Construction of data governance indicator system for manufacturing enterprises based on EFA and CFA

Abstract

A scientific and reasonable indicator system is the key to measuring the level of data governance in manufacturing enterprises (DGME). The study analyzes documents and policy analysis to classify manufacturing enterprise data governance into five dimensions and constructs a manufacturing enterprise data governance indicator system with 5 primary indicators and 23 secondary indicators. Study 1 utilized Exploratory Factor Analysis (EFA) analysis to determine that the data governance indicator system of manufacturing companies presents a five-factor structure; Study 2 used Confirmatory Factor Analysis (CFA) and Average Variance Extracted (AVE) to determine that the overall fitness of the model and the reliability and validity of each level of the indicator measurement items were at the desired level. The proposed indicator system can provide a reference for data governance in manufacturing enterprises.

Keywords

DGME indicator system EFA CFA AVE DAM

Introduction

As the main body of the national economy, the manufacturing industry is the foundation of the country, the tool for the development of the country, and the basis for the strengthening of the country. In order to effectively respond to the demand for massive multi-source heterogeneous data management¹ and value mining in the context of the big data era, manufacturing enterprises need to manage their massive data assets by means of data governance,² that is, to effectively control the whole process of data generation, processing, transmission, use, and destruction, and to solve the problems and needs faced in the process of releasing the value of data. Making data available and accessible in a systematic way. With the development of the times, data governance gradually has obvious industry characteristics, and relevant research around data governance in manufacturing enterprises³ has gradually become an important field. By combing and segmenting the data governance literature and research directions in manufacturing companies, it is concluded as follows:

Research on the concept of data governance “Information Technology Service Governance Part 5: Data Governance Specification” (hereinafter referred to as the “Data Governance Specification”) defines data governance as the collection of control activities, performance, and risk management associated with data resources and their applications. Janssen et al.⁴ define data governance as a model of an organization and its personnel, applications, control rules, and permissions to guide the normal operation of the entire life cycle of data and algorithms within and across organizations and ensure its institutional system. Bozkurt et al. consider data governance as the integration of strategies for data-related structures and entities with data processes, participants, architecture, and overall data management.⁵ Mercuri and Emerson⁶ argue that data governance is about who should have access, how the data will be used, how the products generated by the data will be shared, and how to ensure that the privacy of the individuals to whom the data relates is respected. Data governance and information governance are two relatively similar concepts. Many scholars even interchange these two concepts. Information governance is a broader definition based on the research perspective of information technology issues.⁷

Research on how to conduct data governance. Wei et al.⁸ established a manufacturing supply chain data governance platform using secure and reliable block-chain technology. Zorrilla and Yebenes⁹ constructed framework models for industrial data governance practices. Janssen et al.⁴ established a data governance framework for trusted Big Data Algorithmic Systems (BDAS) to enable external review, trusted information sharing within and between organizations, risk-based governance, system-level control, and data control through shared ownership and autonomous identity. Mao et al.¹⁰ constructed a new government data governance framework based on the concept of data intermediary platform to understand the detailed requirements and functions of a government data governance framework. To the dramatic increase in data volume and the accelerated pace of data assetization, Zhang et al.¹¹ developed a framework to explain how organizations configure data governance activities and take relevant strategic actions through an in-depth case study of a Chinese gold mining company. Li et al.³ proposed a blockchain-based “value-standard-process” collaborative framework for data governance in manufacturing organizations, which helps to ensure a high level of data security, high reliability of collaborative tasks, and high transparency of value transformation.

Research on the role of data governance. Abueed and Aga¹² applied PLS-SEM¹³ to model how corporate data governance contributes to sustainable knowledge creation in companies listed on the Amman Stock Exchange using survey data from (n = 180) listed companies and judgement sampling techniques. Both Yin and Li¹⁴ explored the relationship between corporate data governance and green technology innovation performance by drawing on knowledge management theory and dynamic capability theory. Gegenhuber et al.¹⁵ proposed a three-dimensional distributed Open Social Innovation (OSI) data governance framework consisting of openness, accountability, and power and applied it to reflect on OSI projects.

Research on building a data governance indicator system. Abraham et al.⁷ used a literature analysis of 145 research papers and practitioner publications published between 2001 and 2019 to identify the main building blocks of data governance and decompose them in six dimensions. Five research areas were also identified and a total of 15 research questions were formulated to support future research on data governance. Janssen et al.⁴ state that data governance indicators include assessment of data quality and bias, data sharing, data separation, distributed data storage, data collection, and usefulness of data. These indicators provide a comprehensive framework for building data governance for trusted AI systems. A systematic literature review of 191 articles on digital servitization was conducted by Smania et al.¹⁶ The findings provide six dimensions of data relationships in the digital servitization ecosystem: data strategy, data partner management, data reliability, data security and privacy, data interoperability, and data lifecycle. These dimensions cover key areas of data governance and provide a comprehensive data management framework for organizations in the digital servitization ecosystem.

