to
tech innovations.
tech innovations. First, our simulation of S-shaped curves indicated
both the hype stage and the implementation stage Gompertz functions show better fit than logistic
tech innovations. This result, which agrees with results from
means that the high-tech innovation process is not symmetrical around the
for both stages. Second, polynomial functions of degree 9 demonstrated the
with adjusted R-squared values of more than 0.99 for CRM
r cloud computing. This means that hype cycle curves can be simulated
by polynomial functions.
to claim that this study makes the following contributions: (1) it
simulating hype cycle curves, and (2) it shows the mathematica
that can be used to understand the position along the hype cycle. However, t
noteworthy limitations to this study. First, when examining newspaper articles, we identified
sing the titles (specifically, titles mentioning organizational changes or
, May 2015
11
with mathematical
. By applying polynomial functions, the current position along the curve could be
in comparison with other
when trying to
indicated that during
Gompertz functions show better fit than logistic
agrees with results from our
ovation process is not symmetrical around the
. Second, polynomial functions of degree 9 demonstrated the
for CRM,0.99
. This means that hype cycle curves can be simulated
makes the following contributions: (1) it
and (2) it shows the mathematical functions
However, there are two
we identified the
sing the titles (specifically, titles mentioning organizational changes or

12
Despite this, these organizational changes are not the only indicators that could be used to
confirm the implementation stage. Second, we used only polynomial functions of degrees 5, 7,
and 9. We need to apply polynomials of higher degrees, such as 11, 13, and 15, and examine
other functions that to see whether they are more appropriate. Such extensions are left for future
study.
REFERENCES
[1] Gartner,Inc.,(n.d.)“Gartner Hype Cycle”, Retrieved from
http://www.gartner.com/technology/research/methodologies/hype-cycle.jsp (21 March, 2015).
[2] Fenn, Jackie. & Raskino, Mark, (2008) Mastering the Hype Cycle –How to Choose the Right
Innovation at the Right Time, Harvard Business Press.
[3] Gartner, Inc., (2014) “Gartner's 2014 Hype Cycle for Emerging Technologies Maps the Journey to
Digital Business”,Retrieved from http://www.gartner.com/newsroom/id/2819918 (21March, 2015).
[4] Kim,Song-kyoo, (2013) “General framework for management of technology evolution”, The Journal
of High Technology Management Research, Vol. 24, No. 2, pp130-137.
[5] Gray, Glen. Chiu, Victoria. Liu, Qi. & Li, Pei, (2014) “The expert systems life cycle in AIS research:
What does it mean for future AIS research?”, International Journal of Accounting Information
Systems, Vol. 15, pp423–451.
[6] Lente, Harro. Spitters, Charlotte. & Peine, Alexander,(2013)“Comparing technological hype cycles:
Towards a theory”, Technological Forecasting and Social Change, Vol. 80, No. 8, pp 1615-1628.
[7] Budde, Björn. Alkemade, Floortje. &Hekkert, Marko,(2013) “On the relation between
communication and innovation activities: A comparison of hybrid electric and fuel cell vehicles”,
Environmental Innovation and Societal Transitions, Vol.14, March 2015, pp.45-59.
[8] Vahid, Garousi, (2012) “Classification and trend analysis of UML books (1997-2009)”, Software &
Systems Modeling. Vol. 11, No. 2, pp273-285.
[9] Jun, Seung-Pyo,(2012)“A comparative study of hype cycles among actors within the socio-technical
system: With a focus on the case study of hybrid cars”, Technological Forecasting and Social Change,
Vol. 79, No. 8, pp1413-1430.
[10] Konrad, Kornelia. Markard, Jochen. Ruef, Annette. & Truffer, Bernhard, (2012) “Strategic responses
to fuel cell hype and disappointment”, Technological Forecasting and Social Change, Vol. 79, No. 6,
pp1084-1098.
[11] Kim, Jinhyung. Hwang, Myunggwon. Jeong, Do-Heon. & Jung, Hanmin,(2012) “Technology trends
analysis and forecasting application based on decision tree and statistical feature analysis”, Expert
Systems with Applications, Vol. 39, No.16, pp12618-12625.
[12] Ruef, Annette.& Markard, Jochen,(2010) “What happens after a hype? How changing expectations
affected innovation activities in the case of stationary fuel cells”, Technology Analysis & Strategic
Management, Vol. 22, No. 3, pp317-338.
[13] Konrad, Kornelia, (2006)“The social dynamics of expectations: The interaction of collective and
actor-specific expectations on electronic commerce and interactive television”, Technology Analysis
& Strategic Management, Vol. 18, No. 3/4, pp429-444.
[14] Sasaki, Hiroshi, (2014) “Time lags related to past and current IT innovations in Japan: An analysis of
ERP, SCM, CRM, and big data trends”, International Journal of Business Analytics, Vol. 1, No. 1,
pp29-42.

Simulating hype cycle curves with mathematical functions some examples of high tech trends in japan

More Related Content

Similar to Simulating hype cycle curves with mathematical functions some examples of high tech trends in japan (20)

More from IJMIT JOURNAL (20)

Recently uploaded (20)

Simulating hype cycle curves with mathematical functions some examples of high tech trends in japan