1 Introduction

As far as the depth and range of products are concerned, the manufacturing industry is one of the most complex. The manufacturing industry is categorized into two major segments, viz. discrete manufacturing and process manufacturing, based on the production processes they follow. It is further classified based on the product such as metal & mining, chemicals, aerospace, automotive, pharmaceuticals, etc. The industry is strongly affected by digitization, which has led to digital services like predictive maintenance and disruptive product innovations [1]. Today, one of the critical success factors for manufacturing companies is the efficient and effective management of data and its analysis [2]. With the advent of industry 4.0 technologies like IoT, blockchain, AI, and big data, it is becoming standard practice to support decision-making based on a comprehensive evaluation of data collected from varied sources viz. production equipment, sensors, and enterprise resource planning systems. Also, the advancement of industry 4.0, such as IoT, has resulted in an incremental increase in global data [3, 4]. These advancements create areas where big data analytics (BDA) tools and techniques can be applied in the manufacturing sector. BDA will revolutionize in making informed decisions, identify variables that affect production, monitor and mitigate risk, increase operational efficiency, carrying out market research, competitor analysis for any particular product, etc. [5, 6]. The concept of BDA was derived from internet giants like Amazon, Netflix, and Google, who analyze consumer activity data to provide customizable and personalized offerings [7].

There is colossal literature suggesting the importance of BDA and its possible revolutionary impact on operational and strategic decision making. Researchers are still investigating its application and decision-making impact in new business areas.

Despite the importance of BDA in revolutionizing the business operation and activities, most of the organization has failed at the implementation stage. A report suggested that around 80% of the organizations failed at the implementation stage [8,9,10]. Half of the big data projects implemented with high expectations failed to deliver [11]. In another report published in InfoWorld suggest 100% failure, even in the presence of mature Technology [12].

The majority of manufacturing industry leaders want to implement BDA for improving operational efficiency and have a competitive edge. But they are unable to because of a lack of awareness about BDA, related infrastructure, and many other organizational barriers. Thus, it is vital to investigate and analyze barriers to BDA implementation in today's technologically advanced world. It will help companies to formulate and plan better before implementing BDA in their organizations.

There is limited research on the barriers to BDA implementation [13, 14]) conducted a comprehensive review of the barriers and developed a framework on big data analytics throughout the product lifecycle [15] qualitatively investigated the drivers and barriers of industry 4.0. Still, there is limited literature available on barriers to the implementation of BDA in the manufacturing sector.

For an emerging economy like India, manufacturing is a critical driving force of the economic engine. It offers unparalleled job opportunities that transform societies and result in a high growth rate. Hence, developing countries like India must implement BDA for their rapidly expanding manufacturing sector and remain competitive. The implementation of BDA will help in uncovering new avenues to gain a competitive advantage through innovation. Manufacturing companies face an obstacle in the implementation of BDA. Hence, it is vital to conduct a quantitative analysis of the barriers in adopting BDA in the manufacturing sector focused on India. Such a study will guide and assist industry leaders in framing better policies for its implementation.

The problem statements of the study are:

  1. 1.

    What are the barriers to the implementation of BDA in the manufacturing sector?

  2. 2.

    How can industry leaders quantitatively examine the barriers?

  3. 3.

    Can the final result help industry leaders to formulate better strategies for BDA implementation?

Following are the objectives of this research for answering the above problem statements:

  • To determine the barriers to the implementation of Big data analytics in the manufacturing sector.

  • To quantitatively evaluate the barriers using analytical approach—ISM and MICMAC

  • To identify the most critical barrier using Fuzzy AHP

This paper explores the existing literature available to identify the barriers, categorized based on different functions of an organization. A total of 16 barriers are determined from the rigorous review of existing research. The opinion and judgment on these barriers are taken from industry experts from the automobile, steel, automotive parts manufacturer, and electrical equipment industries.

ISM and MICMAC are the analytical techniques used to classifies the barriers into different impact levels and importance. ISM helps to establish interrelationships among these barriers, and MICMAC is used to classify barriers into driving and dependence power. Independent factors (barriers) have high driving power and are the key factors that were further analyzed using Fuzzy AHP to find their comparative priority/importance.