Through the above literature analysis, it is found that few scholars study data governance in terms of the overall application scenarios of manufacturing enterprises, and the research on the data governance index system of manufacturing enterprises still remains in the conceptual discussion stage. Manufacturing enterprises in the process of data governance implementation, there are problems such as insufficient awareness of data governance, fuzzy concepts, unclear boundaries, unclear perceptions, and inconsistent standards, which are serious impediments to data management and application. In order to effectively measure the level of data governance development and industrial maturity of manufacturing enterprises, and to grasp the current status and problems of data governance development in manufacturing enterprises in a timely manner, there is an urgent need for a set of indicator system to transform data governance from conceptual to operable variables. Although some scholars have proposed a conceptual model of data governance, the lack of a systematic scientific argumentation shows the problem of insufficiency. In fact, quantitative measurement is an indispensable part of the process, and purely qualitative methods can only be used for theoretical construction but not for theoretical conceptual validation, and need to be combined with the use of quantitative research and analysis in order to help test the conceptual validity of the theory. Therefore, this paper focuses on manufacturing enterprise data governance research results and policy documents to build a manufacturing enterprise data governance indicator system consisting of 5 level-1 indicators and 23 level 2-indicators, through EFA and CFA to clarify the dimensions of manufacturing enterprise data governance and the specific measurement indexes, to provide reference for the evaluation and enhancement of data governance of manufacturing enterprises, and to help manufacturing industry transformation and upgrading.

Predefined indicator system for data governance in manufacturing enterprises

Data governance needs to be analyzed in detail based on the specific governance subjects, governance objects, governance goals, and cultural environment. Due to differences in industry background, research motivations, focus objects, and target expectations, the data governance indicator systems proposed by different scholars or institutions are different. Each has its own advantages and disadvantages and is not universal, but it can still provide an important reference for research on data governance in manufacturing enterprises. This paper is based on the academic results of “data governance,” based on the content of the “Data Governance Specification,” “Data Governance Standardization White Paper,” and other documents, combined with the actual needs of data governance in manufacturing enterprises, the basic situation of data governance in manufacturing enterprises is divided into five dimensions, including data strategy, data asset management, data infrastructure, organizational guarantee, and institutional protection. In particular:

Data strategy (DS)

DS¹⁷ refers to determining the goals and direction of enterprise data governance in terms of systems and processes, clarifying the task blueprint, and evaluating the implementation status and implementation results of the strategy. There are four secondary indicators under this dimension, namely, strategic objectives, business process management,¹⁸ strategy implementation, and monitoring and evaluation.

Data asset management (DAM)

DAM¹⁹ refers to planning, controlling, and providing data as an important enterprise asset, thereby controlling, protecting, delivering, and improving the value of data assets. There are seven secondary indicators under this dimension, namely, metadata,²⁰ master data,²¹ data quality management,²² data application,²³ data sharing,²⁴ data security,²⁵ and data life cycle.²⁶

Data infrastructure (DI)

DI²⁷ is the technical support for enterprise data governance. DI refers to the platforms, tools, and software systems used to standardize the entire life cycle of data, such as data aggregation, storage, development, application, operation and maintenance, and security. There are four secondary indicators under this dimension, namely, database, big data platform,²⁸ data analysis and mining,²⁹ and data integration and fusion.³⁰

Organizational guarantee (OG)

According to the stakeholder theory,³¹ manufacturing enterprise data governance involves the participation of multiple entities such as enterprises, industries, and governments. In the process of data governance, attention should be paid to focusing on the responsible parties and their rights and responsibilities, allocation of duties, staffing, communication and collaboration, and leadership and decision-making bodies. There are four secondary indicators under this dimension, which are authority and accountability system, people, data culture,³² and continuous improvement and optimization.³³

Institutional protection (IP)

Institutional protection is a protective barrier for data governance in manufacturing enterprises. Institutional protection refers to the development and implementation of data management processes by enterprises under the guidance of the government and other management authorities, and the systematic implementation of various data management tasks in order to obtain data that follows a series of standard specifications such as industry standards, national standards,³⁴ and so on. There are four secondary indicators under this dimension, which are standard norms, management system, laws and regulations,³⁵ and policy leadership.

In this study, a manufacturing enterprise data governance indicator system is established based on the above five dimensions. As shown in Figure 1, the system contains five first-level indicators and twenty-three second-level indicators.

Figure 1.

