Next Article in Journal
A Model Approach for Yield-Zone-Specific Cost Estimation of Greenhouse Gas Mitigation by Nitrogen Fertilizer Reduction
Previous Article in Journal
Daily Monitoring of Shallow and Fine-Grained Water Patterns in Wet Grasslands Combining Aerial LiDAR Data and In Situ Piezometric Measurements
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

How Can Big Data Complement Expert Analysis? A Value Chain Case Study

1
Department of Electrical and Computer Engineering, Sungkyunkwan University, 164 Suwon, Korea
2
Department of Industrial Engineering, Ajou University, 164 Suwon, Korea
*
Author to whom correspondence should be addressed.
Sustainability 2018, 10(3), 709; https://doi.org/10.3390/su10030709
Submission received: 2 January 2018 / Revised: 14 February 2018 / Accepted: 23 February 2018 / Published: 6 March 2018

Abstract

:
In the world of big data, there is a need to investigate how data-driven approaches can support expert-based analyses during a technology planning process. To meet this goal, we examined opportunities and challenges for big data analytics in the social sciences, particularly with respect to value chain analysis. To accomplish this, we designed a value chain mapping experiment that aimed to compare the results of expert-based and data-based mappings. In the expert-based approach, we asked an industry expert to visually depict an industry value chain based on insights and collected data. We also reviewed a previously published value chain developed by a panel of industry experts during a national technology planning process. In the data-driven analysis, we used a massive number of business transaction records between companies under the assumption that the data would be useful in identifying relationships between items in a value chain. The case study results demonstrated that data-driven analysis can help researchers understand the current status of industry structures, enabling them to develop more realistic, although less flexible value chain maps. This approach is expected to provide more value when used in combination with other databases. It is important to note that significant effort is required to develop an elaborate analysis algorithm, and data preprocessing is essential for obtaining meaningful results, both of which make this approach challenging. Experts’ insights are still helpful for validating the analytic results in value chain mapping.

1. Introduction

Big data is a term that refers to an enormous volume of structured and unstructured digital data extracted from various sources [1]. According to IBM, approximately 2.5 quintillion bytes of data are generated every day [2]. Moreover, 90% of the data we have today were created in the past two years, indicating the tremendous recent growth in data collection and storage [3]. This growth has been observed for diverse entities, including individuals, businesses, and the private and public sectors. Effectively using these data is now regarded as a key component of successful management. The private sector uses big data to predict customer requirements and behaviors, develop core competencies, and create novel products and services. On the other hand, the public sector uses big data to increase transparency, facilitate civic participation in public life, prevent crimes, improve national security, and support social welfare through education and health care management [4,5,6,7,8]. Indeed, big data are being used in diverse disciplines for various purposes that add value to our society.
Data-driven efforts have also been made in the field of innovation management. One of the most representative approaches in this field is patent analysis, as patent documents contain a significant amount of technology development information and are easy to access [9]. Data-driven studies, whether the data be patents, customer complaints, or App store data, commonly show that big data can deliver insights that would otherwise be impossible to obtain. At the same time, however, because these data are derived from past records, data-driven analytic results may not be useful for predicting a future characterized by great uncertainties. That is, for very complex problems, experts’ insights may provide more helpful results than big data analysis.
To maximize the value of data-driven approaches in the world of big data, it is essential to understand their possibilities as well as limitations, compared to expert-based approaches. That is, it is necessary to investigate how a data-driven approach can support an expert-based analysis, and to assess whether such an approach can provide more knowledge than expert insights. However, efforts made to address this issue have been limited. To satisfy such a need, this study aimed to examine opportunities and challenges for big data analysis in the social sciences, particularly with respect to industry analysis, as an attempt to answer the following research question: how can big data complement expert analysis for technology analysis and strategic planning, supporting to navigate the increasingly complex industry dynamics? Here, it should be noted that we restricted our focus to social sciences because different opportunities and challenges for big data analytics can be expected in different areas. The social science was chosen as a research target; the social science phenomena are so complex that it is sometimes difficult to find meaningful implications only from big data analysis. Thus, on one hand, it may be good to rely on experts’ insights on understanding such phenomena. However, on the other hand, data analysis can offer unexpected surprising findings due to such complexity. Moreover, expert insights may have limitations in capturing all the rapidly changing social phenomena. The existence of both opportunities and challenges is evident in the social science and so identifying them is essential to overcome challenges and seize opportunities from big data analytics. Among various areas for experiments, we focused only on technology planning. Systematic decision-makings based on expert- or data-driven analyses are required for technology planning and thus opportunities are easy to be found.
To accomplish this, we designed a simple experiment for comparing the results of an expert-based approach and a data-driven approach. In our experiment, we focused on analyzing an industry value chain for a photovoltaic system in Korea. The value chain is a diagram that explains the chained linkage of activities, i.e., buyer-seller relationships, in the physical world. Both its concept and tool have been widely used in the last 30 years for analyzing industries [10,11]. It was initially developed to help understand the value creation activities at the firm level and then started to be applied at the industry level to model the activities at the aggregate level. Figure 1 presents an industry value chain for semiconductor. According to the figure, the semiconductor products, having two types of users—[end user] and [systems user], are produced by [assembly and test] followed by [fabrication]. Then, for fabrication, the components need to be designed [design] and the [equipment and materials] required for the fabrication need to be acquired. The value chain was proven to be a useful tool to portray a process of value creation in industries [12] and further understand the industry structure.
Analyzing a value chain is closely related to technological forecasting and planning. Technology roadmapping, which is a process of looking forward, requires a value chain diagram to make a valid decision. Moreover, developing a value chain diagram is regarded as a difficult job because it requires a comprehensive understanding of the industry. When it is developed by a single person, the diagram may depend on a restricted view. When more experts are involved in the process of creating the diagram, it would be difficult to obtain a consensus. For the expert-based approach, we asked an industry expert to visually represent an industry value chain based on insights and data collection. We also reviewed a previously published value chain developed by a panel of industry experts during a national technology planning process. For the data-driven analysis, we used a massive number of business transaction records between companies, which we expected to reveal a network of business interrelationships.
Our experimental results show that the data-driven analysis enabled even a non-industry expert to identify basic industry structures. The analysis occasionally proposed a value chain item that had not been discovered by our industry expert, where the item is defined as a product or service that belongs to the value chain; this was particularly common for emerging items. Moreover, the unit of the value chain items could be identified based on the actual business transaction, which is more realistic but less flexible as the value chain items are determined by the transaction data. Finally, we were able to identify core companies for each of the value chain items and promote a company-level analysis by examining the characteristics of such items. While a data-driven approach is advantageous in many respects, considerable effort is required to develop an elaborate analysis algorithm, and data preprocessing is essential for obtaining meaningful results. Experts’ insights are still helpful for validating the analytic results. This study is one of the earliest attempts to compare the data-driven approach with the expert-driven approach during industry analysis and technology planning, and to discuss how the two approaches can complement one another. It is important to take a balanced view on the use of data-driven approach, especially in the social sciences.
This paper consists of six sections. Section 2 provides a basic overview of big data analysis and value chain analysis. The overall research process is described in Section 3, and the analytic processes and results from data-driven and expert-driven value chains are explained in Section 4 and Section 5, respectively. The pros and cons of the two approaches are compared in Section 6 to show how a data-driven industry analysis approach can be complemented by an expert-driven approach. Finally, the contributions and limitations of this study are discussed in Section 7, along with future research directions.

