Technological Forecasting & Social Change 76 (2009) 754768

Contents lists available at ScienceDirect

Technological Forecasting & Social Change

Monitoring the organic structure of technology based on the patent

development paths
Changwoo Choi, Yongtae Park
Seoul National University, Seoul, Republic of Korea

a r t i c l e

i n f o

Article history:
Received 29 August 2008
Received in revised form 31 October 2008
Accepted 31 October 2008
Keywords:
Technology monitoring
Technology development path
Patent citation network, Systematic approach

a b s t r a c t
As the strategic importance of understanding changes in technology for successful business of
most rms increases, the ability to analyze and monitor the current stage and history of
technology is reckoned as a critical asset both for gaining competitive advantage and
identifying promising niches. Patent citation networks have been widely used for systematic
and empirical analysis of technology development. Understanding of technology's detailed
changes in large patent citation networks, however, is difcult to achieve because of these
networks' large and complex structures. To overcome this problem, we suggest an algorithm
that identies patent development paths from a large patent citation network by evaluating the
weight of citations between patents. We then apply this algorithm to ash memory patents in
an empirical study. Our algorithm is a new methodology that can be used to analyze the
dynamic and complex structure of individual technologies.
2008 Elsevier Inc. All rights reserved.

1. Introduction
Technology is one of the most important elements for providing companies with remarkable revenue in the current
competitive environment. Even when a company dominates a competitive market based on its technological advantage, the
company should continue technology development activity to create dominant products or services by identifying, adopting, and
leading the changes in technology in the competitive eld. Hence, companies operating in competitive environments demanding
new product development, process improvement, and technology-enhanced services must obtain and organize information on
emerging technologies [1].
Firms may conduct research and development (R&D) activities and invest in technology, but it is not easy for rms to orient
their strategies to this technological environment and use them to their own benet [2]. This is why investing based on forecasts of
promising markets, preparing products, and selecting emerging technologies for those products is very risky. In this process, one of
the critical factors for rms to establish technological strategies is to identify and understand the technological development
trends. The increasing importance and benets of technology have led to a wide range of applications and studies examining the
management of technological forecasting. The objective of these applications and studies is to track developments in a particular
area that may serve to forecast technology [3], set up research-funding priorities [4], investigate product development history [5],
monitor technology trends [6], integrate technology management processes [7], identify technological opportunities [1], and
visualize technological information [8].
Previous approaches to analyzing and forecasting technological development, however, have several limitations. First, these
approaches cannot provide systematic information about the technology development process based on objective technological
data. Objective information is a critical factor in successful technological forecasting. It is, however, very difcult to explain the

Corresponding author. Department of Industrial Engineering, School of Engineering, Seoul National University, San 56-1, Shillim-Dong, Kwanak-Gu, Seoul 151742, Republic of Korea. Tel.: +82 2 880 8358; fax: +82 2 889 8560.
E-mail address: parkyt@cybernet.snu.ac.kr (Y. Park).
0040-1625/$ see front matter 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.techfore.2008.10.007

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

755

Table 1
Technological forecasting techniques [15].
Technique

Characteristics

Description

Consensus
method
Delphi
method
Structural
models
Scenarios

Subjective procedure

Panel of experts debate the subject face-to-face. It is an intuitive procedure

Subjective procedure

Panel of experts answer several rounds of questionnaires, but they usually do not meet each
other face-to-face. It is
Attempting to develop a mathematical or analytic model for accomplishment of the forecasting

Quantitative procedure

Combined procedure of subjective, cognitive

and quantitative procedure
Technological Combined procedure of subjective, cognitive
vigil
and quantitative procedure

It is not a formal technique. It serves as a guide which helps to foresee the future. It tires to
identify treats and opportunities for the rms
It is the administration of the ow of the scientic, technical and technological information, in
order to aid the innovation process