Data governance indicator system for manufacturing enterprises.

Design of questionnaire scale

Based on the diagram of the predefined indicator system for data governance in manufacturing companies, the research team designed a questionnaire on the degree of data governance implementation in manufacturing companies. Subsequently, the questionnaire was revised after integrating the comments from various parties, and it was finally determined that the questionnaire, except for the general questions (such as gender, age, and years of experience), was in the form of a structured questionnaire—a five-point Likert scale, which consisted of 23 items, with each secondary indicator corresponding to one item, as shown in Table 1. Respondents scored each question item, “strongly inconsistent” represents 1 point, “does not conform” represents 2 points, “medium” represents 3 points, “conforms represents” 4 points, and “very consistent” represents 5 points.

Table 1.

Questionnaire scale on the degree of data governance implementation in manufacturing enterprises.

Primary indicators	Secondary indicators	Scale entries
Data Strategy	Strategic Objectives	Q1.1 I am in an organization that can guarantee operational compliance, risk control, and value realization in the application of data
	Business Process management	Q1.2 I am in a unit that can use the data collected to gain insight into customer needs and market trends
	Strategy implementation	Q1.3 I am in a unit that has a well-built data governance implementation mechanism and pathway
	Monitoring and evaluation	Q4.1 The unit I work has an evaluation of the performance of data governance
Data asset management	Metadata	Q2.1 To avoid different definitions and terminology for the same data element in different departments, my unit has created a unified data glossary
	Master data	Q2.2 The organization I work for designs a unified master data standard for core business entity data (such as products, materials, equipment, suppliers, customers, and accounting accounts
	Data quality management	Q2.3 I work in a unit that uses technology to improve data quality planning, implementation, and control of a range of activities
	Data Application	Q2.4 The unit I am in will manage, process, and apply the data in a unified manner
	Data sharing	Q2.5 My unit was able to break through the data barriers between departments and between upstream and downstream industry chain entities
	Data Security	Q2.6 I am in a unit that can take certain measures to keep the data in a state of effective protection and legal use
	Full data Lifecycle	Q2.7 The unit I am in is able to supervise all aspects of data collection, transmission, storage, cleansing, integration, analysis, application, presentation, archiving, and destruction
Data infrastructure	Database	Q3.1 The unit I work in has electronic file cabinets that store different types of data
	Big data Platform	Q3.2 I work in a unit with a software platform that collects, stores, processes, computes, and analyzes large amounts of information and solves complex problems
	Data Analytics Mining	Q3.3 I work in a unit that understands business needs and cleans up incomplete, noisy, and inconsistent data in a timely manner
	Data Integration and fusion	Q3.4 I am in a unit that can integrate and combine data sets from different sources and forms to improve the quality and accuracy of the data
Organizational Guarantee	Authority And responsibility System	Q4.1 My unit can clarify the implementation organization of data governance and rationalize the authority and responsibility for the whole life cycle of data collection, sharing, opening, utilization, and destruction
	Personnel	Q4.2 Business and IT staff in my organization have expertise in data governance
	Data culture	Q4.3 The unit I work in already has core values and drivers when it comes to data governance
	Continuous improvement and optimization	Q4.4 The organization I work for has specific improvement plan and implementation strategy for data governance optimization
Institutional Protection	Standards Specification	Q5.1 The data I obtained in the unit is in line with national or industry standards
	Management system	Q5.2 I work in an organization that has a data governance system in place and effectively implements data management processes
	Laws and regulations	Q5.3 I work in a unit that fully understands the legal system constructed by the state and government departments at all levels for the sake of national development, national security, social stability. and public interest
	Policy Leadership	Q5.4 The unit I work in is able to respond positively to the policy requirements and promote the specific process of enterprise data governance

Exploratory factor analysis

This study uses exploratory factor analysis³⁶ to study the measurement items and factor structure, and initially verifies the structural dimensions of the manufacturing enterprise data governance indicator system. The research tool is SPSS27.0,³⁷ and the 23 items are selected through the model generation method to select secondary indicator items.

Sampling and measurement

The questionnaire was distributed offline, with IT personnel engaged in manufacturing companies, business personnel, and data management researchers of manufacturing companies as the main survey targets (see Table 2).

Table 2.

Basic demographic characteristics (n = 109).