2. Literature Review

This section reviews previous studies on big data analysis in technology planning and value chain analysis.

2.1. Big Data Analysis in Technology Planning

Big data refers to any data set that is extremely large and complex that it cannot be easily managed by current software and data management techniques [14]. According to [15], a big data set has three characteristics that make it more difficult to analyze than conventional data. The first is variety, that is, the complexity and diversity of the data. Big data are assembled from a wide range of sources and take three different forms: structured, semi-structured, and unstructured. Big data’s second main characteristic is volume, or the overall size of the data. The current volume of big data typically exceeds terabytes or even petabytes; such data volumes surpass modern data storage capabilities and analysis technologies. The final key characteristic of big data is velocity, that is, the speed at which data is generated. To maximize the value of big data, we must be able to process those data at an adequate speed.
Because a big data set is so voluminous and tends to rapidly change in complex ways, there are several analysis techniques essential for obtaining meaningful knowledge from it [3]. Numerous approaches, such as machine learning, text mining, data mining, and crowdsourcing, have been applied to big data analysis. Relevant tools for coping with data volume and complexity have developed rapidly over the past several years [16]. Now, senior managers across industries and regions wonder how to get full value from the massive amount of data they collect every day as well as the already have within their organizations [17]. Using the data, those managers need to know what actions should be taken to cope with the changes expected in the future. Despite this development, big data analysis is facing other challenges, including the integration of internal and external data sources, as well as privacy and security issues. For example, customer data can be collected for purposes that add value to customers. At the same time, however, companies that possess these data should be able to guarantee their customers that such data are securely protected and cannot be used for other purposes.
These opportunities and challenges are present in big data analysis used for technology planning, which commonly includes value chain analysis. Big data sources in this field include patents, trademarks, company profiles and transaction data, and social network service data, although most of existing studies have focused on the analysis of patent data. Thus, far, patent information has been considered as the most valuable source of technology intelligence [18]. Different types of patent data, such as bibliometric, descriptive, and citation information, have been analyzed for a broad range of purposes, including trend analysis [19], competitor analysis [19,20], and opportunity analysis [21,22,23,24]. Focusing on the semi-structured and unstructured patent data, which are gaining an increasing interest, a patent analysis system that enables to analyze such a large size of data in an efficient manner was proposed as well [25]. More recent attempts at big data analysis for technology planning have included the introduction of other available technology planning data. These databases include trademarks, where product information extracted from trademarks is analyzed and linked to patent information as a means for identifying new technology opportunities [26], online customer complaints, where customer requirements are identified and considered as part of a new product development [27], mobile service applications, where online user behavior patterns are observed in order to suggest new business opportunities [28,29,30], and other data collected from various channels (e.g., wireless sensor networks) to be used for business operations and risk management [31].
These analytic results can provide managers significant knowledge for use in technology planning, although data integration, privacy, and data security must also be considered. For example, an automotive technology mining method was proposed to identify prospective technologies in the automobile industries based on the demands of automakers [32]. Technology planning is a future-oriented process of decision-making, which is required to establish a vision of the future, explore various options to achieve such vision, and choose the best option among them. On one hand, the quality of decision-making may improve with the amount of available data. On the other hand, the fact that this process is future-oriented may decrease the value of a data-driven approach because of the uncertainties associated with the future. Technology planning is not only concerned with describing the current technology landscape, but also with predicting the technological future. Data-driven approaches often assume that past trends will continue in the future, which may not be true for technology planning and for other fields (e.g., pattern recognition in speech and language processing). It might be dangerous to make decisions based solely on data; experts’ insights should be used to complement findings derived from data.
In this study, we examined the possible use of business transaction data for developing a value chain that meets the three conditions of big data, namely variety, volume, and velocity because: (1) the transaction data include both structured (e.g., the amount of transaction in a monetary value) and unstructured data (e.g., company name, transaction items), meeting the requirement of variety; (2) the size of transaction data is large enough, meeting the requirement of volume; and (3) the transaction data are updated for every new transaction, meeting the requirement of velocity. Although the data set used for this study was limited to 17,068,062 due to the data accessibility, its size can be increased dramatically because the set contains transaction data as well as patent data, trademark data, and company profiles. Accordingly, each data record has numerous fields and so the data size is quite large for a data set with 17,068,062 records. In those data fields, unstructured data (e.g., patent title, patent abstract, product/service titles) as well as structured data are observed; a variety of data types are given in the database. Finally, the data set used in this study can change every moment according to the entry and exit of companies, corporate patenting and trademarking activities, and new business transactions between companies. If data are cumulated over time, data size can increase dramatically, enabling time-series analysis. Furthermore, we adopted only a subset of all Korean companies, but a larger number of companies can be used. Similarly, we considered patents published only in Korea, but patents published in other countries can be involved in the analysis. Hence, the data used for this analysis had limitations, but its basic attributes can meet the condition of big data.
Considering that big data analytics is the process of collecting, organizing, and analyzing a large set of data (big data) to discover useful information, the question on whether a set of transaction data can replace the work by experts in developing a value chain is worth investigating. Nevertheless, our analysis will be based only on a simple approach, as the purpose of this study is not to suggest a novel data-driven algorithm to construct a valid value chain but to show the potential of the existing databases as solutions to replace or support the analysis by experts. If a simple approach can even produce reasonable results, a greater impact can be expected from big data analytics, where more advanced approaches are used.

2.2. Value Chain Analysis

A value chain is the series of activities, functions, and processes that are directly and indirectly related to a product or service during the process of delivering value to a customer [33]. This concept was developed by Porter in 1985 as a strategic tool for internal corporate analysis. Porter claimed that value chain activities are the foundation of corporate competitive advantages, and that these activities are not just a collection of independent activities, but also a systematic structure of interrelated activities. From this perspective, he tried to explain the sources of cost and distinct advantages by investigating all activities in a company and their interrelationships.
More recently, value chain analysis has been applied to assess inter-enterprise competitiveness and promote inter-industry or international competitiveness [34]. The use of value chain analysis has expanded to industry analysis. Value chain analysis at the industry level allows for a comprehensive look at industries and can be used for strategic technology planning or industry policy-making. Accordingly, there have been previous attempts to identify an industry value chain. Yan and Wang analyzed the value flow of iron resources from China by using material flow and value chain analyses [34]. El-Sayed et al. analyzed the value chain of the aquafeed sector in Egypt, which includes feed input suppliers, aquafeed producers and marketers, and fish farmers [35]. A value chain analysis applied to the scrap tire reverse logistics chain (STRLC) was shown to verify whether STRLC can become sustainable from an environmental, economic, and social perspective [36]. Irvine demonstrated that value chain analysis provides a robust, systems-based approach that can be used as a systematic framework for evaluating a livestock health surveillance system [37]. He further demonstrated the context in which disease information and associated surveillance data are placed in a surveillance system value chain by the people, businesses, and organizations involved. Jussani et al. developed a global value chain for urban light electric vehicle in Korea and Japan, aiming to identify the main features of the value chain [38]. Analyzing an industry value chain is regarded as an essential step for technology planning, policy-making support, and strategic planning. At the firm level, the value chain analysis results allow a firm to comprehend its position in the industrial ecosystem and identify its direct and indirect competitors as well as its clients and suppliers. At the national (industry) level, the value chain analysis results can help in revealing a country’s strengths and weaknesses. As described above, both data-driven and expert-driven approaches have been adopted for analyzing an industry value chain.