The organization of this paper is as follows: Literature review on the topic is summarized in Sect. 2. Section 3 presents the research methodology, i.e., ISM-MICMAC. Section 4 discusses the final result and the Fuzzy AHP analysis. Section 5 concludes the outcome of the paper. Section 6 suggests future scope. The paper ends with references and an Appendix.

2 Literature review

This section reviews big data analytics concepts, its significance, application, and importance in the manufacturing industry. At the last, critical barriers to BDA implementation are identified.

2.1 Big data and big data analytics

Big data's technological development is considered the critical area of advancements in information and communication technology. It has evolved rapidly due to the ever-increasing affordability and availability of electronic devices and networks, driven mainly through social media and the increased presence of the Internet of Things (IoT). Firms can transform themselves into data-driven organizations due to technological advancements in big data awareness, infrastructure, analytics, and related services. Every firm wants to stay competitive; hence they need to build their future strategies surrounding big data owing to its potential of being a game-changer. Big data is usually used to describe a massive volume of datasets that are very complex and difficult to handle using conventional data analysis applications [16, 17]. In other words, the dataset size is so massive that current tools and storage systems are not capable enough to store, manage, and process it at an adequate time.

Organizations and practitioners may have different concepts on big data, but the definition introduced by [18] is widely used. According to the definition, big data has three dimensions, viz. volume, variety, and velocity (3 V). Several other definitions exist for big data like [19] describe it in three-part viz. high volume-velocity-variety that assists data-driven decision-making and can't be managed by conventional tools. The big data concept is further expanded by [20] when they claimed that it should have Veracity as a fourth dimension. Veracity refers to the unreliability and uncertainty in data sources, resulting from inaccuracy, inconsistency, incompleteness, and subjectivity in data. There are further additions to the above concepts by various organizations, but the general idea is the same.

We lack an integrated view of big data in the industry. Figure 1 clearly explains the interrelationship between different dimensions of big data [21]. Representation of the three dimensions of big data is done with the help of the triangular framework with the three edges representing them. Inside space is occupied by five more dimensions, i.e., complexity, decay, value, integrity, and variability, which are affected by the growth of three primary dimensions. Thus basic framework of 4 V is modified to a 7 V framework, which includes: volume-velocity-variety-veracity-value-variability-volatility [22,23,24]. The growth of the three dimensions negatively impacts Veracity but positively affects the other four dimensions. Traditional data is a subset of big data with the same three dimensions but with a smaller scope than big data. With the growth of three edge dimensions, the magnitude of big data also expands. The expansion of all three edge dimensions is interlinked to each other. Expansion of any one dimension will affect all the other seven dimensions.

Fig. 1
figure 1

An integrated view of big data [21]

Full size image

From a broader perspective, big data also refers to the tools and advanced analytical techniques used for managing and processing this complex and massive dataset, which is evolving rapidly to extract valuable information and facilitate decision-making [25]. When it comes to the manufacturing industry, the data analytic process is more widely familiar as Business Intelligence (BI) due to some marked difference between BDA and BI based on the type of data and questions answered by its analysis. For example, a database about BI is finite to a particular timeline in an organizational data architecture. The data are accessed via a specialized data mart developed from a data warehouse in an offline mode, contrary to the real-time data format in BDA. Also, most of the time, data sources are limited to organizations' internal and external databases having low complexity. Some marked differences between BI and BDA are listed in Table 1.

Table 1 Business intelligence vs. Big data analytics
Full size table

2.2 Big data analytics in the manufacturing industry

The manufacturing industry faces the wind of change with a more volatile and fragmented business environment than ever before. The industry must adapt to remain competitive or risk completely fading away. The next question firms face, where to start and how to stay competitive? The answers lie in BDA or BI as widely known in the industry, where the advanced analytical technique of data management is utilized to aggregate datasets in a structured way. These aggregated structured datasets help extract valuable information that aids in data-driven decision-making [7]. Many world-class firms in the manufacturing sector, i.e., automotive, steel, chemical, power, pharmaceutical, and garments industry, are using BDA tools to improve operational efficiency and minimize process flows, increase productivity, reduce downtime and enhance the quality of a product [26,27,28].