Description	Number	Percentage (%)
Gender
Male	71	65.1
Female	38	34.9
Education level
College and below	5	4.6
Undergraduate	68	62.4
Master’s degree	36	33.0
Unit type
Manufacturing company	85	78.0
University or research institute	14	12.8
Other	10	9.2
Years of work
≤5 years or less	49	45.0
Within 5–10 years	31	28.4
Within 10–20 years	18	16.5
≥20 years	11	10.1

Project analysis

Item analysis method using extreme group analysis method, according to the sum sorting high and low grouping up and down 27% of the standard will be divided into two groups of item scores, test the difference between high and low groups in the item, according to the results of the test of significance to retain those who identify the effect of a better item, otherwise consider deleting the item. The results of the post-discrimination analysis showed that all 23 questions reached the level of significance, that is, the discriminant validity of the 23 questions of the scale was good.

Result analysis

Before performing the EFA test, the fitness test of the data was first performed. According to the data requirements of the exploratory factor analysis method, the sample data were to pass the KMO measure and the Bartlett’s Spherical Test, with the rule of determination being KMO >0.7, and for the Bartlett’s Spherical Test, a significance level of p < 0.05 was required. The KMO value of the data collected in this study was 0.885, while the moderateness of the correlation matrix between the variables according to the Bartlett’s spherical test (χ² = 1843.002, p < 0.001) indicated the presence of common factors between the variables, that is, the sample data were suitable for exploratory factor analysis.

In this study, the factors were extracted by principal component analysis,³⁸ and the factors with eigenvalues greater than 1, and the options with a factor loading of >0.5 and no cross-loading of the question items were extracted from the 23 question items of the data governance index system of the manufacturing enterprises by using the method of Orthogonal Great Rotation. Judging from the results of exploratory factor analysis, the data governance index system of manufacturing enterprises presents a five-factor structure. According to the item composition of each factor, they are named as follows: data strategy (4 items), data asset management (7 items), data infrastructure (4 items), organizational guarantee (4 items), and institutional protection (4 items) (see Table 3). As shown in Table 4, the Cronbach’s Alpha coefficient³⁹ of the 23-item scale obtained through exploratory factor analysis was 0.931, indicating that the scale has good stability.

Table 3.

Results of EFA of data governance index system of manufacturing enterprises (n = 109).

Secondary index question items	Factor loadings
Secondary index question items	F1	F2	F3	F4	F5
Q2.1 Metadata	0.836
Q2.4 Data application	0.829
Q2.7 Data full lifecycle	0.820
Q2.6 Data security	0.816
Q2.2 Master data	0.801
Q2.3 Data quality management	0.796
Q2.5 Data sharing	0.791
Q4.3 Data culture		0.852
Q4.2 Personnel		0.851
Q4.1 Authority and responsibility system		0.786
Q4.4 Continuous improvement and optimization		0.761
Q3.4 Data integration and fusion			0.836
Q3.2 Big Data platform			0.824
Q3.3 Data analysis and mining			0.796
Q3.1 Database			0.781
Q5.2 Management system				0.825
Q5.1 Standard specification				0.811
Q5.4 Policy leadership				0.788
Q5.3 Laws and regulations				0.768
Q1.1 Strategic objectives					0.842
Q1.2 Business process management					0.815
Q1.4 Monitoring and evaluation					0.812
Q1.3 Strategy implementation					0.794
Eigenvalues (non-rotating)	9.281	2.798	2.283	1.819	1.368
Explanation rate (cumulative 76.306%)	40.356%	12.164%	9.926%	7.909%	5.951%

Table 4.

Reliability test of the data governance secondary indicator scale for manufacturing enterprises (n = 109).

Factor	DS	DAM	DI	OG	IP	Overall reliability
Standardized Cronbach’salpha coefficient	0.882	0.939	0.890	0.907	0.892	0.931

Confirmatory factor analysis

The quality of a set of measurement models includes two aspects: external quality and intrinsic quality. External quality is suitable for measuring the overall fitness of the model and is mainly measured by various fitness indicators of the model. Intrinsic quality is reflected in the reliability and validity of measurement items, where validity refers to the convergent validity and discriminant validity of measurement items. In this paper, AMOS28.0 was used for validation factor analysis, and the model was evaluated using the values of various types of fitness indicators, and AVE⁴⁰ and Composite reliability (CR)⁴¹ were calculated on the basis of the validation factor analysis to determine its convergent validity and discriminant validity.

Sampling and measurement

This study used both offline and online methods of resampling. A total of 344 questionnaires were distributed and a total of 327 questionnaires were returned, of which 309 were valid and the basic demographic characteristics are shown in Table 5.

Table 5.

Basic demographic characteristics (n = 309).