3. Research Process

The overall research process consists of four steps, as shown in Figure 2. First, a target industry for analysis was identified. We selected a photovoltaic system as the target industry, but the same analysis was conducted on light-emitting diodes (LED) for the validity test (see Appendix A). Because of their extensive potential applications and increasing efficiency, photovoltaic systems are regarded as a promising power generation technology [39]. Both theoretical and practical contributions are expected from analyzing a photovoltaic system value chain.
Our second step was to develop an expert-driven value chain map for the photovoltaic system. Two approaches were adopted. The first was to search for a relevant value chain map that already existed in the public domain (having already been developed by a panel of industry experts). The second was to ask an industry expert to analyze a value chain specifically for this experiment. In this step, we also developed a value chain map by using a data-driven approach. Data collected from Korea Enterprise Data (KED) (http://www.kedkorea.com) were selected as the main source for analysis. KED is one of the primary companies that provide credit data on Korean businesses. It is an affiliate of the Korea Credit Guarantee Fund, a government subsidiary; therefore, its data are massive, reliable, and up-to-date. KED collects data ranging from company profiles to financial statements. This study used two of KED’s data sources: company profiles and business transaction data. In particular, KED maintains a self-reported data set of each company’s main clients and corresponding business items, which are its main product and service categories for transactions with those clients. If analyzed well, the combination of business transaction data and company profiles can be a strong foundation for value chain analysis. After merging the two data sets, we obtained 17,068,062 records. As part of data preprocessing, we performed data cleansing on those records. The basic steps for constructing a value chain are as follows: (1) the companies offering the end products of a value chain are identified; (2) supplier companies that have provided products or services to the companies identified in the first step are extracted from the transaction data; (3) the supplier companies identified in step 2 are grouped by standard industry classification (SIC) codes; (4) the group name is decided based on the SIC codes, main business items, and main transaction items of the group members; and (5) the same procedure is followed for other groups.
Finally, the value chain constructed using this data-driven approach is compared with the value chain created via the expert-driven approach. A discussion on the pros and cons of these approaches can help us determine the ways in which data-driven approaches can be used in combination with expert-driven approaches.

4. Expert-Driven Value Chain Development

This section explains the data collection process and the results for expert-drive value chain development.

4.1. Data Collection

An expert-based value chain was constructed in two ways. First, we obtained photovoltaic value chain maps that had previously been analyzed and published. Most of these maps were developed by a panel of experts involved in the Korean government’s long-term technology planning. We were able to obtain a value chain from an industrial technology roadmap report published in 2012 by the Korea Institute for Advancement of Technology (KIAT). The report is the outcome of an effort to develop special R&D strategies for helping industrial technologies to cope with industrial technology policies and environmental changes. During the roadmapping process, a value chain was investigated as a means for understanding the industry structure. Five to seven industry experts provided the relevant knowledge and background information required to construct the value chain. We were concerned that this previously existing, public-domain value chain might be outdated at the time of our study, and that it may have been affected by the underlying purpose of its research; that is, we are aware that different perspectives behind a value chain analysis may lead to different results. Therefore, we also asked an industry expert to construct an original photovoltaic value chain for our study. We recruited industry experts capable of overviewing the photovoltaic industry in June 2015. Three candidates were identified as appropriate experts, and one was selected as the most suitable. He was given two weeks to construct a value chain, and we received his results in July 2015.

4.2. Value-Chain Development Results

Figure 2 and Figure 3 present the results of the expert-driven value chain analysis. The value chain map in Figure 2 has a relatively simple structure, including four second-level and four third-level items. The unit of value chain items, such as the solar collector and solar absorber, is mostly the unit of business in the industry. Figure 4 shows the value chain map created for this study, which includes more specialized items than the value chain map shown in Figure 3. The second-level items were grouped by different technological principles, which were based on a more in-depth understanding of the industry and associated technologies.

5. Data-Driven Value Chain Development

This section describes the data collection and preprocessing process for data-driven value chain development, followed by the data analysis methods and results for its development.

5.1. Data Collection and Preprocessing

A company profile data set consisting of 3,107,069 records and a business transaction data set consisting of 17,068,062 records were collected from KED and were then merged into a single data set with 17,068,062 records. To do this, the company ID (business registration certificate) was used as a key field to link company profile data set with business transaction data set. A company name together with company address and representative name was used to link company profile data set to patents and trademarks. This integrated database, which was built on the basis of transaction data, is composed of three parts: (1) supplier data (identity code, registration number, company name, main products, and SIC code); (2) customer data (identity code, registration number, company name, main products, and SIC code); and (3) transaction data (transaction value in Korean won and transaction items). A data cleansing process was then undertaken to delete missing and erroneous data. Specifically, data for companies bankrupted or unrecognizable company names were eliminated, which corresponds to less than 1% of total data.

5.2. Data Analysis Methods and Results

First, we identified the end-product manufacturers of photovoltaic systems in Korea, as first-level companies. This was performed with a keyword search by using the term “photovoltaic system” from the main products field of the database used for analysis. Here, synonyms (e.g., solar power system for photovoltaic system), or terms for equivalent products (e.g., solar collector and solar panel for solar cell, or LCD for liquefied crystal display) need to be considered for the best use of the database. Thus, such were defined in advance for each product, although the terms in the value chain consist of the most representative terms. More specifically, we used Boolean searching to combine search terms using the three operators, AND, OR, and NOT.
As a result, 24 first-level companies were identified at the first (end-product) level. Then, companies having recorded transactions with the end-product manufacturers were identified from the database as second-level companies, and their SIC codes were extracted. The SIC codes have six alphanumeric characters; only the first three characters were used to define the items in the value chain. Then, to determine which SIC codes would be included in the value chain, a cut-off criterion of 5% was applied—if a particular SIC code appeared in more than 5% of second-level companies, then that SIC code was selected as a value chain element; however, the value for the cut-off criterion can be flexible. In theory, the best cut-off criterion will be that value which produces the most valid and least noisy companies. However, the best criterion is likely to be case-sensitive, and therefore, we recommend that such value should be determined by users and not by testing. If users want to investigate a large number of items including those with relatively lower significance, a small cut-off value is recommended. On the other hand, if users want to see only core items, a large cut-off value will be a better choice. We designated 220 companies as second-level companies, which were widely spread over 26 SIC codes. For these 220 companies, only seven SIC codes, each of which was observed in more than 11 second-level companies (5% of 220 companies), were included in the value chain as core elements (see Table 1).
The table above shows that industry titles barely reveal detailed information about the actual items in the value chain. Therefore, industry titles must be converted to item-level titles. For each group of companies with the same industry code, two data fields (main products and transaction items) were retrieved in order to name the representative business items for the group in the context of photovoltaic systems. If necessary, we visited the websites of companies to identify their main business items. When more than 70% of companies in a group are offering a particular item, we put the item on the value chain as a core element. For example, there are 49 companies with an SIC code of C28. Among these 49 companies, more than 36 are producing an “inverter.” As 36 is greater than 34.3, which corresponds to 70% of 49, the item inverter is determined to be a representative item for group C28 and a core item in the photovoltaic system value chain. As part of this process, we delete a group from the value chain if the group is concerned with too many items, and hence, does not have a generalizable name for its business items, e.g., C29 and G4; no item was shared by more than 70% of companies in each of these two groups. Additionally, we delete a group from the value chain if it is obvious that the business items for the group are not related to the value chain, e.g., C32; the industry title of this group is “manufacture of furniture,” where more than 70% of companies belonging to this group produce office furniture, desk, chair, and others. However, the purchase of these items is apparently not for manufacturing a solar system but for general use in the office.
After identifying all the second-level business items, the third-level business items were determined as presented in Table 2. The same procedures used to identify the second-level items were applied to the third-level items (see Appendix B for detailed information). The depth of analysis, that is, at which level the value chain map will be deployed, is determined according to the overall purpose of the analysis; in this study, it was set to three. The data-driven value chain analysis results are illustrated in Figure 5. This figure shows that a photovoltaic system industry consists of one first-level item, six second-level items, and seven third-level items.