Business Intelligence refers to leveraging a range of tools that transform data to provide quick and actionable insights about an organization's current state. It offers detailed intelligence to business owners through reports, dashboards, graphs, charts, and summaries based on data visualization tools. As shown in Table1, BI tells you the current state of the business, i.e., what's happening now and what happened in the past, to lead us to the present state. BDA, on the contrary, is—predictive analytics, which tells what is going to happen in the future, and prescriptive analytics, which explain what one should do to have better outcomes. A review of the analytical methods employed on Big Data to extract the Hindsight, Insight, and Foresight from the data are summarized in [22, 23].

2.3 Importance of data analytics in the business environment

Significant contributions can be made to areas such as supply chain, distribution optimization, new product development, demand prediction, and customer relationship through big data analytics. The importance of data analytics is demonstrated by various researchers, such as [3]. They highlighted that the supply chain problem could be quantitatively resolved by making available useful quality data [29] mentions how manufacturing performance can be improved with the help of BDA [30] explored BD application in product lifecycle management [7] examined BD application for predictive analysis to measure the supply chain sustainability.

We can make the following inferences from the above scholarly articles: (1) the information from data points should be extracted and presented with the motive of helping end-users. (2) proper data integration and supply chain data analysis will help increase visibility upward and downward of the supply chain. (3) Spend analysis of data can reduce procurement costs by a strategic allocation of contracts among vendors. (4) product line profitability can be impacted by using sales data analytics. In other words, BDA is the natural evolution of big data in the manufacturing industry [31]. Numerous literature suggests the importance of BDA and its possible revolutionary impact on operational and strategic decision making. Owing to the BDA influence, more and more research is in progress. Table 2 compiles the current studies on the importance of BDA in different and new sectors or operational areas.

Table 2 List of recent studies on the importance of BDA
Full size table

2.4 Barriers of data analytics in the manufacturing industry

In the BD era in this global market, companies in the manufacturing sector have begun to embrace BDA tools and techniques to sustain and remain competitive. Experts believe that India is one of the ideal countries to benefit from big data analytics. The government of India started the campaign of Make-In-India, Digital India, etc., to boost the manufacturing sector. These initiatives aim to help Indian industries strive for sustained business growth and development [42]. A study on Indian firms focused on the impact of BDA on social performance, environmental performance, and collaborative performance showed a positive effect [43, 44]. A recent survey done be [45] on Indian manufacturing industries indicated that 65% of the organization had set a top priority for BDA investment in the next couple of years. The study reflected predictive maintenance, connected supply chain, reduced energy consumption, production optimization, lower price of sensors and high computing needs, and connected customers as the key factors driving digital manufacturing in India. Still, the adoption of digital technologies in India is in its infancy. The top Indian Government office of the Department of Science & Technology expressed BDA as crucial for the Indian manufacturing sector. Polaris India Head and MD also pointed out the importance of BDA towards capturing, cleaning, and analyzing machine data to reveal insights that can improve performance [46].

Still, limited research exists on BDA implementation in the Indian manufacturing industry. Research article done in the Indian context are summarized in Table 3. It is observed that only the sales and marketing division is adopting data analytics to a certain extent for production based on sales over time. Manufacturers are reluctant to implement data analytics due to many hurdles and the myth they possess due to a lack of awareness. Hence it is of utmost importance to investigate these barriers faced in the implementation of big data analytics.

Table 3 Summary of existing research in Indin context on BDA
Full size table

In this paper, the existing literature is examined using keywords like a barrier in the implementation of BDA, BDA in the manufacturing industry, barriers in adoption of BDA, issues faced in the implementation of big data analytics, and BDA, barriers to adoption of BDA, etc.

We could find a limited related article. Research by [6] has assessed challenges for the implementation of Industry 4.0 and its implication on process safety and environmental protection [14] explores the barriers to BDA in Bangladesh's manufacturing supply chain [51] investigated the barriers of BDA in the automotive sector through qualitative analysis [52] recommended specific tactics for addressing BD barriers by qualitative analysis [53] undertook an interpretive study to explore the adoption challenge of BDA in the South African telecom industry. Table 4 summarizes the research related to BDA implementation.

Table 4 Summary of existing research on the Barrier to BDA implementation
Full size table

3 Research methodology

This paper uses the ISM technique for identifying contextual relationships among 16 barriers in the implementation of data analytics in the manufacturing industry. MICMAC is further used to categorize the barriers into four groups based on driver and dependence power. Four groups are autonomous, dependent, linkage, and independent. Barriers in the independent group are further prioritized using Fuzzy AHP. Flow chart for research methods is shown in Fig. 2.