detailed processes of technology development using traditional forecasting methodologies. Hence, providing sufcient
information about technological change is still necessary for successful technological forecasting. Second, a systematic approach
is necessary to explain the detailed development processes of technology. Even though various methodologies such as statistical
analysis, diffusion modeling, and trend extrapolation can be applied to technological forecasting to enhance the objectivity of
analysis results, they cannot explain the complex structure of the detailed development of technologies. They can describe only the
overall directions and processes of technological development at the macro level. To overcome these limitations, we take a
systematic approach to identifying organic and complex structures of current technology.
In this context, we propose a technology development map to analyze changes in technology in detailed micro-level. This map
is developed using patent analysis, because this is the most widely used methodology in formalized and systematic approaches to
identifying and managing technological change. Patents are the primary output of the R&D activity of rms. The utilization of
patent data, however, is not limited to protecting the legal right to developed technology. Patents and patent data can support
many aspects of technology management [9].
In the remainder of the paper, rst we describe technological forecasting, patent analysis, and patent citation analysis as the
background for this research, in Section 2. Second, the proposed algorithm to identify the patent development path is explained in
Section 3. As an example and application of the suggested approach, development paths for ash memory system patents are
presented and discussed in Section 4. The implications of this research are discussed in Section 5. Finally, in Section 6, we conclude
with a summary and the implications of our results.
2. Background
2.1. Technological forecasting
Technological forecasting is of great interest for both theory and practice in the establishment of technological strategy and
planning. Technological change requires a new set of engineering and scientic principles. It may reinforce the dominance of rms in
the market or open up new market [7,1012]. On the other hand, many studies show that an inadequate reaction to technological
change may lead to the demise of established company [10,13,14]. Insufcient information on technological trends and managerial
incompetence are the main reasons for failure in the market. Hence, societies, scientists, planners and decision-makers should
endeavor to discover the current status of technology and anticipate future events [15]. This has been called technological forecasting.
Many researchers have suggested various observations of technological trends to forecast technological change at an early stage
and increase the effectiveness of technological decision-making. In the literature, many terms are used for this process of
acquisition, assessment and communication of information on technological trends to detect opportunities and threats in a timely
manner [7]. Furthermore, a detailed analysis of studies that were carried out in different industries shows that the ability of rms
to forecast technological change is a major factor in managing the risk of organizational failure in the face of rapid technological
development [16,17].
Porter et al. (1991) argued that the cornerstone for technological forecasting is identication of current technology [18]. It is vital in
its own right to comprehend who is doing what now with respect to a given technology. This underpins forecasting in two critical
ways forthcoming technological change is foreshadowed by current developments and will be inuenced by changes in related
technologies, and relevant contextual inuences are the most essential ingredients in effective technological forecasting [19].
Typical techniques for identifying and forecasting technological change are the consensus method, the Delphi method,
structural models, scenarios and technological vigilance. The characteristics of these techniques may be described with subjective
procedure, quantitative procedure and combined procedure as shown in Table 1 [15]. First, the consensus method and the Delphi
method are subjective procedures. The subjective procedure depends on the experts' qualitative and intuitive knowledge. This,
however, may be biased because of the subjectivity of experts' knowledge and opinion. Moreover, even after a basic innovation,
some experts may be too pessimistic about progress because they are familiar with the problems and difculties of new technology
[20,21]. The structural model, as a representative quantitative procedure, may eliminate these subjective factors. This model
isolates certain elements that are important to the technological generation process, explaining and expressing mathematically
some of the functional relationships among the elements involved. These models, however, tend to be abstractions. Omissions of

756

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

certain elements that are not judged to be relevant for the formulation of the model may occur. The combined approaches help to
identify threats and opportunities in the decision-making process through the information collected. These approaches, however,
also require a given amount of information and it is still difcult to obtain informative data and to provide objective information for
technological forecasting for technological strategizing [15]. In this context, a systematic approach is needed to gather and analyze
data based on scientic methodologies for technological forecasting.
2.2. Patent analysis
Patents are major outputs of research and development. Hence, they represent the origin and features of a new technology.
They have long been recognized as a very fruitful source of data for the study of innovation and technical change and thus for
technology management research. There are, of course, important limitations on using patent data as an indicator of technology
development. First, not all inventions are patented, because not all inventions meet the patentability [22]. Second, the inventor has
to make a strategic decision to le for a patent registration, as opposed to relying on secrecy or other appropriate means. This
strategic decision-making differs across companies and industries [23]. The results and interpretations of patent analysis are,
therefore, not consistent across technology elds. Third, changes in patent law over the years make it difcult to analyze trends
over time. The protection afforded to patentees worldwide has been improved since the early 1980s, and companies are now more
inclined to le for a patent registration than used to be the case [22].
Patent data, however, should not be discarded entirely as a statistical indicator simply because of these limitations. In fact,
quantitative and empirical analyses of technological innovation using patents have been performed with various objectives [24].
Consequently, patent data represent a unique opportunity to satisfy the need for conceptual or qualitative analysis of technological
change [25]. Furthermore, patent data can empirically explain most aspects of technological innovation in developed countries [23]. In
summary, patent information can help researchers and technology developers make technological decisions, facilitate policy-making
and long-term, national R&D strategies in government ofces and assist in R&D strategy-making and management in individual rms.
Research related to patent analysis can be divided into macro-level research of national or industrial analysis and micro-level
research of particular technological analysis and forecasting. In the macro-level research, the major topics are the economic effect
of technological innovation [26], and the evaluation of technological competitiveness of nations [27]. In the micro level, research
activities, such as identifying technological advantages and disadvantages of competitors and planning of technological
development activities, are conducted with a focus on data related to individual rms [28,29]. This patent information can help
practitioners derive R&D priorities [30], draw a patent map to discover technological vacuums [31], analyze technological trends
and opportunities [32,33], and examine the effects of technological change on rm performance [34].
2.3. Patent citation network
Patent citation analysis is widely used method for advanced analysis on technological change. It is conducted through the
examination of citation links among various patents. Patent citation information can be used to analyze technological valuation,
impact, or diffusion [35]. The number of patent citations can be used as an objective indicator of the market value of innovation
output, or to estimate the rm's value for M&A [36]. Additionally, an economic model for measuring the international knowledge
ow has been developed by using patent citation information as a proxy for knowledge ow between actors and investigating coauthorship, citation, and scientic activities [37,38].
The objective of this research is to identify change in the development and innovation of technology at the micro-level. To this end,
analysis on citation information of individual patent is required. Recent works have attempted to construct a patent citation network at
the individual patent level [3942]. In particular, Mina et al. (2007) constructed a large citation network and applied a path search
algorithm to it to analyze the growth and transformation of medical knowledge [40]. This research was conducted based on an
algorithm suggested by Hummon and Doreian (1989) [41] and developed and utilized by Verspagen (2007) [42]. Hummon and Dorian
(1989) suggested an algorithm for evaluating the weight of an arc (a patent citation in a patent citation network) in a large network
graph. Based on this research, a few studies have been carried out on enhancing the efciency of search path algorithms and on
constructing patent citation networks in the eld of fuel cells applying Hummon and Dorian's (1989) methodology [40,42].
However, these studies have some limitations. Even though early research by Hummon and Doreian (1989) is worthwhile, their
basic algorithm is simply based on the exhaustive path search algorithm of general network theory. This can be an advantage from the
perspective of efciency, since such a method uses existing research in the network theory eld; however, the method doesn't apply
the innate characteristics of patent citation analysis. Specically, the method results in a simple analysis of network structure without
the applying the characteristics of patent citation. Other studies are also extensions of Hummon and Dorian's (1989) DNA algorithm or
use applications of the algorithm provided by commercial software for social network analysis, such as Pejek and Ucinet. The present
research suggests a developed algorithm for identifying the path of a patent's development, applying the characteristics of patent
citation to overcome the limitations of previous research.
3. Suggested algorithm for patent development paths
The number of patents is escalating as the competitiveness of technology development is intensifying and the speed of
technology development is accelerating. Hence, it is difcult to identify and understand the detailed structure of technology
development at the level of individual patents with the large patent citation network, because it has too complex structure. To