6. Big Data Analytics vs. Expert Insights

The value chain results based on big data analytics and expert insights are compared in this section to identify the strengths and weaknesses of each approach and further to understand the opportunities of big data analytics.

6.1. Comparative Analysis of the Two Approaches

The experiment in this study demonstrates how big data analytics can contribute to technology planning. As the purpose of technology intelligence, which is the key role of both data and experts in technology planning, is to offer the right information to the right person at the right time, these three conditions were used as the three criteria to compare the characteristics of data-driven and expert-driven approaches. First, as to “right information”, the quality and reliability of the analysis results were considered. The data-driven analytic results show that the value chain was developed based on the terms and business items that are actually used in the real world. In addition, we were able to find some items that were not identified in the expert-driven analysis, including network infrastructure equipment, LEDs, semiconductor molding and automation equipment, transformers, monitoring systems, main junction boxes, and batteries (orange color in Figure 6). Our expert-driven analysis categorized “frames” as a third-level item, whereas it was a second-level item in the data-driven analysis (blue color in Figure 6). This discrepancy indicates that end-user product manufacturers in this industry have worked directly with frame suppliers, which more closely matches business realities. We presented the results of two approaches to two experts that have been involved in various industry value chain mapping and let them discuss the discrepancies between the two approaches. The experts argued that both approaches have their own pros and cons in terms of information accuracy. The data-driven approach is beneficial in situations where little information is available for developing an industry value chain and thus it can be a good baseline from which expert analysis could start, particularly for investigating emerging industries. Furthermore, there exist several items observed only in the data-driven analysis results. These missing items seemed to be regarded as of little importance by those who deployed the value chain and the perceived importance of item may vary depending on the criteria to be used for evaluating such importance; the data-driven approach can provide standard criteria to determine whether an item should be included in the value chain or not. At the same time, both of the experts concerned about the difficulties in collecting the right data for analysis. In our experiment of data-driven approach, the items imported could not be considered due to the lack of available data, which may distort the true global value chain of photovoltaic industry. Therefore, an expert-driven approach can still be useful to supplement the data-driven analysis results. Considering the shortcomings of both approaches, we constructed an ideal value chain by using the results from both approaches; this ideal value chain is shown in Figure 6.
Second, with respect to “right person,” the flexibility of analysis in terms of its granularity and approaches was evaluated. We found that the expert-based approach could provide a relatively flexible solution to intelligence users as the granularity, as well as the approaches to analysis, is easily customizable to analysis purposes. For example, in our experiment, the expert-driven analysis produced more technical and specialized results than the data-driven analysis. Both micro-level (identifying technical and specialized items) as well as macro-level (identifying business items) analyses were feasible on request with the expert-based approach, while the level of analysis was relatively fixed at macro-level with the data-driven analysis in our experiments. Furthermore, the expert-driven analysis results reflected information that could not be extracted from the data used in this study; the business transaction data used in this study only reflect transactions between Korean companies, offering a value chain within Korea rather than from a global perspective. Products that are primarily imported cannot be identified from this data set, even if they are essential to the value chain. Acquiring all databases required for analysis purposes is the foremost process of obtaining the right insights for the right person, which may decrease the flexibility of the data-driven approach.
Finally, concerning “right time”, the time and cost required to conduct each analysis were examined, because they affect the possibility of providing necessary information with a minimum delivery time. A data-driven approach enables an easy analysis and update of analysis results. It may cost a lot to acquire a database of interest and to develop an algorithm for its analysis. In our experiment, it took almost a year to work on the project. However, once they are developed, having up-to-date analysis results based on real-time data becomes feasible only with a small maintenance cost, which is expected to significantly increase the value of analysis. On the contrary, expert-based analysis may require less cost in its first application but the same amount of production costs will be incurred for the subsequent applications.

6.2. Strengths and Weaknesses of Big Data Analytics

Our experiments showed that big data analytics may have both strengths and weaknesses, compared to expert-based approaches. First, with respect to data collection, the most significant step in big data analytics is to collect the right data for analysis. The data need to be representative, accurate, and aligned with the analysis purpose, as the quality of the results is likely to be greatly affected by the quality of data. For example, the data set obtained from KED is undoubtedly useful for analyzing a value chain for domestic firms but is limited in its capacity for analyzing a global value chain (which would need to consider international business transactions). A precondition of benefiting from big data analytics is the ability to obtain a proper set of data. Because of these limitations, we should be cautious of relying too heavily on data-driven analysis results.
Second, with respect to data analysis, given the appropriate data, big data analytics can help uncover findings that would not have been revealed by experts. Furthermore, data sets can be kept up-to-date relatively easily, which increases the reliability of data analysis. Although we failed to build real-time and time-series data in this study, business transaction data are basically real-time and time-series by nature. As a result, it is possible to develop a system that can monitor the evolution of industry value chains and can detect early signals of changes. The challenge here is developing an algorithm sophisticated enough to construct a trustworthy value chain. Once such an algorithm is in place, a value chain for another industry can be developed relatively easily, requiring little or no additional effort. However, it is still extremely difficult to explain complex social phenomena with an analysis model and relevant algorithms. Even in our data-driven approach, some manual tasks were required to ensure the quality of analysis results. For example, a set of keywords on end-user products, which will affect the list of companies producing such products and further those at the second- and third-levels, need to be identified by the users. Identifying wrong keywords for end-user products may result in an incomplete or even wrong value chain. Similarly, a cut-off value needs to be determined by the users. As there is no one-size-fits-all solution for every value chain, the appropriate criteria rely on the users according to the purpose of analysis.
Furthermore, the title of the main product and service items needs to be determined manually as the information obtained from the data is the list of SIC codes and corresponding companies, not the title of the items. In addition, there were some industry sectors whose value chains could not be analyzed with the algorithm used in this study. The approach used in this study has an intrinsic limitation in that it is more applicable to companies producing only a few items. The transaction data for a company characterized by a small-quantity batch production may have included a significant amount of noisy data. Moreover, the data for a single company covered several value chains, and therefore, choosing a subset of data that corresponded only to a particular industry of interest was a challenge that we failed to overcome in this study.
Finally, with respect to data usage, the data used in this study could be linked to other data, opening up a number of new possibilities. The potential of a data set is not limited to the information that can be directly obtained from it, but also extends to the information that could be obtained by combining it with other databases. For example, transaction data can be merged with company profiles, patent data, trademark data, or international trade data. If these databases are linked via company codes, various other meaningful analyses can be designed to identify the characteristics of a value chain in terms of its strengths, weaknesses, degrees of innovation, etc. However, it is important to note that linking data in this way would require considerable effort and would raise relevant issues regarding data security.