Description	Number	Percentage (%)
Gender
Male	221	71.5
Female	88	28.5
Education level
College and below	16	5.2
Undergraduate	198	64.1
Master’s degree	95	30.7
Unit type
Manufacturing company	253	81.9
University or research institute	33	10.7
Other	23	7.4
Years of work
≤5 years or less	171	55.3
Within 5–10 years	82	26.5
Within 10–20 years	36	11.7
≥20 years	20	6.5

Second-order confirmatory factor analysis

Use the five first-level indicators as the first-level latent variables and the overall “manufacturing enterprise data governance” (DGME) as the second-level latent variables to conduct a second-level CFA test. At the same time, the items corresponding to the 23 secondary classification indicators were used as measurement items to test whether the sample data supports the theoretical model obtained by EFA. As shown before, the model is evaluated by analyzing the fitness index values obtained. The fit of the theoretical model to the questionnaire sample was assessed using absolute fit indices such as chi-squared freedom ratio (χ²/df),⁴² root-mean-squared error of approximation (RMSEA),⁴³ goodness of fit index (GFI), and adjusted goodness fitness index (AGFI). The degree of fit improvement of the theoretical model relative to the null model is judged by normed fit index (NFI), incremental fit index (IFI), and comparative fit index (CFI). The degree of parsimony of the model is mainly judged by parsimony goodness fitness index (PGFI) and parsimony normative fit index (PNFI).

χ²/df < 3 indicates that the model fits well, and RMSEA less than 0.08 indicates that the model fitting goodness is moderate. GFI and AGFI are greater than 0 and less than 1, with the closer to 1 indicating a better model fit. NFI, IFI, and CFI should be between 0 and 1, the closer to 1 the better. From the fitness metrics in Table 6, it can be seen that the whole measurement model presents a good fit, with a chi-square degrees of freedom ratio of χ²/df = 1.066, which is in line with the standard. The root mean square error of approximation RMSEA = 0.015 is ideal, and the fitness index GFI = 0.938 exceeds the goodness standard of 0.9. The adjustment fitness index AGFI = 0.924, exceeding the excellence standard of 0.9. It can be seen that all fitness indicators of the model have reached acceptable levels, indicating that the model structure set in the EFA is reasonable.

Table 6.

Fit indices of this measurement model.

Fitness index	Absolute fitness indexes				Value-added fitness indexes			Parsimony fitness indexes
Fitness index	X²/df	RMSEA	GFI	AGFI	NFI	IFI	CFI	PGFI	PNFI
Model	1.066	0.015	0.938	0.924	0.951	0.997	0.997	0.765	0.846
Reasonable domain value	<3	<0.08	>0.9	>0.9	>0.9	>0.9	>0.9	>0.5	>0.5

In addition to the above-mentioned fitting index that needs to meet the standards, the load values of the measurement items must also meet certain requirements. When the factor loading value is greater than 0.71 the factor explains 50% of the variance in the observed variable, factor loading values greater than 0.63 mean that the factor explains 40% of the variance in the observed variable, and factor loading values greater than 0.55 mean that the factor explains 30% of the variance in the observed variable. When the factor loading value is greater than 0.55, it means the situation is good. The factor loading levels of the model are shown in Figure 2. It can be found that the standardized factor loading of “data strategy,” “data asset management,” “data infrastructure,” “organizational guarantee,” and “institutional protection” are 0.61, 0.69, 0.70, 0.65, and 0.57, respectively. The standardized factor loadings of the 23 secondary indicators were in the range of 0.79–0.88, which were all at desirable levels. The standardized factor loadings of the 23 secondary indicators range from 0.79 to 0.88, all at ideal levels.

Figure 2.

Results of validated factor analysis of data governance indicator system for manufacturing companies^①. Note 1: The Figure 2 is replicated from the output of AMOS. The oval and rectangle represent the latent variable and the measure term, respectively. e1 to e23 and e24 to e28 represent the residual terms of the measure and latent variable, respectively. The numbers on the arrows are the normalized factor loading values.

Convergent and discriminant validity

Convergent validity

First, a first-order CFA test was done with the first-level indicator as the latent variable and all the question items under the first-level indicator as the measurement items. In this study, the method proposed by Fornell and Larcker was used to determine the convergent validity of the data governance indicator system of manufacturing companies through CR and AVE. Taking AVE >0.5 and CR > 0.8 as criteria, when AVE >0.5 proves that the explanatory strength of the extracted factor portion of the measurement term exceeds the variance due to measurement error and has good convergent validity. By testing the convergent validity of the existing scales, the results are shown in Table 7. The AVEs for the five first-order factors (tier 1 indicators) for “Manufacturing enterprise data governance” were 0.648, 0.683, 0.649, 0.702, and 0.689, respectively, which were all greater than 0.5 to meet the criteria. The CRs were 0.881, 0.938, 0.881, 0.904, and 0.899, respectively, which were all higher than the judge mental criterion of 0.8. Therefore, it is possible to establish the internal convergent validity of the data governance indicator system of manufacturing companies.

Table 7.