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

757

Fig. 1. Overall process of suggested algorithm.

overcome this problem, a systematic algorithm is suggested to extract the main paths of patent development and track history of
technology development from patents and patent citation information.
The algorithm suggested for tracing the paths of patent development is composed of the six steps as shown in Fig. 1. First, a
patent's data is collected from the public patents database. Second, a patent citation matrix is constructed, and patents are
classied into four groups based on citation relationships. Third, the weights of all arcs in the patent citation matrix are calculated
based on the number of forward citations. Fourth, all origin patents among the collected patents are selected as the starting points
of the patent development path. Fifth, main development paths among arcs are selected recursively based on the weight of the
arcs. Finally, this algorithm terminates when all of the patent development paths from every origin patent meet the terminus
patents. The detailed process of this algorithm is explained in following sections.
3.1. Step 1: data collection
Patent information should be collected as data to monitor and track the technology development. Patent data is widely used to
explain technology development, and this data is provided to public users as an open database. As the rst step, patents in given
technology eld are collected based on the various search conditions from patent database.
3.2. Step 2: construction of the patent citation matrix
A general network consists of nodes and arcs which link two nodes. Columns and rows of matrix represent the nodes and each
value in the matrix is dened as the existence or strength of arc between two nodes. The nodes and arcs in a patent citation matrix
are dened in this step. The patent citation matrix represents citations among patents. Nodes represent the individual patents, and
arcs between two nodes are citations. The patent citation matrix P is dened with these nodes and arcs as follows:
0

1
p11 p12 : : : p1n
B p21 p22 : : : p2n C
C
P =B
@ v
v
O
v A
p
pn2 : : : pnn
 n1
1 : if patent jcites patent i
pij =
0 : otherwise
The patent citation matrix P is a n n square matrix, where n is the number of patents. If the nodes are arranged by increasing
or decreasing order of patent number, the matrix P is a triangular matrix. The patent citation network is a directed network,

758

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

Table 2
Types of patents in a citation network.
Types of patents

Description

Isolated patents
Origin patents
Terminus patents
Intermediate patents

Patents
Patents
Patents
Patents

that
that
that
that

Condition
have no citation and are not cited
have no citation, but are cited
have citations, but are not cited
have citations and are cited

dI(ni) = dO(ni) = 0
dI(ni) = 0 and dO(ni) N 0
dI(ni) N 0 and dO(ni) = 0
dI(ni) N 0 and dO(ni) N 0