6.3. Opportunities of Big Data Analytics

Analyzing a value chain has been regarded by experts as one of the most difficult tasks involved in technology planning for the following reasons. First, experts may face difficulties in presenting an industry overview, as they are likely to be an expert only in a particular area. Second, published reports on value chains are not always reliable or applicable—they might be outdated, or the value chain analysis results might be significantly influenced by the purpose of the analysis. Third, even after a value chain map has been successfully developed, continuous updating is necessary to maintain it and to ensure that it reflects the changes in industrial environments.
Big data analytics can help overcome the above difficulties. It enables users to grasp the overall structure of an industry, supporting strategic technology planning and policy-making. When multiple databases are combined, more insightful analyses become feasible. Table 3 summarizes the analytic results of value chain items in terms of the characteristics of the companies that supply each of the items. To prepare this information, three databases were combined: the KED database, Association of Science and Technology Information (ASTI) database, and Korea Intellectual Property Rights Information Service (KIPRIS) database. We collected business transaction data and company profiles from the KED database, company lists (as of 2015) from the ASTI database, and patent data for the past three years (as of 2015) from the KIPRIS database. Table 3 indicates that in a value chain of photovoltaic systems, most companies are working on wafer and frame items; it is highly likely that these two items have been the focus of Korean companies. The transformer item shows the greatest increase in sales between 2014 and 2015, though the size of the market (inferred from the average income and sales of relevant companies) is small. The most active innovation activities are observed in the photovoltaic system item, which is an end-user product with an average of 4.9 patent publications per company.
There seem to be enormous opportunities for big data analytics in technology planning because understanding of the current situation is essential to obtain successful planning results. Such knowledge can be gained from both human resources and real-time data generated by various actors that are directly or indirectly related to a company in its innovation network. A company can ensure that its plan stays on track to success by closely monitoring up-to-date trends and thus can cope with today’s complex industry dynamics. However, these opportunities lie more in big data’s capacity to support expert decision-making, and less in big data’s ability to solve future-oriented problems. Solving such problems requires a look-ahead strategy, envisioning the desired future. However, big data analysis usually focuses on forecasting (a prediction based on past data), while future-oriented problems require foresight (a target for the desired future) as well as forecasting. In other words, data analysis results can create a diagram that describes an industry’s value chain, but it is still an expert’s role to verify and finalize such diagram: (1) a different data set may produce different results; (2) the way to present outputs is less flexible when using a data-based approach; and (3) by using secondary data, which is common in big data analysis, analysis results may contain noise even after data preprocessing or may lack some contents. In our analysis, we use a data set only from domestic firms, which limited the view to a domestic rather than a global value chain.
The types of planning suitable for data-driven analysis include fact-based planning. In our case, the data-driven value chain analysis enabled us to understand the overall relationships between firms in the value chain. From the analysis, it is possible to identify the strengths and weaknesses of the value chain at the industry level, based on which an industrial policy can be designed in a way to improve the strengths while overcoming the weaknesses. At the firm level, by looking at what other competitors are doing, a firm can make a better decision on its strategies on future businesses. On the other hand, the types of industries where our approach can have a significant potential to contribute are newly emerging and/or converging industries. In these industries, experts’ knowledge of industry structure is so limited or scatter across so many sources that it will not be easy to develop a value chain based on experts’ knowledge. The data-driven approach could help identify hidden and unexpected relationships between firms in a value chain.

7. Conclusions

This study aims to compare data-driven and expert-driven analytics in the field of technology planning. To accomplish this, we first obtained a value chain mapping case study for the photovoltaic industry, which is an industry that might benefit greatly from a data-driven approach, and one that presents challenges to experts involved in the value chain mapping. We designed a simple experiment to compare a data-driven approach with an expert-driven approach. An appropriate database was identified, and analysis algorithms were developed. Finally, the pros and cons of each approach were investigated in order to draw conclusions regarding opportunities for applying big data analytics to technology planning. This study is one of the earliest attempts to compare the data-driven approach with the expert-driven approach and to discuss how the two approaches can complement one another.
Our case study results demonstrated that big data analytics has obvious advantages over expert insights in that (1) the value chain unit could appropriately reflect real business environments; (2) the analytic results could identify value chain elements that had not been identified by experts; (3) updating value chain maps was relatively easy, a feature that is expected to provide interested parties with an early signal for industry changes; and (4) several databases could be interconnected to produce meaningful implications. However, there are still several challenges: (1) high quality results could not be achieved without both high quality data and sophisticated algorithms for transforming those data into information; and (2) owing to data limitations, the flexibility of the analysis in terms of the contents of value chain items was restricted. That is, how to interpret an industry with respect to its value chain may vary by experts; accordingly, different value chains can be developed by different experts, allowing the creativity and flexibility of analysis, which is not feasible in a data-driven analysis. Data-driven approaches provide a greater opportunity to support expert decision-making, but cannot be a sole solution for technology planning.
Despite these significant findings, this study has several limitations. First, it was exploratory in nature and thus, was not based on a rigorous experimental design. For example, the analytic results (that is, the value of each approach) may change depending on the data we obtain and the algorithms we develop. Thus, in the future experiment, more than one industry expert needs to be involved to achieve reliable and valid results. In addition, we set the depth of value chain to three, because the two value chains developed by experts used for comparative analysis in this study have adopted a three-level structure. However, for a data-driven approach, it would be interesting to have a higher number, which is required as future research.
Second, additional experiments on other industries will be required—different industries may benefit to different degrees from data-driven approaches. The more complex and rapidly changing an industry structure is, the more necessity there is to use the data-driven approaches. Future research need to identify the characteristics of industry necessitating the data-driven approaches.
Third, the approaches to analyze the value chain data can be improved. Only a simple approach was adopted in this study because its purpose was not to design a novel approach to investigating big data for value chain analysis but to show the value and limitations of data-based approaches, compared to expert-based approaches. Nonetheless, the value of big data analytics can increase dramatically by adding more data sources and applying more advanced data-mining algorithms. For example, this study used only a national database owing to the data accessibility, whereas a combination of national and international databases (i.e., the use of overseas transactions) will improve the results significantly. Similarly, an algorithm can be elaborated. The SIC codes play an important role in identifying items of a value chain. However, a single company may have several business areas and correspondingly several SIC codes. In our database, each company has only one SIC code, which best describes its main business area. This information is provided by a company, updated every year, and thus, is relatively accurate. Nevertheless, an algorithm to consider companies in several areas is worth developing by assigning several SICs to one company.
Indeed, the final goal for data-driven approaches should be to develop an algorithm that can create the best value chain without expert intervention. This can be a challenge of data-driven approach; for a more truly data-driven approach, several parameters that were given by experts in this study should be determined by learning from the data; acquiring the right set of data to be trained is prerequisite for a more truly data-driven approach. Thanks to the rapid advances in data analysis techniques, the algorithm for developing a value chain is expected to be elaborated by applying more automatic approaches (e.g., abnormality detection) or conducting a bigger experiment; groups that are common to all businesses (e.g., furniture, or office supplies) are captured to be excluded from the main value chain. Future research will address these issues.

Acknowledgments

This work was supported by the Ajou University research fund. We appreciate KED and KISTI (Korea Institute of Science and Technology Information) for their support in our data collection.

Author Contributions

Kyungtae Kim and Sungjoo Lee conceived and designed the experiments; Kyungtae Kim analyzed the data; Sungjoo Lee contributed analysis tools and wrote the paper.