Fig. 2
figure 2

Research method flow chart

Full size image

3.1 Barrier Identification, expert review and categorization

Sixteen most essential barriers in the implementation of BDA are identified through rigorous literature review. These barriers were validated through the interaction with the relevant industry experts (Appendix B Table 31). The barriers are further categorized under six broad groups (Fig. 3). The barriers identified are listed in Table 5, along with their references.

Fig. 3
figure 3

Categorized barriers to BDA in the manufacturing industry

Full size image
Table 5 Barriers in implementation of data analytics in the manufacturing industry
Full size table

3.2 Interpretive structural modeling (ISM)

Warfield [64] was the first to propose the Interpretive Structural Modeling (ISM) approach to analyze and visualize the complex problem in a hierarchical structure and manage decision making. It is an interactive learning process that structures complex and directly related variables into a comprehensive and well-defined systematic model. It also helps in establishing interrelationships among identified variables. In ISM, practical experience and judgment of subject matter experts are used to decide whether the variables are connected or not and how they behave if connected [65]. Based on situations, the relationship between direct and indirect variables is discovered more precisely compared to seeing individual factors in isolation. Owing to this benefit, the researcher selects this approach for establishing the interrelationships between the identified barriers [66]. Moreover, [67] mentions that when the numbers of elements are large, it is a better approach for solving the complexity of relationships.

The ISM-MICMAC has been a trusted tool for researchers to conduct studies across various fields; however, this research paper creates a unique amalgamation of these tools concerning geography and area of study. The areas in which other researchers have used these tools are tabulated in Table 6.

Table 6 Summary of ISM-MICMAC used in the field of study
Full size table

The steps involved in ISM methodology are below [80]:

  1. (1)

    List the barriers relevant to the study.

  2. (2)

    Establish a contextual relationship between the listed barriers with the help of subject matter experts.

  3. (3)

    A pairwise relationship matrix (self-interaction matrix (SSIM)) is developed for the barriers with the help of subject matter experts.

  4. (4)

    The Reachability matrix is developed from the SSIM, and the matrix is checked for transitivity.

  5. (5)

    The barriers are partitioned into different levels using the reachability matrix

  6. (6)

    Develop a Conical matrix using levels and reachability matrix.

  7. (7)

    Develop Digraph and remove the transitive links.

  8. (8)

    ISM model is created from the final resultant Digraph.

  9. (9)

    The ISM model developed is reviewed to check for conceptual inconsistency

3.2.1 Data collection

The ISM approach is based on the opinion of subject matter experts to develop contextual relationships among various variables. Industry experts from the manufacturing sectors were consulted for the study (Appendix B Table 31). All the individual responses were discussed with the particular respondent to understand their approach. All the different responses were then consolidated.

3.2.2 Structural self-interaction matrix (SSIM)

SSIM Matrix records all pairwise relationships between two barriers (i and j). Barriers in row i is compared with the barriers in column j. Responses are gathered by comparing barriers i leads to barriers j. The Symbols used for establishing the direction of relationships between the barriers (i and j) are as follows:

  • V: Barriers i will lead to barriers j;

  • A: Barriers j will lead to barriers i;

  • X: Barriers i and j will lead to each other; and

  • O: Barriers i and j are unrelated.


Explanation: Lack of infrastructure facility, unavailability of specific data tools, and training facility may lead to unskilled IT personnel, lack of awareness about DA, and higher investment. Similarly, lack of long-term vision, lack of management initiative, and top management commitment will lead to data security and privacy concerns, performance & scalability, poor data quality, absence of any policy to share data among the organization, and complexity of data integration. Lack of management initiative leads to a lack of funding and vice versa. The complexity of data integration leads to high-cost investment and performance issues, whereas it is affected by poor data quality. Insufficient funds will degrade the organizational performance and adaptability to change. Subsequently, it will degrade the quality, security, and privacy of data.

3.2.3 Initial reachability matrix

The initial reachability matrix is formed by replacing the value of V, A, X, and O in the SSIM matrix (Table 7), by binary digits 1 and 0. Rules followed for such transformation is outlined below [81, 82]:

  • Cell (i, j) and cell (j, i) with value V in the SSIM matrix will be replaced by 1 and 0, respectively, in the initial reachability matrix.