therefore all arcs have a direction. An individual matrix element pij is the directed linkage from node i to node j, representing the
fact that patent j cites patent i. In general network theory, node i is named as the sender and node j the receiver [43]. Using this
directed network, various approaches have been proposed for explaining the relationships and characteristics of an individual
node and arc. Wasserman and Faust (1994) classify the nodes in a directed network into four types [43]. In the same manner, we
dene the four types of patents in this research based on their classications, as shown in Table 2. The rst column shows the types
of patents, and the second column describes the citation relationships of each type. The last column shows the mathematical
conditions on the types of patents. The variable dI(ni) is the number of nodes adjacent to the node i and represents the number of
backward citations for patent i. Also, dO(ni) is the number of nodes adjacent from the node i, representing the number of patents in
which patent i is cited, i.e., the number of forward citations.
3.3. Step 3: calculating the weights of arcs
As explained above, it is difcult to understand a complex patent citation network in its entirety. Hence, this research proposes
a methodology for identifying the main development path, which aids in understanding the history of a technology. The main arcs
are used to identify the patent development path. Because this main path is selected based on the weight of all arcs, the method for
dening the weight of arcs is the most important part of identifying the patent development path.
We call the indicator of the weight of the arcs the forward citation node pair (FCNP). The FCNP considers the node pair linked by
a selected arc. The FCNP is calculated by multiplying the number of forward citations of the two linked nodes. In this calculation,
the number of forward citations includes the linked patent itself, thereby preventing the weight of terminus patents from being
calculated as zero. This process is illustrated in Fig. 2.
In Fig. 2 FCNP (Aij) is the weight of arc ij, where patent j cites patent i. Also, ni and nj are the number of forward citations of
patent i and patent j. This FCNP (Aij) indicator can be calculated for each arc in the patent citation network, and the patent
development paths are dened based on the FCNP (Aij) indicators.
The proposed FCNP considers the outward linkages, that is, forward citations. Forward citation and backward citation have
different utilities in patent analysis. Backward citations are used to investigate knowledge ows, or spillovers between patents or
technologies; and forward citations are applied as a measure of inventive quality in terms of technological or economic value
[23,39]. Hence, only the number of forward citations is applied to the indicator to reect the technological or economic value of
patents in selecting the patent development paths.
3.4. Step 4: selecting origin patents
All origin patents in the original patent citation network are selected in this stage. The origin patents are developed and
registered in the early time of technology history. These patents are the starting point of development paths that derived from the
original patent citation network. Hence, the identication process starts from the origin patents.
3.5. Step 5: selecting the linked patent with the highest FCNP
The next step is extending the patent development path from the origin patents. Every arc starting from an origin patent is
selected. These arcs are linked to the patents that cite the origin patent. Arcs are evaluated based on the comparison of the FCNP of

Fig. 2. Calculation of weight of arc ij.

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

759

Table 3
Large components in the network of patent development paths for ash memory system.
Components

Size (number of patents)

Components

Size (number of patents)

C1
C2
C3
C4
C5
C6
C7

113
33
30
29
28
21
20

C8
C9
C10
C11
C12
C13

17
16
11
11
10
10

each arc. One or more arcs that have the highest FCNP among linked arcs from the origin patent are selected. In this process,
generally only one arc is selected. Two or more arcs, however, can be selected if the FCNPs of these arcs are same. The selected arcs
are the rst elements that compose the development path from an origin patent.
A patent located at the end of the selected arc becomes a new starting point for another arc of development path. That is, the arc
and patent located at the end of an arc are reclusively added to the development path by evaluating the arcs linked to the recently
added patent based on the FCNP.
3.6. Step 6: evaluation of the exit condition
The algorithm is terminated when a latterly selected patent is a terminus patent. Viewing the algorithm as a whole, the patent
development path starts from origin patents and ends at terminus patents. The origin patents are the oldest patents and the
terminus patents are most recently registered patents. The process of identifying a patent development path for selected patents in
a given eld of technology is performed from every origin patent to every terminus patent. Therefore, various patent development
paths in a specic technology eld can be identied with this process.
Our algorithm keeps the highly weighted arcs in the original patent citation network and reduces the arcs linked from the
origin patents. Patents that are linked to the selected arcs, hence, can be evaluated as highly valued patents. Also, patents located at
the position where various development paths gather are interpreted as points of converging technology, where a technology is
developed by converging two or more technologies that have different objectives or characteristics. These patents can be the main
targets of technology monitoring.
4. Illustrative example: Patent development path in ash memory
4.1. Data collection
In this section, the development paths of ash memory patents are analyzed using the suggested algorithm. Patent data for
ash memory were collected from the United States Patent and Trademark Ofce (USPTO) homepage. The search keyword was
ash memory system, and a total of 2796 patents before June, 2007 were collected.
4.2. Components in the patent development paths
The components in a general network graph mean the connected subgraphs. In a directed graph, the two nodes are in the same
weak component if all pairs of nodes are weakly connected [43]. Various components are identied by applying the algorithm after
eliminating the isolate patents among the 2796 patents collected. Each component in the network of patent development paths
can be interpreted as an independent technology group. Table 3 shows the thirteen large components in the network of patent
development paths for ash memory systems.
4.3. Patent development paths for the largest component
As an illustrative example, we constructed the patent development paths for the largest component in order to apply the
proposed algorithm. From the thirteen components, we selected component 1, which has 113 patents, to draw the patent
development paths shown in Fig. 3. The color of each node denotes the types of patents. White nodes are origin patents, gray nodes
are intermediate patents and black nodes are terminus patents. The direction of the arrow represents the development direction of
forward citations. The main development paths for ash memory systems are composed with connected arcs from many origin
patents to many terminus patents. Of these patents, some can be evaluated as valuable patents located where two or more
development paths gather. For example, 6901499 (System and method for tracking data stored in a ash memory device) is located
in the position at which ve development paths from various origin patents gather and develop toward the new technology,
7076599 (Transactional le system for ash memory) and 7178061 (Power failure detection and correction in a ash memory
device). And two terminus patents, 7111121 (USB storage device and program) and 7102671 (Enhanced compact ash memory
card), can be evaluated as valuable patents since they are located in the gathering positions.

760

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

Fig. 3. Patent development path for ash memory systems.