Conflicts of Interest

The authors declare no conflict of interest.

Appendix A. Value Chain Analysis for LED

Appendix A.1. Expert-Driven Value Chain Development

Similar to the case of photovoltaic systems, two types of expert-driven value chain maps were acquired for LED: one from the “2012 knowledge map on green technology” published by the Ministry of Education, Science and Technology (MEST) in Korea to support small- and medium-sized enterprises (see Figure A1), and the other map from an expert who was requested to construct it (see Figure A2).
Figure A1. Expert-driven value chain analysis—public domain (MEST, 2012).
Figure A1. Expert-driven value chain analysis—public domain (MEST, 2012).
Sustainability 10 00709 g0a1
Figure A2. Expert-driven value chain analysis—private request.
Figure A2. Expert-driven value chain analysis—private request.
Sustainability 10 00709 g0a2

Appendix A.2. Data-Driven Value Chain Development

Setting two representative LED applications, namely LED lighting and Back Light Unit, as first-level items, the companies producing these two applications were identified. Then the companies selling products or services to the first-level companies were retrieved to define the second-level items. A total of 12 SIC codes (3 alphanumeric characters) were found as indicated in Table A1. The cut-off value was 7.5 (5% of total frequency of 150); thus, six codes (C26, C29, C23, C30, C27, and J58) were regarded as key items, among which, the main product and service items could be identified as “Package” only for two codes (C26 with a ratio of 82% and C29 with a ratio of 77%).
Table A1. Value chain analysis results: second-level SIC codes for LED.
Table A1. Value chain analysis results: second-level SIC codes for LED.
SIC CodesIndustry TitleFrequency
C26Manufacture of electronic components, computer, radio, television and communication equipment and apparatuses30
C29Manufacture of other machinery and equipment27
C23Manufacture of non-metallic mineral products25
C30Manufacture of motor vehicles, trailers, and semitrailers18
C27Manufacture of medical, precision, and optical instruments, watches and clocks13
J58Publishing activities12
C24Manufacture of basic metal products7
C22Manufacture of rubber and plastic products6
J62Computer programming, consultancy, and related activities5
C33Other manufacturing3
C20Manufacture of chemicals and chemical products except pharmaceuticals and medicinal chemicals2
F41General construction2
In the same way, the third-level SIC codes were extracted (see Table A2). The top seven items (C26, C28, C46, C30, C20, C27, and C29) had frequencies greater than the cut-off value, while Epitaxial wafer, LED chip, and Phosphor were found to be the main items produced by the third-level companies.
Table A2. Value chain analysis results: third-level SIC codes for LED.
Table A2. Value chain analysis results: third-level SIC codes for LED.
SIC CodesIndustry TitleFrequency
C26Manufacture of electronic components, computer, radio, television and communication equipment and apparatuses35
C28Manufacture of electrical equipment31
G46Wholesale trade and commission trade, except of motor vehicles and motorcycles26
C30Manufacture of Motor Vehicles, Trailers and Semitrailers25
C20Manufacture of chemicals and chemical products except pharmaceuticals and medicinal chemicals25
C27Manufacture of Medical, Precision and Optical Instruments, Watches and Clocks18
C29Manufacture of Other Machinery and Equipment15
C24Manufacture of Basic Metal Products10
C25Manufacture of Fabricated Metal Products, Except Machinery and Furniture8
C22Manufacture of Rubber and Plastic Products4
M72Architectural, Engineering and Other Scientific Technical Services4
G45Sales of automobiles and parts3
C18Printing and Reproduction of Recorded Media3
C32Manufacture of Furniture2
C33Manufacture of other products2
C31Manufacture of other transportation equipment2
The analysis results at the fourth level indicated that nine codes (C26, C20, C22, C24, G46, C25, C23, J62, and C29) were found to be the main ones (see Table A3) corresponding to six items including Sapphire, Silicon Carbide (SiC) substrate, Gallium nitride (GaN) ingot, Aluminum nitride (AIN) ingot, Wafer mounter, Indium gallium nitride (InGan) ingot, and Dispenser machine.
The final value chain map based on the data-driven approach is presented in Figure A3.
Table A3. Value chain analysis results: fourth-level SIC codes for LED.
Table A3. Value chain analysis results: fourth-level SIC codes for LED.
SIC CodesIndustry Title (Three Digits)Frequency
C26Manufacture of electronic components, computer, radio, television and communication equipment and apparatuses45
C20Manufacture of chemicals and chemical products except pharmaceuticals and medicinal chemicals37
C22Manufacture of rubber and plastic products30
C24Manufacture of basic metal products30
G46Wholesale trade and commission trade, except of motor vehicles and motorcycles26
C25Manufacture of fabricated metal products, except machinery and furniture25
C23Manufacture of other non-metallic mineral products24
J62Computer programming, consultancy, and related activities20
C29Manufacture of other machinery and equipment18
F42Special trade construction10
C28Manufacture of electrical equipment5
C25Manufacture of fabricated metal products, except machinery and furniture5
G45Sales of automobiles and parts4
C15Manufacture of leather, bags, and footwear3
C32Manufacture of furniture3
Figure A3. Data-driven value chain analysis.
Figure A3. Data-driven value chain analysis.
Sustainability 10 00709 g0a3

Appendix A.3. Big Data Analytics vs. Expert Insights

The items that were not found from the expert-based approach were generated from the data-driven approach, such as Dispenser machine, Wafer mounter, InGan ingot, and Phosphor. The data-driven approach seemed to reflect the real business transactions but still required some manual works. On the other hand, the expert-driven approach is more customizable and less time-consuming, similar to the observations from the case of photovoltaic systems.
Figure A4. Integrated value chain analysis.
Figure A4. Integrated value chain analysis.
Sustainability 10 00709 g0a4

Appendix B

Table A4. Value chain analysis result: third-level SIC codes.
Table A4. Value chain analysis result: third-level SIC codes.
SIC CodesIndustry TitleFrequency
G46Wholesale trade and commission trade, except of motor vehicles and motorcycles51
F42Special trade construction21
C26Manufacture of electronic components, computer, radio, television and communication equipment and apparatuses15
C22Manufacture of rubber and plastic products14
C28Manufacture of electrical equipment13
G47Retail trade, except motor vehicles and motorcycles13
C29Manufacture of other machinery and equipment12
F41General construction11
C25Manufacture of fabricated metal products, except machinery and furniture11
C27Manufacture of medical, precision, and optical Instruments, watches and clocks5
M72Architectural, engineering, and other scientific technical services5
C20Manufacture of chemicals and chemical products except pharmaceuticals and medicinal chemicals4
C23Manufacture of other non-metallic mineral products4
E38Waste collection, disposal, and materials recovery4
C10Manufacture of food products3
C14Manufacture of wearing apparel, clothing accessories and fur articles3
C16Manufacture of wood and of products of wood and cork; except furniture3
C31Manufacture of other transport equipment3
J58Publishing activities3
C13Manufacture of textiles, except apparel2
C30Manufacture of motor vehicles, trailers, and semitrailers2
C33Other manufacturing2
A02Forestry1
C18Printing and reproduction of recorded media1
C24Manufacture of basic metal products1
D35Electricity, gas, steam, and air conditioning supply1
N74Business facilities management and landscape services1
Table A5. Value chain analysis result: third-level items.
Table A5. Value chain analysis result: third-level items.
SIC CodesIndustry TitleMain Product and Service ItemsRatio
G46Wholesale trade and commission trade, except of motor vehicles and motorcyclesVarious items including monitor, printer, furniture, and others-
F42Special trade constructionVarious service items including electric work, painting, and others-
C26Manufacture of electronic components, computer, radio, television and communication equipment and apparatusesTransformer82
LED79
C22Manufacture of rubber and plastic productsBack sheet81
C28Manufacture of electrical equipmentSemiconductor/automation equipment70
C25Manufacture of fabricated metal products, except machinery and furnitureWafer78
Ribbon71
G47Retail trade, except motor vehicles and motorcyclesVarious items including pen, stapler, portable storage device, and others-
C29Manufacture of other machinery and equipmentNetwork infrastructure equipment76
F41General constructionConstruction service-