  • Cell (i, j) and cell (j, i) with value A in the SSIM matrix will be replaced by 0 and 1, respectively, in the initial reachability matrix.

  • Cell (i, j) and cell (j, i) with value X in the SSIM matrix will be replaced by 1 and 1, respectively, in the initial reachability matrix.

  • Cell (i, j) and cell (j, i) with value O in the SSIM matrix will be replaced by 0 and 0, respectively, in the initial reachability matrix.

Table 7 Structural Self-Interaction Matrix (SSIM)
Full size table

The initial reachability matrix developed, as shown in Table 8. Note that a barrier affects itself so, the diagonals will have value as 1.

Table 8 Initial reachability matrix
Full size table

Transitivity is incorporated to make the expert opinion obtained on the contextual relationship consistent as per step 4. The transitivity matrix is obtained by replacing any inconsistency by 1* in an initial reachability matrix, as shown in Table 9.

Table 9 Transitivity matrix
Full size table

Hence, the final reachability matrix is shown in Table 10. Table 10: Final Reachability Matrix contains pairwise relationships among barriers along with some inferred entries.

Table 10 Final reachability matrix
Full size table

3.2.4 Level partitions

Reachability, antecedent, and interaction sets are obtained for each barrier from the final reachability matrix. The reachability set of a specific barrier consists of all the barriers in j columns with cell (i,j) value of 1. The antecedent set of a particular barrier includes all the barrier in i rows with cell (i,j) value of 1. common barriers between the reachability and antecedent set determines the intersection set. Subsequently, the barriers for which the reachability and the intersection sets are the same acquire higher levels in the ISM hierarchy. It signifies that the particular barrier at the highest level will not lead to the other barriers above their level in the ISM model. Once it is identified, the top-level barrier is removed from the other remaining variables [67]. The same iterative process is repeated until the level of each barrier is found. The ISM model is developed using these identified levels. In this paper, it took ten iterations for the level identification process of 16 barriers.

Table 11 shows the level partition wherein barriers 9, 11, and 14 are grouped at the level I. Subsequent levels are determined through further iterations. Detailed level partitions from iteration I to iteration X are tabulated in Appendix A (Tables 21, 22, 23, 24, 25, 26, 27, 28, 29, and 30).

Table 11 Level partition table: Iteration I–Iteration X
Full size table

3.2.5 Conical matrix

After obtaining the level of each barrier, a conical matrix is developed by rearranging barriers of the final reachability matrix at the same level across rows and columns. The value of driving and dependence power of the barriers is the sum of once in respective rows and columns [83, 84]. Next, the barriers are ranked on the driver and dependence power scale. Rank 1 is given to barriers with the highest value for respective power, i.e., dependence and driver.

3.2.6 Digraph and ISM model

An initial digraph with transitive links is drawn from the conical form (Table 12) of the reachability matrix. It is used to represent the barriers and their interdependencies by nodes and lines of edges. In one way, a digraph is a visual representation of all the barriers and their interdependencies [85]. After removing transitivity, a final digraph is obtained. Barriers at the level I are kept at the top of the hierarchy, meaning they won't lead to any other barrier. The barriers having the same level are grouped on the same level of the hierarchy.

Table 12 Conical matrix
Full size table

Finally, a digraph is changed into an ISM model (Fig. 4) by replacing nodes of the barriers with statements. The model depicts how various barriers impact the implementation of data analytics. It shows that lack of commitment from top management being the most critical barrier (based on high driving power) cause in the implementation of data analytics in the manufacturing industry.

Fig. 4
figure 4

ISM based model for barriers

Full size image

3.3 MICMAC analysis

MICMAC (Cross-impact matrix multiplication applied to classification) is based on the multiplication properties of matrices [86]. The purpose of MICMAC analysis is to identify and analyze the key barriers that drive the model. For this purpose, all the barriers are categorized into four quadrants (Fig. 5) according to their dependence and driving power: Independent, Linkages, Autonomous, and Dependent [87, 88]. The first quadrant is of the barrier with weak driving and weak dependence power. These barriers are relatively disconnected from the system; hence they are called autonomous or excluded barriers. The second quadrant comprises those barriers which have weak driving power but strong dependence power. They are called dependent barriers. The third quadrant is of those barriers which have strong driving as well as dependence power. These barriers are unstable compared to others; hence, any effect on them leads to an impact on other barriers. They are called linkage variables. The fourth quadrant barriers are those that have strong driving power but weak dependence power. They are called independent barriers. The position coordinates of barriers, as derived from the reachability matrix, as shown in Table 13.