4.4. Clustering of the largest component

We carried out clustering analysis to understand the detailed inner structure of the networks. Hierarchical clustering analysis
was applied to the largest component of the 113 patents in the previous section. Five clusters were delivered based on the citation
relationship. The title and patents of each cluster are described in Table 4, and Fig. 4 shows the result of clustering analysis. We
asked two experts on ash memory system to make the titles of clusters. The titles were decided by discussion of the experts on
ash memory system. Cluster 1, programmed value sensing, includes technologies used for sensing programmed value in memory

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

761

Fig. 4. Clustering result of the largest component.

cells. There are two kinds of technologies in cluster 2: software download to ash memory and diagnostic systems. Patents for
data-handling technology are gathered in cluster 3. Cluster 4, le systems, includes data recovery, tracking data, transactional le
systems, and power failure detection technologies for ash memory systems. Patents in cluster 5 are practical applications of
removable storage or portable memory.
4.5. Converging patents in patent development paths
Of the various patents in this map, those located at the position where various development paths converge are interpreted as
converging patents. Because this map is drawn by starting from origin patents and selecting arcs of the highest weight, it is very

762

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

Table 4
Five clusters of patents for ash memory systems.
Cluster

Title

1
2
3
4
5

Programmed value sensing

Software download to ash memory and diagnostic system
Data handling
File system
Usage as portable memory

likely that patents located at the converging positions would be economically or technologically valuable ones. Furthermore, these
patents are expected to serve as seeds for another advanced technology. Table 5 shows the converging patents in each cluster and
additional attributes of registered year, main classications, type and number of inows.
4.6. Development stages of ash memory technology
Visualization of the technology development over time based on the changes of development path map can provide more
useful information for identifying the complex structure of technology. The development process of ash memory technology,
especially, selected 113 patents in ve clusters (as mentioned before, these patents are not all patents for ash memory
technology because the only one component of 113 patents is selected from the constructed patent development paths for the
convenience of illustration) can be divided into four stages. Table 6 shows the development process from introduction stage to
converging stage.
First, early origin patents, especially, included in the cluster 2 (Software download to ash memory & diagnostic system) and
cluster 4 (le system) started to be registered at the introduction stage in Fig. 5. This means that the technology for le systems and
software download was developed earlier than other technology elds. Representative patents cited more than other patents in
the introduction stage are 5440632 (reprogrammable subscriber terminal) and 5463766 (system and method for loading
diagnostics routines from disk) in cluster 2 and 5410707 (bootstrap loading from external memory including disabling a reset from
a keyboard controller while an operating system load signal is active) and 5463757 (command interface between user commands
and a memory device) in cluster 4.
Second, at the growth stage, origin patents in all ve technology elds in Table 4 were developed and early part of technology
development path can be seen in cluster 3 (data handling) and cluster 4 (le system) in Fig. 6. Technology development is most
active in cluster 4. 6279069 (interface for ash EEPROM memory arrays) is outstanding as a converging patent in cluster 4.
Third, at the maturity stage, there are various technology development paths in all technology elds including cluster 1
(programmed value sensing), cluster 2 (software download to ash memory and diagnostic system), and cluster 5 (usage as
portable memory) in Fig. 7. This means that technology development activity was carried out actively at this stage and many
outputs were registered as patents. Various converging patents were emerged and the structure of the development path became
more complex. Converging patents such as 6615404 (method and apparatus for downloading software into an embedded-system),
6901499 (system and method for tracking data stored in a ash memory device), and 6842794 (method for starting a data
processing system via a ash memory device) are observed.
Table 5
Converging patents in patent development paths.
Cluster

Patent number

Year

Main class

Type

Number of inows

1
2

6914823
6449735
6615404
7055148
7000064
7102671
7215580
6279069
6901499
6842794
7111121

2005
2002
2003
2006
2006
2006
2007
2001
2005
2005
2006

365/185.22
714/25
717/173
717/172
711/103
348/231.9
365/189.02
711/103
711/205
710/10
711/115

Terminus patents
Intermediate patents
Intermediate patents
Terminus patents
Intermediate patents
Terminus patents
Terminus patents
Intermediate patents
Intermediate patents
Terminus patents
Terminus patents

3
4
4
3
3
3
3
7
5
4
4

4
5

Table 6
Development stages of ash memory technology.
Stage

Description

Development path map

Introduction stage ( 1995)

Growth stage ( 2001)
Maturity stage (2005)
Converging stage (2007)

Development of early origin patents

Development of origin patents in all ve technology elds
Expansion of technology development paths
Emergence of links among various technology elds

Fig.
Fig.
Fig.
Fig.

5
6
7
4

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

763

Fig. 5. Technology development map for ash memory at introduction stage (till 1995).

Finally, all technology elds from cluster 1 to cluster 5 were connected at the converging stage in Fig. 4. Many converging
patents were registered and linked among technology elds. Representative converging patents are 7102671 (enhanced compact
ash memory card), 7215580 (non-volatile memory control), 7076599 (transactional le system for ash memory) and 7111121
(USB storage device and program).
With this technology development map on ash memory system, we asked two experts to identify key patents on ash
memory. Patent 5627784 was selected as one of the core ash memory technologies. This patent includes technology for
nonvolatile data storage structure for memory, and control parameters and methods. A NAND ash memory using the technology
in that patent has been produced. Flash memory is evaluated in the semiconductor market as a fusion semiconductor that will
migrate previous high-quality technologies, single-level cell (SLC) and multilevel cell (MLC), into one chip. This technology

764

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

Fig. 6. Technology development map for ash memory at growth stage (till 2001).

integrates SLC's high functionality and MLC's high capacity. This patent is one example that shows that the suggested map reects
the development of technology and product in the market.
5. Discussion
The role of technology as a resource for building competitive advantage has been critical not only in manufacturing
industries, but also in service industries. With the increasing importance of technologies, practitioners are interested in learning
how to manage and plan technologies more effectively, and how to develop organizational strategies. However, technology
management is a challenging topic, because technological complexity, the rate of technological change, and global technology

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

765

Fig. 7. Technology development map for ash memory at maturity stage (till 2005).

sourcing are more dynamic than ever. In this context, this research provides a systematic tool for analyzing complex and
dynamic structures of technological change to support strategic technology management. In particular, the patent-based
technology development map identies the development paths of a technology. It can be a useful tool to represent the dynamic
history of technology development and to understand its past and present. It shows the dynamic path of technology
development and also reects the composition of increasingly complex systems. It is a helpful tool in ascertaining likely future
development paths based on the objective database, that is, each path to the terminal patents is a possible path for likely future
technology development.
This can help researchers, R&D managers, and decision makers to evaluate technology, analyze competitors more objectively,
and to develop technology strategy. First, it can be a data source for decisions on technology transactions. It can be used to identify

766

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

Table 7
Comparison of proposed approaches with general technological forecasting methods.
Technique

Characteristics

Comparison with proposed approaches

Consensus
method
Delphi
method
Structural
models

- Provide various options for experts to formulate more correct

technology forecasting
- Share and purify knowledge by group dynamics

- Use practical technology database to identify current trends of technology

development

- Eliminates subjective factors

- Focus on the analysis of the complex structure and dynamic change of the
structure

Scenarios

- Explicate and express mathematically some functional

relationships among the elements involved
- Identify threats and opportunities for business

Technological - Help manager by providing information in the decision-making

vigil
process

- Provide objective information about the current structure and change of

technology
- Identify and understand the complex structure of technological
development based on systematic methodologies

the core technology in a specic technology eld. For example, if a rm seeks a technological resource and legal rights for a le
system using ash memory technology, 6279069 (interface for ash EEPROM memory arrays) and 6901499 (system and method
for tracking data stored in a ash memory device) are the rst candidates for a technology transaction. Second, opportunities for
licensing patent rights can be discovered using the suggested map. We can understand the trend of technology development and
identify the companies that have the legal right to a specic technology by tracking the development paths. This process can be
used to determine the licensing opportunities required for the business operation of a company. Third, this can provide objective
information for the development of new technology. The development paths represent the development history of technology
itself. Therefore, we can identify the development of a specic technology. For example, if a rm is preparing for the development
of the next-generation portable USB memory, the R&D manager can track the development paths of cluster 5 in the suggested
technology development paths.
The patent-based technology development map proposed in this paper has different characteristics than traditional
technological forecasting techniques (Table 7). First, the consensus method and the Delphi method are subjective techniques
though they are used differently for specic forecasting processes. This paper applies a systematic approach to draw objective
analysis results from a practical technology database. Second, a structural model eliminates subjective factors and expresses
development trends using mathematical functions. However, the forecasting area that may be analyzed based on mathematical
functions is limited. Furthermore, mathematical functions can deal only with overall trends at the macro level. Even though this
approach is very useful for forecasting sales, since demand and stock prices support strategic decisions surrounding general
management issues, it is not applicable for wide use in technological forecasting. Therefore, quantitative serial data are
inadequate for technological forecasting because technology is naturally uncertain and unclear. Therefore, patent-based
technology development paths focus on the analysis of the complex structure and dynamic change of the structure, and not on
the forecasting of numerical changes in technological data. Third, the scenario model serves as an outline to use in predicting the
future as it is a combined procedure of intuitive cognitive and mechanistic approaches. The accuracy and usefulness depends on
how rich the information about current trends is in these kinds of approaches. In this context, the main purpose of these
proposed methodologies is to provide objective information about the current structure and change in specic technology
elds. Finally, technological vigilance includes various analytical methods for identifying and understanding technology
development. The main purpose of this technique is to provide information to managers to help in the decision-making process.
However, it is difcult to provide objective information about technological development because technology is generally
embedded in products, processes or services. In this context, patent-based technology development analysis are practical
approaches to identifying and understanding the complex structure of technological development at a detailed level based on
systematic methodologies.
The method for analysis of complex structure and change in technology proposed in this research provides benecial
information for establishing technological strategy in the private sector and technology policy in the public sector. Our approach
overcomes the problems with traditional patent analysis in the following ways. First, systematic and quantitative analysis of
technology development may be conducted. It is possible to analyze issued patent data from an empirical perspective to
understand the process of technology development. Second, this quantitative analysis supports expert intuition and judgment.
This research may also provide valuable support for decisions on investment in and development of specic technologies in both
the private and public sectors. Third, we may draw a technology development map for multiple technology elds. Analysis by
experts may provide accurate and effective output in single specic technology elds; however, it is difcult to assemble all the
expert opinion required to assess all technology elds. By comparison, the analysis in this paper covers the wide scope of
technology elds while using a specic patent database. Fourth, comparisons between technology elds may be conducted using
these maps. Our approach may also be used to analyze the complex and dynamic structure of technology development. In previous
research, various patent analyses were conducted based on the simple counting of patents. Such a traditional approach is useful for
understanding the overall trend at the macro level, but it fails to illustrate the detailed structure and trends of technology
development at the micro level. These results may be used to understand these complex relationships and developments.
These results should allow decision makers to (1) assess the attractiveness of technologies, especially new technologies posing
a threat, or a new opportunity for existing business, (2) recognize strategic changes in the rm's competitive environment, (3)

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

767

identify and assess external sources of knowledge generation, and (4) assess the patent situation in new business areas that may be
explored. For these reasons, the results may be taken as a strategic information source, contributing to the effective and efcient
management of technology.
6. Conclusion
We have suggested an algorithm for identifying patent development paths based on patent citation analysis and we have
constructed an example of patent development paths for ash memory systems. By identifying the development history of
individual patents, the proposed map provides insights into the technology development process and into forecasting the direction
of technology development. It can be a powerful tool to identify the historical development of technology. It represents the
dynamic paths of technology development and simplies the complex and organic structure of technology development.
Moreover, this map can be used to obtain ideas for new technology development, and identify core technology and search for
patent technology for licensing.
We believe that with the suggested algorithm it is possible to understand the microprocess of technology innovation and that
this algorithm is applicable to various research domains. However, this research also has some limitations. First, it does not
overcome the innate limitations of patent analysis, that is, fundamental problems such as the differences in patent application and
registration strategy across various industries and the inuence of legal change [22]. Second, information loss occurs during
analysis because only the major development path is constructed, and patents on other paths are discarded. Analysis of the
unselected paths and patents cannot be conducted with the suggested algorithm. This is why the main objective of our algorithm,
is to effectively visualize technological development and to enhance the readability of large patent maps. Trade-offs between
readability and information are inevitable. Third, the algorithm suggested in this paper is focused on a limited sample. That is, the
development paths are constructed based on only the retrieved patents. To identify development paths outside the selected
patents, additional retrieval of patents is necessary. This, however, can be an expanded application of the suggested algorithm. For
example, the development paths over two or more technology elds can be identied with patents retrieved from those
technology elds.
Some possible future research topics related to this work are as follows. First, additional indicators could be developed based on the
suggested algorithms. Second, comparisons between industries or countries based on the suggested map and indicators could be
conducted. Third, the suggested algorithm starts from origin patents, screens the highly weighted arcs and selects intermediate patents
and terminus patents. This process focuses on high-value, recent patents and those located at converging points. If we began this
algorithm with terminus patents, we could analyze the important early patents and identify those located at points of divergence.
References
[1] A.L. Porter, M.J. Detampel, Technology opportunities analysis, Technol. Forecast. Soc. Change 49 (3) (1995) 237255.
[2] H. Taskin, M.R. Adali, E. Ersin, Technological intelligence and competitive strategies: An application study with fuzzy logic, J. Intell. Manuf. 15 (4) (2004)
417429.
[3] M.E. Mogee, Using patent data for technology analysis and planning, Res. Technol. Manag. 34 (4) (1991) 4351.
[4] C. van der Eerden, F.H. Saelens, The use of science and technology indicators in strategy planning, Long Range Plan. 24 (3) (1991) 1825.
[5] M. Iansiti, Technology integration: managing technological evolution in a complex environment, Res. Policy 24 (4) (1995) 521542.
[6] S. Bhattacharya, M.T.R. Khan, Monitoring technology trends through patent analysis: a case study of thin lm, Res. Eval. 10 (1) (2001) 3345.
[7] E. Lichtenthaler, Managing technology intelligence processes in situations of radical technological change, Technol. Forecast. Soc. Change 74 (8) (2007)
11091136.
[8] D. Zhu, A.L. Porter, Automated extraction and visualization of information for technological intelligence and forecasting, Technol. Forecast. Soc. Change 69 (5)
(2002) 495506.
[9] H. Ernst, Patent information for strategic technology management, World Pat. Inf. 25 (3) (2003) 233242.
[10] R. Henderson, K.B. Clark, Architectural innovation: the reconguration of existing product technologies and the failure of established rms, Adm. Sci. Q. 35 (1)
(1994) 930.
[11] F.T. Rothaermel, Complementary assets, strategic alliances, and the incumbent's advantage: an empirical study of industry and rms effects in the
biopharmaceutical industry, Res. Policy 30 (8) (2001) 235251.
[12] J.D. Linton, Forecasting the market diffusion of disruptive and discontinuous innovation, IEEE Trans. Eng. Manage. 49 (4) (2002) 365374.
[13] C.M. Christensen, F.F. Suarez, J.M. Utterback, Strategies for survival in fast changing industries, Manage. Sci. 44 (12) (1998) 207220.
[14] M.A. Schilling, Technological lockout: an integrative model of the economic and strategic factors driving technology success and failure, Acad. Manage. Rev. 23
(2) (1998) 267284.
[15] A.D. Lemos, A.C. Porto, Technological forecasting techniques and competitive intelligence: tools for improving the innovation process, Ind. Manage. Data Syst.
98 (7) (1998) 330337.
[16] J.M. Utterback, Mastering the Dynamics of Innovation, Harvard Business School Press, Boston, Massachusetts, 1994.
[17] C.M. Christensen, R.S. Rosenbloom, Explaining the attackers advantage: technological paradigms, organizational dynamics, and the value network, Res. Policy
24 (2) (1995) 233257.
[18] A.L. Porter, A.T. Roper, T.W. Mason, F.A. Rossini, J. Banks, Forecasting and Management of Technology, Wiley, NY, 1991.
[19] R.J. Watts, A.L. Porter, Innovation forecasting, Technol. Forecast. Soc. Change 56 (1) (1997) 2547.
[20] M. Godet, Reducing the blunders in forecasting, Futures, 15 (3) (1983) 181192.
[21] F. Betz, Forecasting and planning technology, in: G.H. Gaynor (Ed.), Handbook of Technology Management, McGrow-Hill, NY, 1996.
[22] H. Dernis, D. Guellec, Using patent counts for cross-country comparisons of technology output, Special Issue on New Science and Technology Indicators, STI
Review, vol. 27, OECD, Paris, 2002, pp. 129146.
[23] M. Trajtenberg, A penny for your quotes: patent citations and the value of innovations, in: A. Jaffe, M. Trajtenberg (Eds.), Patents, Citations and Innovations - A
Window on the Knowledge Economy, MIT Press, Cambridge, MA., 2002, pp. 2549.
[24] D. Harhoff, F.M. Scherer, K. Vopel, Citations, family size, opposition and the value of patent rights, Res. Policy 32 (8) (2003) 13431363.
[25] H. Ernst, Patent portfolios for strategic R&D planning, J. Eng. Technol. Manag. 15 (4) (1998) 279308.
[26] Z. Griliches, Patent statistics as economic indicators: a survey, J. Econ. Lit. 28 (4) (1990) 16611707.
[27] X. Tong, J. Frame, Measuring national technological performance with patent claims data, Res. Policy 23 (2) (1994) 133141.

768

C. Choi, Y. Park / Technological Forecasting & Social Change 76 (2009) 754768

[28] F. Narin, E. Noma, Patents as indicators of corporate technological strength, Res. Policy 16 (2-4) (1987) 143155.
[29] M.E. Mogee, R.G. Kolar, International patent analysis as a tool for corporate technology analysis and planning, Technol. Anal. Strateg. Manag. 6 (4) (1994)
485503.
[30] M. Hirschey, V.J. Richardson, Are scientic indicators of patent quality useful to investors? J. Empir. Finance 11 (1) (2004) 91107.
[31] B. Yoon, C. Yoon, Y. Park, On the development and application of a self-organizing feature map-based patent map, R&D Manage. 32 (4) (2002) 291300.
[32] B. Yoon, Y. Park, A systematic approach for identifying technology opportunities: Keyword-based morphology analysis, Technol. Forecast. Soc. Change 72 (2)
(2004) 145160.
[33] B. Yoon, Y. Park, A text-mining-based patent network: analytical tool for high-technology trend, J. High Technol. Managem. Res. 15 (1) (2004) 3750.
[34] E.F. Levitas, M.A. McFadyen, D. Loree, Survival and the introduction of new technology: a patent analysis in the integrated circuit industry, J. Eng. Technol.
Manag. 23 (3) (2006) 182201.
[35] M.M.S. Karki, Patent citation analysis: a policy analysis tool, World Pat. Inf. 19 (4) (1997) 269272.
[36] A. Breitzman, P. Thomas, Using patent citation analysis to target/value M&A candidates, Res. Technol. Manag. 45 (5) (2002) 2836.
[37] M. Meyer, Tracing knowledge ows in innovation systems, Scientometrics 54 (2) (2002) 193212.
[38] D.M. Waguespack, J.K. Birnir, Foreignness and the diffusion of ideas, J. Eng. Technol. Manag. 22 (1/2) (2005) 3150.
[39] von Wartburg, T. Teichert, K. Rost, Inventive progress measured by multi-stage patent citation analysis, Res. Policy 34 (10) (2005) 15911607.
[40] A. Mina, R. Ramlogan, G. Tampubolon, J.S. Metcalfe, Mapping evolutionary trajectories: applications to the growth and transformation of medical knowledge,
Res. Policy 36 (5) (2007) 789806.
[41] N.P. Hummon, P. Doreian, Connectivity in a citation network: the development of DNA theory, Social Network 11 (1) (1989) 3963.
[42] B. Verspagen, Mapping technological trajectories as patent citation network: a study on the history of fuel cell research, Adv. Complex Systems 10 (1) (2007)
93115.
[43] S. Wasserman, K. Faust, Social Network Analysis: Methods and Applications, Cambridge University Press, Cambridge, UK, 1994.
Changwoo Choi is a doctoral candidate in the Department of Industrial Engineering of Seoul National University (SNU). He holds MS in industrial engineering from
SNU. His research interests lie in R&D policy, technology management, patent analysis, and new service creation. He has authored published papers in
Technological Forecasting and Social Change, and Expert Systems with Applications.

Yongtae Park is a professor in the Department of Industrial Engineering at SNU. He holds a BS in industrial engineering from SNU, and an MS and PhD in operations
management, both from the University of Wisconsin-Madison. His research topics cover a wide variety of areas including technology management, technological
innovation, knowledge management, and service engineering.