References

  1. Provost, F.; Fawcett, T. Data science and its relationship to big data and data-driven decision making. Big Data 2013, 1, 51–59. [Google Scholar] [CrossRef] [PubMed]
  2. Data, I.B. IBM-Bringing Big Data to the Enterprise. 2015. Available online: http://www-01.ibm.com/software/data/bigdata (accessed on 26 February 2018).
  3. Maltby, D. Big data analytics. In Proceedings of the 74th Annual Meeting of the Association for Information Science and Technology (ASIST), New Orleans, LA, USA, 9–12 October 2011; pp. 1–6. [Google Scholar]
  4. Cardenas, A.A.; Manadhata, P.K.; Rajan, S.P. Big data analytics for security. IEEE Secur. Priv. 2013, 11, 74–76. [Google Scholar] [CrossRef]
  5. Liu, W.K.; Yen, C.C. Optimizing Bus Passenger Complaint Service through Big Data Analysis: Systematized Analysis for Improved Public Sector Management. Sustainability 2016, 8, 1319. [Google Scholar] [CrossRef]
  6. Chui, K.T.; Alhalabi, W.; Pang, S.S.H.; Pablos, P.O.D.; Liu, R.W.; Zhao, M. Disease Diagnosis in Smart Healthcare: Innovation, Technologies and Applications. Sustainability 2017, 9, 2309. [Google Scholar] [CrossRef]
  7. Kim, G.H.; Trimi, S.; Chung, J.H. Big-data applications in the government sector. Commun. ACM 2014, 57, 78–85. [Google Scholar] [CrossRef]
  8. Jee, K.; Kim, G.H. Potentiality of big data in the medical sector: Focus on how to reshape the healthcare system. Healthc. Inform. Res. 2013, 19, 79–85. [Google Scholar] [CrossRef] [PubMed]
  9. Ernst, H. Patent information for strategic technology management. World Patent Inf. 2003, 25, 233–242. [Google Scholar] [CrossRef]
  10. Porter, M.E. Competitive Strategy: Techniques for Analyzing Industries and Competitors; Simon and Schuster: New York, NY, USA, 2008. [Google Scholar]
  11. Porter, M.E. Competitive Advantage: Creating and Sustaining Superior Performance; Free Press: New York, NY, USA, 1985. [Google Scholar]
  12. Peppard, J.; Rylander, A. From value chain to value network: Insights for mobile operators. Eur. Manag. J. 2006, 24, 128–141. [Google Scholar] [CrossRef] [Green Version]
  13. Mowery, D.C.; Macher, J.T. E-Business and the Semiconductor Industry Value Chain: Implications for Vertical Specialization and Integrated Semiconductor Manufacturers; East West Center Working Paper: Economic Series; East-West Center: Honolulu, HI, USA, 2001; Volume 17. [Google Scholar]
  14. Fan, W.; Bifet, A. Mining big data: Current status, and forecast to the future. ACM sIGKDD Explor. Newsl. 2013, 14, 1–5. [Google Scholar] [CrossRef]
  15. Sagiroglu, S.; Sinanc, D. Big data: A review. In Proceedings of the 2013 International Conference on Collaboration Technologies and Systems (CTS), San Diego, CA, USA, 20–24 May 2013; pp. 42–47. [Google Scholar]
  16. Zakir, J.; Seymour, T.; Berg, K. Big data analytics. Issues Inform. Syst. 2015, 16, 81–90. [Google Scholar]
  17. LaValle, S.; Lesser, E.; Shockley, R.; Hopkins, M.S.; Kruschwitz, N. Big data, analytics and the path from insights to value. MIT Sloan Manag. Rev. 2011, 52, 21. [Google Scholar]
  18. Kim, J.; Lee, S. Patent databases for innovation studies: A comparative analysis of USPTO, EPO, JPO and KIPO. Technol. Forecast. Soc. Chang. 2015, 92, 332–345. [Google Scholar] [CrossRef]
  19. Daim, T.U.; Rueda, G.; Martin, H.; Gerdsri, P. Forecasting emerging technologies: Use of bibliometrics and patent analysis. Technol. Forecast. Soc. Chang. 2006, 73, 981–1012. [Google Scholar] [CrossRef]
  20. Lee, S.; Yoon, B.; Lee, C.; Park, J. Business planning based on technological capabilities: Patent analysis for technology-driven roadmapping. Technol. Forecast. Soc. Chang. 2009, 76, 769–786. [Google Scholar] [CrossRef]
  21. Lee, S.; Lee, S.; Seol, H.; Park, Y. Using patent information for designing new product and technology: Keyword based technology roadmapping. R&D Manag. 2008, 38, 169–188. [Google Scholar]
  22. Lee, S.; Yoon, B.; Park, Y. An approach to discovering new technology opportunities: Keyword-based patent map approach. Technovation 2009, 29, 481–497. [Google Scholar] [CrossRef]
  23. Park, Y.; Lee, S.; Lee, S. Patent analysis for promoting technology transfer in multi-technology industries: The Korean aerospace industry case. J. Technol. Transf. 2012, 37, 355–374. [Google Scholar] [CrossRef]
  24. Seol, H.; Lee, S.; Kim, C. Identifying new business areas using patent information: A DEA and text mining approach. Expert Syst. Appl. 2011, 38, 2933–2941. [Google Scholar] [CrossRef]
  25. Shin, J.; Lee, S.; Wang, T. Semantic Patent Analysis System Based on Big Data. In Proceedings of the 2017 IEEE 11th International Conference on Semantic Computing (ICSC), San Diego, CA, USA, 30 January–1 February 2017; pp. 284–285. [Google Scholar]
  26. Lee, M.; Lee, S. Identifying new business opportunities from competitor intelligence: An integrated use of patent and trademark databases. Technol. Forecast. Soc. Chang. 2017, 119, 170–183. [Google Scholar] [CrossRef]
  27. Park, Y.; Lee, S. How to design and utilize online customer center to support new product concept generation. Expert Syst. Appl. 2011, 38, 10638–10647. [Google Scholar] [CrossRef]
  28. Kim, J.; Lee, S.; Geum, Y.; Park, Y. Patterns of innovation in digital content services: The case of App Store applications. Innovation 2012, 14, 540–556. [Google Scholar] [CrossRef]
  29. Kim, J.; Lee, S.; Park, Y. User-centric service map for identifying new service opportunities from potential needs: A case of App Store applications. Creat. Innov. Manag. 2013, 22, 241–264. [Google Scholar]
  30. Kim, J.; Lee, S.; Park, Y. A visual context-based market analysis of mobile application services. Manag. Decis. 2016, 54, 2106–2132. [Google Scholar] [CrossRef]
  31. Choi, T.M.; Chan, H.K.; Yue, X. Recent development in big data analytics for business operations and risk management. IEEE Trans. Cybern. 2017, 47, 81–92. [Google Scholar] [CrossRef] [PubMed]
  32. Liu, G.R.; Mao, L.X.; Wang, L.L.; Zou, S. Automotive prospective technology mining method based on big data content analysis. In Proceedings of the 2017 IEEE 2nd International Conference on Cloud Computing and Big Data Analysis (ICCCBDA), Chengdu, China, 28–30 April 2017; pp. 171–174. [Google Scholar]
  33. Sturgeon, T.J. How do we define value chains and production networks? IDS Bull. 2001, 32, 9–18. [Google Scholar] [CrossRef] [Green Version]