Results of AVE and CR analysis of the five first-order CFA tests.

Factor	DS	DAM	DI	OG	IP
AVE	0.648	0.682	0.649	0.702	0.689
CR	0.881	0.938	0.881	0.904	0.899

Discriminant validity

The average value of the measurement items under each first-level indicator is used as the first-level indicator score to test the discriminant validity between the first-level indicators. When the squared Pearson correlation coefficients between the level 1 indicators are all below AVE, it means that the five level 1 indicators have both common attributes and their own independence from each other. The results, as shown in Table 8, show that each level of indicators that make up the data governance indicator system of manufacturing enterprises has a certain degree of correlation between the indicators and at the same time has a good differentiation of validity, so as to form an organic whole.

Table 8.

Descriptive statistics and discriminant validity of data governance primary indicators of manufacturing enterprises^② (n = 309).

Factor	Mean	Standard deviation	DS	DAM	DI	OG	IP
DS	3.040	1.346	[0.648]	0.165	0.201	0.176	0.102
DAM	3.008	1.362	0.406***	[0.682]	0.279	0.197	0.128
DI	2.978	1.351	0.448***	0.528***	[0.649]	0.154	0.156
OG	3.053	1.360	0.419***	0.444***	0.392***	[0.702]	0.194
IP	2.907	1.350	0.320***	0.358***	0.395***	0.441***	[0.689]

Note 2: In Table 8, *** indicates p < 0.001, the numbers in [ ] on the diagonal line are the AVEs of the corresponding primary indicators, and the squared correlation coefficients of the scores of the corresponding paired primary indicators are in the upper right corner of the matrix. AVEs are greater than the numbers above their same column and the numbers to the right of their peers, indicating that each primary indicator has sufficient discriminant validity.

Comparison of competitive models

In order to further confirm the optimality of the measurement model, the second-order 5-factor model was used as the baseline model, and four competing models were formed by combining the intrinsic relationship of the five factors and the results of empirical research. One of the competing models is the first-order five-factor model, which treats the five factors as five latent variables that are only correlated with each other and do not belong to “manufacturing enterprise data governance.” The second competitive model is a single-factor model, which skips the first-level indicators and attributes the 23 second-level indicators to the latent variable of “manufacturing enterprise data governance.” The third competitive model is a first-order four-factor model, that is, organizational guarantee and institutional protection are merged into one factor, and then are only related to each other with data strategy, data asset management, and data infrastructure. The fourth competitive model is a first-order three-factor model that combines data strategy, organizational guarantee, and institutional protection into one latent variable. Furthermore, it is composed of three latent variables that are only related to each other and do not belong to one high-order latent variable.

The main fitness indices of the baseline model and the competing models are shown in Table 9; comparing with the baseline model, the first-order 4-factor model, the first-order 3-factor model, and the single-factor model have poor fit indices, and the first-order six-factor model is the most effective model among the first-order models. Based on the previous study, it is known that there may be a higher-order factor for the five dimensions of data governance in manufacturing companies, with correlation coefficients taking values between 0.320 and 0.528, all of which are at the significant level, indicating that there is indeed a higher-order co-factor between the five factors. In the second-order 5-factor model, the loadings of the observed variables on the corresponding factors are between 0.78 and 0.88, indicating that the loadings are relatively high. The relationship between the first-order factor and the second-order factor is between 0.57 and 0.70, indicating a strong relationship. The above analysis shows that the manufacturing enterprise data governance indicator system proposed in this article has high internal and external quality.

Table 9.

Comparison of the fit indexes of the baseline and competitive models.

Fit index	X2/df	RMSEA	GFI	AGFI	NFI	IFI	CFI	PGFI	PNFI
Baseline model	1.066	0.015	0.938	0.924	0.951	0.997	0.997	0.765	0.846
First-order 5-factor	1.047	0.012	0.940	0.924	0.953	0.998	0.998	0.749	0.829
One-factor model	—	0.178	0.490	0.387	0.501	0.526	0.524	0.408	0.456
First-order 4-factor	3.684	0.093	0.766	0.712	0.833	0.872	0.872	0.622	0.737
First-order 3-factor	5.872	0.126	0.656	0.581	0.730	0.765	0.764	0.539	0.655
Reasonable domain value	<3	<0.08	>0.9	>0.9	>0.9	>0.9	>0.9	>0.5	>0.5

Research conclusion

This paper analyzes the literature and the implications of relevant policy documents to summarize the composition of data governance indicators in manufacturing enterprises as “one body, five sides.” That is, data governance in manufacturing enterprises consists of five dimensions, namely, data strategy, data asset management, data infrastructure, organizational guarantee, and institutional protection in that order, with more secondary indicators divided under each dimension. A manufacturing enterprise data governance indicator system containing 5 first-level indicators and 23 second-level indicators was constructed, and a questionnaire containing 23 items was designed based on the second-level indicators. In order to improve the reliability and validity of the scale, EFA was conducted on 109 questionnaire data using SPSS27.0. It is determined that the data governance indicator system of manufacturing enterprises presents a five-factor structure, with a cumulative variance contribution rate of 76.306%. The reliability analysis results also show that the internal consistency reliability of the first-level indicators is higher than 0.79, indicating that the corresponding measurement items of the five first-level indicators of the scale have good consistency. With the help of AMOS28.0, CFA was carried out on 309 pieces of data. First, the second-order CFA found that the overall model showed good fit. Secondly, the AVE and combined reliability were used to test the convergent validity and discriminant validity. The results show that the measurement items corresponding to the five first-level indicators all have good convergent validity and discriminant validity, and the five first-level indicators are jointly attributed to the higher-order factor of manufacturing enterprise data governance. Finally, through comparison with various fitness indicators of the competitive model, it was confirmed that the second-order five-factor concept of manufacturing enterprise data governance is optimal. This conclusion provides reference for data governance evaluation and data governance capability improvement of manufacturing enterprises, and helps manufacturing enterprises digital transformation.

Shortcomings and prospects

The data governance index system of manufacturing enterprises constructed in this paper can improve the data governance level of manufacturing enterprises and provide certain reference for promoting the digital construction of manufacturing enterprises. The shortcomings are as follows: (1) data governance is not a project work, but a persistent work, dynamic changes in the process of practice, in the practice of exploration, there will certainly be other elements of the indicators of data governance on the manufacturing enterprise, to give a specific data governance indicator system model has the suspicion of “cutting the feet to fit the shoes.” (2) Due to the large scope of the manufacturing industry, most of the data in this study were collected from questionnaires distributed by team members attending manufacturing-related conferences, the sample size is too small, and more data will be collected in the future to verify the validity of the model. A complete and practical indicator system needs to determine the weights of all levels and indicators, which is one of the next research directions.

Footnotes

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This paper is a collaborative innovation project of The University Synergy Innovation Program of Anhui Province, “Research and Development of Embodied Intelligent Composite Robot for AI Popularization of Science Education” (Project No.: GXXT-2023-108) and General Project of National Social Science Foundation of China “Research on the Theoretical Construction and Operational Mode of Scientific and Technological Security Intelligence in the New Era” (Project No.: 19BTQ089).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Song

Gao

Nan

, et al. Ecological restoration for mega-infrastructure projects: a study based on multi-source heterogeneous data. Eng Construct Architect Manag 2024; 31(9): 3653–3678.

Nadal

Jovanovic

Bilalli

, et al. Operationalizing and automating data governance. J Big Data 2022; 9(1): 117.

Liang

. Blockchain-based and value-driven enterprise data governance: a collaborative framework. Sustainability 2023; 15(11): 8578.

Janssen

Brous

Estevez

, et al. Data governance: organizing data for trustworthy artificial intelligence. Gov Inf Q 2020; 37(3): 101493.

Bozkurt

Rossmann

Konanahalli

, et al. Toward urban data governance: status-quo, challenges, and success factors. IEEE Access 2023; 11: 85656–85677.

Mercuri

Emerson

. Normative challenges in data governance: insights from global health research. Adv Health Sci Educ 2024; 29(4): 1453–1461.

Abraham

Schneider

vom Brocke

. Data governance: a conceptual framework, structured review, and research agenda. Int J Inf Manag 2019; 49: 424–438.

Wei

Yang

, et al. Blockchain-based design of a government incentive mechanism for manufacturing supply chain data governance. Sustainability-Basel 2023; 15(8): 6968.

Zorrilla

Yebenes

. A reference framework for the implementation of data governance systems for industry 4.0. Comput Stand Interfac 2022; 81: 103595.

10.

Mao

Qiao

, et al. Government data governance framework based on a data middle platform. Aslib J Inf Manag 2022; 74(2): 289–310.

11.

Zhang

Sun

Zhang

. Data matters: a strategic action framework for data governance. Inform Manage-Amster 2022; 59(4): 103642.

12.

Abueed

RAI

Aga

. Sustainable knowledge creation and corporate outcomes: does corporate data governance matter? Sustainability 2019; 11(20): 5575.

13.

Russo

Stol

. PLS-SEM for software engineering research: an introduction and survey. ACM Comput Surv 2021; 54(4): 78.

14.

Yin

. Data governance and green technological innovation performance: a curvilinear relationship. J Clean Prod 2022; 379: 134441.

15.

Gegenhuber

Mair

Lührsen

, et al. Orchestrating distributed data governance in open social innovation. Inform Organ-UK 2023; 33(1): 100453.

16.

Smania

Ayala

Coreynen

, et al. Data-related tensions in digital servitization ecosystems: a systematic literature review. Ind Mark Manag 2024; 123: 31–48.

17.

Götz

le Gall

Konrad

, et al. LEAPS data strategy. Eur Phys J Plus 2023; 138(7): 617.

18.

Van Looy

. A quantitative and qualitative study of the link between business process management and digital innovation. Inform Manage-Amster 2021; 58(2): 103413.

19.

Xie

Peng

. Data asset management and visualization based on intelligent algorithm: taking power equipment data as an example. Teh Vjesn 2022; 29(4): 1109–1119.

20.

Khalid

Zimányi

. Repairing raw metadata for metadata management. Inf Syst 2024; 122: 102344.

21.

Gualo

Caballero

Rodríguez

, et al. A data quality model for master data repositories. Informatica-Lithuan 2023; 34(4): 795–824.

22.

Zhang

Jeong

Lee

. Data quality management in the internet of things. Sensors-Basel 2021; 21(17): 5834.

23.

Chen

Choe

Cong

, et al. Data-driven mergers and personalization. Rand J Econ 2022; 53(1): 3–31.

24.

Gao

. Company data sharing, product innovation and competitive strategies. Expert Syst Appl 2023; 234: 121083.

25.

Chen

Liu

. Promotion and advancement of data security governance in China. Electronics-Switz 2024; 13(10): 1905.

26.

Koo

Kang

Kim

. Security and privacy in big data life cycle: a survey and open challenges. Sustainability 2020; 12(24): 10571.

27.

Clutterbuck

. Data infrastructures and the governance of their accompanying narratives. Br J Sociol Educ 2022; 43(1): 120–139.

28.

Bonnard

Arantes

Lorbieski

, et al. Big data/analytics platform for Industry 4.0 implementation in advanced manufacturing context. Int J Adv Manuf Technol 2021; 117(5): 1959–1973.

29.

Garouani

Ahmad

Bouneffa

, et al. Towards big industrial data mining through explainable automated machine learning. Int J Adv Manuf Technol 2022; 120(1-2): 1169–1188.

30.

Xiang

Yin

Wang

, et al. Systematic method for big manufacturing data integration and sharing. Int J Adv Manuf Technol 2018; 94(9-12): 3345–3358.

31.

Bridoux

Stoelhorst

. Stakeholder theory, strategy, and organization: past, present, and future. Strateg Organ 2022; 20(4): 797–809.

32.

Staudt

Hoffmann

. How a utility company established a corporate data culture for data-driven decision making. MIS Q Exec 2024; 23(1): 3.

33.

Lyu

Chen

. Effect of founder control on equity financing and corporate performance-based on moderation of radical strategy. Sage Open 2022; 12(2): 21582440221085013.

34.

Huang

Yong

Zhao

, et al. Review of the development of China's Eco-industrial Park standard system. Resour Conserv Recycl 2019; 140: 137–144.

35.

Cue

Doornink

George

, et al. Data governance in the dairy industry. Animals 2021; 11(10): 2981.

36.

Goretzko

. Factor retention in exploratory factor analysis with missing data. Educ Psychol Meas 2022; 82(3): 444–464.

37.

Qiao

Huang

. Analysis on the stress relief of college students by physical exercise. Psychiatr Danub 2022; 34: S924–S928.

38.

Fraino

. Using principal component analysis to explore multi-variable relationships. Nat Rev Earth Environ 2023; 4(5): 294.

39.

Toro

Pena-Sarmiento

Avendaño-Prieto

, et al. Empirical analysis of cronbach’s alpha coefficient as a function of question response options, sample size and outliers. Rev Iberoam Diagn Ev 2022; 2(63): 17–30.

40.

Libório

Diniz

AMA

Vieira

DAG

, et al. Subjective–objective method of maximizing the average variance extracted from sub-indicators in composite indicators. Soc Indic Res 2024; 13: 100431.

41.

Assis

Coelho

AJC

Rezende

, et al. Unsupervised machine learning techniques applied to composite reliability assessment of power systems. Int T Electr Energy 2021; 31(11): e13109.

42.

Martins

AMEDL

Santos

AMR

Alencar

, et al. Psychometric properties of an oral health literacy scale for people living with diabetes. Braz Oral Res 2023; 37: e022.

43.

Fitzgerald

Estabrook

Martin

, et al. Correcting the bias of the root mean squared error of approximation under missing data. Methodology-Eur 2021; 17(3): 189–204.