Fig. 5
figure 5

MICMAC Analysis

Full size image
Table 13 Position coordinates of identified barriers
Full size table

4 Results and discussion on the outcome of ISM-MICMAC

This paper identifies sixteen barriers that create hindrances in the implementation of data analytics in the manufacturing industry from the literature review. ISM Modelling is used to structure and analyze these barriers. Figure 4 depicts the 16 barriers into ten levels through the ISM model. From the values of Table 13, a driving-dependence power graph (Fig. 5) is obtained from MICMAC analysis. MICMAC gives insights into the interdependencies among the 16 barriers. Significant findings of this study are summarized into four clusters:


First quadrant (Q-I): This is an autonomous quadrant. Barriers in this quadrant have weak driving and weak dependence power. These factors are generally disconnected and rarely influence the system. From the result of this study, poor data quality (DRB3) and data security & privacy (DRB1) appear in this quadrant, respectively, at levels III and IV in the ISM model. They may be essential but, compared to other barriers, are least connected with the system. These barriers are shown not to be against the implementation of data analytics. Data quality plays a critical role in Big data analytical techniques for extracting Hindsight, Insight, and Foresight [22, 23, 89].


Second quadrant (Q-II): This is a dependent quadrant. The barriers in this quadrant have weak driving power but strong dependence power. According to this study, the following barriers appear in this quadrant viz. time constraints (ORB1), no policy to share data among organization (ORB2), lack of awareness about data analytics (KSRB2), performance & scalability (DRB2), and complexity of data integration (DRB4). The absence of policy to share data among various departments in an organization creates a major hurdle in any data analytics project. It is also coupled with the complexity of data integration at the level I and II in the ISM model, which creates resistance to the implementation of data analytics.


Third quadrant (Q-III): This is a linkage quadrant. The barriers in this quadrant have both strong driving and dependence power. In this study, lack of management initiative (MRB2) and high cost of investment (FRB1) appear in this quadrant. Both of these barriers are positioned at level I in the ISM model. They play an important role in the implementation of data analytics in the manufacturing industry.


Fourth quadrant (Q-IV): This is an independent quadrant. Barriers in this quadrant have strong driving power and weak dependence power. As per this study, the following barriers appear in this quadrant: lack of infrastructure facility (ITRB1), unavailability of specific data tools (ITRB2), lack of training facility (ITRB3), lack of long term vision (MRB1), lack of commitment from top management (MRB3) and lack of funding (FRB2). Due to high driving power, these are the key barriers that impact the implementation of data analytics projects. All three barriers related to information technology positioned at level VII, and VIII in the ISM model, appear in this quadrant. While management-related barriers (MRB1, MRB3) are at level IX and X in the ISM model, it shows that targeting them can lead to the successful implementation of data analytics.

Manufacturing firms to implement BDA may require an expert to use, which is barrier 6 “Lack of skilled IT personal.”. Does it mean that hiring skilled IT personal will solve the problem of BDA implementation? The answer may be “NO.” Even if skilled IT personal are hired, the pressure will mount on “Lack of Availability of Specific Data Tools (barrier 2) and “Lack of Training Facility (barrier 3) because these are linked and have higher driving power thus need to ensure their availability else the implementation may face pressure.

Similarly, merely thinking about developing IT infrastructure facilities (barrier 1) may not help if sufficient funding is not available, and most importantly, the management must have the vision (barrier 8) and commitments (barrier 10). Management commitments and long-term vision have the highest driving power and can drive to remove all the obstacles in implementing BDA in manufacturing.

4.1 Prioritization of drivers using Fuzzy AHP

The ISM-MICMAC analysis produced six barriers into the fourth quadrant. The indicators grouped in this quadrant are most important in terms of drivers and independence. These are the indicators that drive all other indicators. In this section, our motive is to prioritize these indicators and find the most prominent barrier. Subsequently, it will also help in establishing the relative criticality of the obtained critical barriers (high driving power) through the ISM-MICMAC model. These barriers will help management in framing the BDA strategies.

Fuzzy MCDM refers to the fuzzy objective requirement and fuzzy decision making [90, 91]. The steps of fuzzy AHP for finding out the weights of each criterion are as follows [92].


Step 1: Make a pairwise comparison matrix and determine the fuzzy triangular values using standard fuzzification as in Table 14.

Table 14 Fuzzy triangular nos for criteria
Full size table

Pairwise comparison matrix: Five experts from the manufacturing industry were asked to evaluate the barriers on [93, 94] comparative scale. The combined outcome is shown in Table 15.

Table 15 Expert relative opinion on the barriers
Full size table

Consistency indices, CI, RI, and CR values, respectively obtained, are 0.048, 1.24, and 0.039. These indices indicate responses received are consistent.

The standard triangulated fuzzy pairwise comparison matrix is created by converting expert opinion in Table 15 using the standard triangulated criteria in Table 14. The resulting Fuzzy matrix is shown in Table 16.

Table 16 Fuzzy pairwise matrix
Full size table

Step 2: Calculated Fuzzy geometrical mean value (ri):

Obtained by fuzzy number multiplication: \(A_{1} \times A_{2} = \left( {l_{1} ,m_{1} ,u_{1} } \right) \times \left( {l_{2} ,m_{2} ,u_{2} } \right) = \left( {l_{1} *l_{2} ,m_{1} *m_{2} ,u_{1} *u_{2} } \right)\)

$$r_{i} = \left( {l_{i1} *l_{i2} \cdots l_{in} } \right)^{{{\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 n}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$n$}}}} , \left( {m_{i1} *m_{i2} \cdots m_{in} } \right)^{{{\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 n}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$n$}}}} ,\left( {u_{i1} *u_{i2} \cdots u_{in} } \right)^{{{\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 n}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$n$}}}}$$

where l, m, and u represents a lower, middle, and upper fuzzy number. n represents the number of columns.

Ex. For the first row: geometrical mean would be.

r1 = [{(1*2*1/4*1/6*1/8*1/4)^(1/6)},{(1*3*1/3*1/5*1/7*1/3)^(1/6)}, {(1*4*1/2*1/4*1/6*1/2)^(1/6)}].

The computed ri values for the six high impact barriers are shown in Table 17.

Table 17 Fuzzy geometrical mean
Full size table

Step 3: Evaluate Fuzzy weights

The formula used for getting weight is:

$$w_{i} = r_{i} *\left( {r_{1} + r_{2} + \cdots + r_{m} } \right)^{ - 1}$$

where “m” represents the number of rows.

$$\left( {r_{1} + r_{2} + r_{3} + r_{4} + r_{5} + r_{6} } \right) = \, \left( {{7}.{243},{ 8}.{8234},{ 1}0.{54}} \right)$$
$$\left( {r_{1} + r_{2} + \cdots + r_{m} } \right)^{ - 1} = \, \left( {0.0{95}, \, 0.{1133}, \, 0.{138}} \right)$$

Fuzzy weights obtained using the formula for \(w_{i}\) are displayed in Table 18.

Table 18 Fuzzy weights
Full size table

Step 4: De-Fuzzification

Now, De-Fuzzify the number using the center of area method (COA) using the expression given below

$$w_{i} = \left( {\frac{l + m + u}{3}} \right)$$

Fuzzy weights in Table 18 are Defuzzified as in Table 19, using the above expression.

Table 19 De-Fuzzify weights
Full size table

Step 5: Normalized weights

The De-fuzzified weights obtained as in Table 19 are normalized in 0–1 and 0–100% scales as in Table 20.

Table 20 Normalized weights
Full size table

Therefore, the priority of the independent barriers are as follows: MRB3 > MRB1 > FRB2 > ITRB3 > ITRB1 > ITRB2.

4.2 Consistency check for FAHP

A modified method for calculating λmax (relative weight), consistency index (CI), and consistency ratio(CR) is developed.

λmax = average (λlmax, λmmax, λumax); where l, m and u are suffix representing matrix formed by lower, medium, and upper values respectively.

λmax= average (5.147, 6.231, 7.60) = 6.327

CI = (λmax-n)/(n−1); n= matrix order = 6

CI = (6.327−6)/(6−1) = 0.0654;

CR = CI/RI, RI for corresponding “n=6” is 1.24 [93].

Condition for consistency: CR<0.1

CR = 0.0654/1.24 = 0.0527 <0.1 hence consistent.

5 Conclusion

Manufacturing firms in emerging economies are eager to implement BDA techniques and tools to remain competitive and maintain sustained firms' operations and performance. The adoption of BDA technology is at a nascent stage in manufacturing industries. The Indian manufacturing firms, in particular, are facing barriers to the implementation of BDA. Therefore, this paper is instrumental to the BDA literature concerning the manufacturing landscape by evaluating the importance of each barrier by using the ISM-MICMAC and Fuzzy AHP analytical approaches.

Sixteen barriers were obtained through literature study, and subject-related experts categorized the barriers into six groups. These barriers were analyzed using ISM. Barriers were structured in the ISM model and ranked based on their driving and dependence power to understand where they lie in MICMAC analysis. The research concludes that Management, Infrastructure & Technology related barriers are critical barriers (High driving power) to BDA implementation. Six sub-barriers, namely lack of commitment from top management (MRB3), lack of long term vision (MRB1), lack of funding (FRB2), lack of infrastructure facility (ITRB1), lack of availability of specific data tools (ITRB2), and lack of training facilities (ITRB3), are the critical barriers in driving for the implementation of BDA in the manufacturing industry. Lack of commitment from top management (MRB3), lack of long-term vision (MRB1) are management-related and are at Xth and IXth level in the ISM model. These levels depict the highest driving power. Lack of funding (FRB2) is related to finance while, lack of infrastructure facility (ITRB1), lack of availability of specific data tools (ITRB2), and lack of training facilities (ITRB3) are related to IT infrastructure & Technology.

Six sub-barriers are further investigated using Fuzzy-AHP for finding their relative criticality. The Sub-barrier lack of commitment from top management (MRB3) is the most critical driver, which confirms the outcome of the ISM model. The top two critical Drivers are management-related.

A similar study done in Bangladesh's manufacturing supply chain found IT infrastructure as a vital barrier followed by data privacy [14]. A study on the green lean six sigma implementation issue found that IT infrastructure and management are prominent factors [76], while Knowledge of Technology was a critical barrier to m-commerce adaptation SME’s in the UK [77]. Our findings are supported by the study done previously in different domains and geography, at least on critical barriers; Management and Infrastructure & Technology. This paper will help decision-makers in manufacturing companies formulate strategies related to BDA in their respective firms.

5.1 ISM MICMAC for academia

In this era of digitization, the absorbing power of students and researchers has increased tremendously. Thus, academic professionals can use this technique to figure out an active classroom pedagogical approach for application-based teaching to be better equipped to handle real-life scenarios in the business context. Work on trapezoidal fuzzification version of ISM-MICMAC may be developed and tested for its accuracy and reliability.

5.2 ISM MICMAC for industry

This methodology has varied applications across the industry. For example, it can be used for long-range planning. This methodology can also be used for designing processes, planning of career, strategic planning, solution of complex engineering enablers, design of products, re-engineering, financial decision making, Strategic human resource management (HRM), and e-commerce.

6 Limitations and future scope

This paper shows that 16 barriers have been identified, posing a hindrance to the implementation of data analytics in the manufacturing industry. Using ISM and MICMAC analysis, a model is developed. First, enough literature related to the topic of this study is not available. For this paper, literature related to sustainable supply chain management, reverse logistics, and remanufacturing is found, but there are hardly any research papers focused on India attempting linkages between barriers that impact the implementation of data analytics in the manufacturing sector. The barriers are identified through the literature review of journals and discussed with subject matter experts from the industry. The present study is entirely based on the subjective judgments of experts from the industry. This study can be expanded by identifying and incorporating the most significant barriers which are related to the implementation of data analytics in different sectors of the manufacturing firms in India. Some more factors like Data governance issues, Aligning BDA with corporate strategy, Ethical uses of BDA, Legislative and regulatory compliance, etc., can also be explored before finalizing the framework. The study was focused on the Indian manufacturing sector. The study may be extended to explore cross-country and cross-sector impact.