  34. Yan, L.; Wang, A. Based on material flow analysis: Value chain analysis of China iron resources. Renew. Sustain. Energy Rev. 2014, 91, 52–61. [Google Scholar] [CrossRef]
  35. El-Sayed, A.F.M.; Dickson, M.W.; El-Naggar, G.O. Value chain analysis of the aquaculture feed sector in Egypt. Aquaculture 2015, 437, 92–101. [Google Scholar]
  36. De Souza, C.D.R.; Márcio de Almeida, D.A. Value chain analysis applied to the scrap tire reverse logistics chain: An applied study of co-processing in the cement industry. Renew. Sustain. Energy Rev. 2013, 78, 15–25. [Google Scholar] [CrossRef]
  37. Irvine, R.M. A conceptual study of value chain analysis as a tool for assessing a veterinary surveillance system for poultry in Great Britain. Agric. Syst. 2015, 135, 143–158. [Google Scholar] [CrossRef]
  38. Jussani, A.C.; Wright, J.T.C.; Ibusuki, U. Battery global value chain and its technological challenges for electric vehicle mobility. Rev. Adm. Inov. 2017, 14, 333–338. [Google Scholar]
  39. Devabhaktuni, V.; Alam, M.; Depuru, S.S.S.R.; Green, R.C.; Nims, D.; Near, C. Solar energy: Trends and enabling technologies. Renew. Sustain. Energy Rev. 2013, 19, 555–564. [Google Scholar] [CrossRef]
Figure 1. Semiconductor industry value chain [13].
Figure 1. Semiconductor industry value chain [13].
Sustainability 10 00709 g001
Figure 2. Overall research process.
Figure 2. Overall research process.
Sustainability 10 00709 g002
Figure 3. Expert-driven value chain analysis—public domain (KIAT, 2012).
Figure 3. Expert-driven value chain analysis—public domain (KIAT, 2012).
Sustainability 10 00709 g003
Figure 4. Expert-driven value chain analysis—private request.
Figure 4. Expert-driven value chain analysis—private request.
Sustainability 10 00709 g004
Figure 5. Data-driven value chain analysis.
Figure 5. Data-driven value chain analysis.
Sustainability 10 00709 g005
Figure 6. Integrated value chain analysis.
Figure 6. Integrated value chain analysis.
Sustainability 10 00709 g006
Table 1. Value chain analysis results: second-level SIC codes.
Table 1. Value chain analysis results: second-level SIC codes.
SIC CodesIndustry Title Frequency
C28Manufacture of electrical equipment49
G46Wholesale trade and commission trade, except of motor vehicles and motorcycles35
F42Special trade construction23
C29Manufacture of other machinery and equipment21
C26Manufacture of electronic components, computer, radio, television and communication equipment and apparatuses16
C25Manufacture of fabricated metal products, except machinery and furniture14
C32Manufacture of furniture12
C16Manufacture of wood and of products of wood and cork; except furniture6
C27Manufacture of medical, precision, and optical instruments, watches and clocks6
C10Manufacture of food products4
C18Printing and reproduction of recorded media4
C23Manufacture of other non-metallic mineral products4
C30Manufacture of motor vehicles, trailers, and semitrailers4
J58Publishing activities4
C20Manufacture of chemicals and chemical products; except pharmaceuticals and medicinal chemicals2
C22Manufacture of rubber and plastic products2
C24Manufacture of basic metal products2
C33Other manufacturing2
M72Architectural, engineering, and other scientific technical services2
C13Manufacture of textiles; except apparel2
C14Manufacture of wearing apparel, clothing accessories and fur articles1
C17Manufacture of pulp, paper, and paper products1
F41General construction1
G47Retail trade; except motor vehicles and motorcycles1
J62Computer programming, consultancy, and related activities1
M71Professional services1
Note 1. The texts in bold indicate the items to be included in the value chain. Note 2. The frequency indicates the number of firms for each SIC code.
Table 2. Value chain analysis results: third-level items.
Table 2. Value chain analysis results: third-level items.
SIC CodesIndustry TitleMain Product and Service ItemsPercentage
C28Manufacture of electrical equipmentInverter85%
Battery80%
Main junction box72%
G46Wholesale trade and commission trade, except of motor vehicles and motorcyclesVarious items including thermometer, flooring, LCD, livestock, and iron-
F42Special trade constructionMonitoring system76%
C29Manufacture of other machinery and equipmentVarious items including environmental pollution measuring devices, noise blanker, and shopping cart-
C26Manufacture of electronic components, computer, radio, television and communication equipment and apparatusesSolar cell (monocrystalline/polycrystalline)89%
C25Manufacture of fabricated metal products, except machinery and furnitureFrame74%
C32Manufacture of furnitureOffice furniture, desk, chair, and others-
Table 3. Characteristics of companies in a value chain by items.
Table 3. Characteristics of companies in a value chain by items.
LevelValue Chain ItemsNumber of CompaniesAverage Assets
(2015)
(Million won)
Average Sales
(2015)
(Million won)
Average Income
(2015)
(Million won)
Average Sales Increase
(2014–2015)
(%)
Average Number of Patents *
(29 August 2013–29 August 2016)
3Network infrastructure equipment1846532586127−35.330.0
LED23159,042143,8084737−5.682.0
Semiconductor molding and automation equipment2435,41226,94729889.720.0
Transformer37785612,64614142.330.0
Wafer54255,56684,456134,108−43.661.5
Back sheet47332,307195,17719,292−6.672.5
Ribbon4369,21851,0271341−7.961.0
2Monitoring system343143530923028.762.0
Power converter (inverter)4292,749141,6532239−11.113.0
Solar cell (monocrystalline/polycrystalline)48194,64353,908−95730.491.7
Main junction box3250485439337−32.773.7
Frame57286,102350,31814,8822.640.0
Battery48504,945483,29940,060−4.413.0
1Photovoltaic system2433,85745,262699−8.374.9
Note * Average number of patents granted.

Share and Cite

MDPI and ACS Style

Kim, K.; Lee, S. How Can Big Data Complement Expert Analysis? A Value Chain Case Study. Sustainability 2018, 10, 709. https://doi.org/10.3390/su10030709

AMA Style

Kim K, Lee S. How Can Big Data Complement Expert Analysis? A Value Chain Case Study. Sustainability. 2018; 10(3):709. https://doi.org/10.3390/su10030709

Chicago/Turabian Style

Kim, Kyungtae, and Sungjoo Lee. 2018. "How Can Big Data Complement Expert Analysis? A Value Chain Case Study" Sustainability 10, no. 3: 709. https://doi.org/10.3390/su10030709

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop