Big data analytics and the use of artificial intelligence in the services industry: a meta-analysis

ABSTRACT

Big data analytics have impacted nearly every service industry in the last decade. Furthermore,using artificial intelligence in big data analytics has introduced a new trend, resulting in different performance types, e.g. sales, marketing, innovation, organizational, financial, and operational. A systematic review of the empirical results from publications addressing big data analytics in the services industry becomes necessary to understand these performances better. Based on this rationale, this study conducted a meta-analysis to identify the relevant dimensions of big data analytics and evaluate artificial intelligence as a potential moderator of its effects on service performance. The results demonstrate that environmental dynamism, resources and capabilities, and competitive pressure drive big data analytics adoption. Environmental dynamism, followed by resources and capabilities, has greater effects on adopting big data analytics. The findings suggest that adopting big data analytics powered by artificial intelligence enhances service performance more than adopting big data analytics without using artificial intelligence.

在过去的十年里，大数据分析几乎影响了每一个服务行业。此外，人工智能在大数据分析中的使用成为了一种新趋势， 从不同方面影响了企业服务绩效，如销售、市场营销、创新、组织、财务和运营绩效。因此，我们认为有必要去系统性地回顾涉及服务行业中大数据分析的实证结果，以更好地理解这些影响。据此，本研究进行了元分析，以确定大数据分析的不同维度，并评估了人工智能作为其对服务绩效影响的潜在调节作用。结果表明，环境动态、资源与能力以及竞争压力推动了大数据分析的采用。其中，环境动态性，其次是资源与能力，对大数据分析的采用有更大的影响。研究发现，采用由人工智能驱动的大数据分析比未使用人工智能的大数据分析更能提高服务绩效。

KEYWORDS:

• Big data analytics
• artificial intelligence
• environmental dynamism
• competitive pressure
• service performance

关键词:

• ：大数据分析，人工智能，环境动态性，竞争压力，服务绩效。

1. Introduction

The emergence of big data has created opportunities for the services industry due to the growth of mobile technologies, social networking, search engines, e-commerce, prediction algorithms, and other technologies associated with the digital economy (Ciampi et al., 2021; Maroufkhani et al., 2020). Managing big data analytics is one of the most challenging tasks in the data era (Bag, Gupta, et al., 2021; Dubey et al., 2019; Huang et al., 2022; Yaşar et al., 2024). Understanding information management for decision-making due to the large volume of data has become one of the most valuable assets for the services industry (Acharya et al., 2018; Bag et al., 2023). In recent years, extensive research on the impacts of big data analytics has been available in the services literature (Auh et al., 2022; Belk et al., 2023; Blanche et al., 2024). This research discusses the advantages (Belk et al., 2023; Samara et al., 2022) and disadvantages (Belanche et al., 2024) of using analytics in service companies. Although the current services literature provides evidence of the importance of using analytics, little is known about how and when analytics translates into firm performance (Auh et al., 2022). Furthermore, there is a need to integrate analytical techniques (descriptive, predictive, and prescriptive) to understand their effect on various stages of service provision, e.g. the service customer journey and touchpoints) (Auh et al., 2022; Dam et al., 2019). A set of recent studies demonstrates the need to systematize (Samara et al., 2020), refine (Marian et al., 2018 and integrate (Dam et al., 2019) current literature on services to gain a better understanding of customers in the service area.

Several service industries began to use big data analytics and artificial intelligence amid this dispersion of results. For example, in Gunasekaran et al.’s (2017) study, big data analytics influences organizational performance. In Dubey et al.’s (2021) study, big data analytics influences organizational and financial performance. Sommanawat et al. (2019) demonstrate that big data analytics influences organizational performance. These discrepancies in performance types demonstrate that artificial intelligence is a prominent issue in today’s big data analytics (Bag et al., 2023; Paschen et al., 2019).

The complexity of data, several new technologies, and the search for efficiency and effectiveness have caused the service industry to invest more in big data analytics through computers and machines that can reason, learn, and act in ways that normally require human intelligence (Akter et al., 2021; Bag et al., 2023; Dubey et al., 2022).

Overall, our study aims to analyze the effects of big data analytics on service performance. More specifically, this study investigates three research questions associated with the dispersion of results in big data analytics and the use of artificial intelligence in the services industry: (1) What is the influence of environmental dynamism, resources and capabilities, and competitive pressure on adopting big data analytics? (2) What are effect sizes that measure big data analytics as drivers of service performance? and (3) What is the effect of using artificial intelligence in big data analytics? We developed a theory-based model using data from preliminary papers estimating the parameters through Meta-Analytic Structural Equation Modeling (MASEM) to answer the first and second questions. To answer the third question, we analyzed the heterogeneity of the relationship between big data analytics and service industry performance through the AI-driven versus no AI-driven moderator by estimating the parameters through hierarchical linear meta-analysis (HiLMA).

This study aims to contribute to the services field by addressing several research questions. Firstly, it responds to recent calls for research on consolidating big data analytics findings through a meta-analysis (Ansari & Ghasemaghaei, 2023; Oesterreich et al., 2022). Previous meta-analytic studies have underscored the importance of studying big data analytics and artificial intelligence (Ansari & Ghasemaghaei, 2023; Blut et al., 2021). This meta-analysis distinguishes itself academically by presenting a model that interprets the effects of big data analytics in services. Recent studies in the service sector have emphasized the necessity for academic work that aims to consolidate, synthesize, and refine findings to guide future agendas for technology adoption in the service sector (Auh et al., 2022; Belanche et al., 2024; Belk et al., 2023). Understanding the effects of big data analytics presents an opportunity to enhance customer value in the service sector (Belk et al., 2023). Longitudinal studies tend to increase confidence in analytical results regarding higher firm performance (Auh et al., 2022). This meta-analysis applies an integrated framework to comprehend the state of the art, the application of analytical techniques, and their potential impacts on performance.

Secondly, big data analytics is a prominent topic in academic and practitioner discussions in the services industry. Several theories, such as the resource-based view, dynamic capability view, technology-organization-environment framework, diffusion of innovation theory, institutional theory, and organizational learning theory, among others, are employed to understand big data analytics adoption. However, these theories are scattered across primary papers. Our meta-analysis consolidates these theories within a singular model to understand the differences in using big data analytics with or without artificial intelligence.

Thirdly, big data analytics implementations have been undertaken in various countries, services, and contexts using diverse approaches. Consequently, different perspectives on big data analytics and conflicting results emerge from antecedents and consequents. This diversity leads to heterogeneity in findings, increasing levels of heterogeneity. We endeavor to mitigate the heterogeneity of effect sizes from previous studies using moderators.

Finally, the interaction between big data and artificial intelligence to enhance service industry performance warrants further investigation. There are instances of big data analytics powered by artificial intelligence (BDA-AI) (Bag et al., 2023; Paschen et al., 2019) and big data analytics without the use of artificial intelligence (BDA) (Ciampi et al., 2021; Maroufkhani et al., 2020) affecting service performance. However, an aggregate analysis of the effects of employing or not employing artificial intelligence on service industry performance is still necessary. Through meta-analytical analysis of previous research, we classify the use of BDA and BDA-AI, revealing that incorporating AI in big data analytics tends to yield greater service performance than without AI.

To answer the three research questions, this article first shows the theoretical model of the relationship between big data analytics and the use of artificial intelligence and its possible relations with the hypotheses. Second, the data collection for the meta-analysis process is outlined, followed by the interpretation. Data was collected in the Scopus Platform using the keywords ‘big data analytics’ and ‘big data’ in the title and/or summary or keywords or abstract. We emphasized papers that involved service focus in this set of studies. The initial data collection resulted in 2,139 published studies. We excluded 2,057 articles from the database if they were qualitative studies, literature reviews, or articles without statistical data or insufficient statistics for converting effect sizes. The number of articles selected for analysis containing statistical correlation data or with possible conversions was 82. Meta-analytic structural equation modeling (MASEM) and hierarchical linear meta-analysis (HiLMA) were the analytical techniques used. Based on the findings, the paper discusses the implications for practice and further service industry research.

2. Proposed framework for big data analytics in the service industry

Big data analytics provides a technique widely used in today’s service industry to process inherently voluminous and complex data with high ‘velocity and variety’ (Duan & Xiong, 2015; Gupta et al., 2020). From a simplistic perspective, big data analytics can be understood as a technique for unwinding voluminous data for decision-making and better company performance. The complex nature of big data has recently given rise to tools that assist in data analysis, such as artificial intelligence (Bag et al., 2023; Paschen et al., 2019). Service companies have used artificial intelligence as a powerful tool to understand the complexity of data (Gupta et al., 2020). Adopting big data analytics generates more efficiency and effectiveness in understanding consumers and the market, trying to increase service industry performance (Bag et al., 2023; Dubey et al., 2022; Gupta et al., 2020).

Recent studies in the service sector have discussed the importance of employing technologies linked to big data analytics (Auh et al., 2022; Belk et al., 2023; Dam et al., 2019). Auh et al. (2022) emphasize the significance of understanding analytics and their direct impact on service performance. The authors note that despite the importance of analytics for service companies, there is currently little debate about how much a service company is prepared to incorporate an analytics strategy and how much this translates into better performance. From an optimistic perspective, Belk et al. (2023) assert that these new technologies provide a better understanding of customers in the service sector. In their view, new technologies represent an opportunity to enhance customer value in the service sector. Therefore, recent studies should aim to understand these technologies’ distinctive features and challenges in services marketing (Belk et al., 2023). Similarly, Dam et al. (2019) demonstrate that the service sector must integrate analytical techniques, including descriptive, predictive, and prescriptive approaches, to attempt to understand the real effect of using these technologies. Samara et al. (2020) point out that using these technologies tends to increase strategies’ efficiency, productivity, and profitability. From a pessimistic perspective, Blanche et al. (2024) demonstrate the negative side of these technologies for the service sector. Analytics are strategically important for service firms as they improve personalization, prioritization, and cost minimization levels. On the other hand, critical aspects related to the implementation of technologies in services need to be explored further as they still reflect on managers’ concerns (e.g. loss of privacy and lack of emotion and empathy, among others) (Belanche et al., 2024).

Discussions regarding the negative and positive aspects of using technologies such as analytics and artificial intelligence have demonstrated the importance of developing more robust models for understanding the effects on service performance (Marian et al., 2018; Dam et al., 2019; Marian et al., 2018). Rialti et al. (2019) report that although numerous scientific publications regarding big data and dynamic capabilities exist, this literature needs systematization. Marian et al. (2018) describe the importance of refining studies to understand the effect of analytics on services better. Dam et al. (2019) assert that analytical techniques must be integrated to understand the use of technologies in services. Despite using different theoretical lenses to understand the effects of analytics on service performance, these studies converge in indicating the need to propose models that understand the complexity of using these technologies in service companies. Our theoretical model is founded on three elements that can be viewed within a triad: big data analytics, artificial intelligence, and service industry performance. Figure 1 shows the meta-analytic framework in which big data analytics in the service industry is the focal construct.

Figure 1. Theoretical model.

2.1. Drivers of big data analytics

We consider the variables of environmental dynamism, resources and capabilities, and competitive pressure as drivers of big data analytics. Environmental dynamism represents the volatility and unpredictability of the firm’s external environment, an important factor for Dynamic Capabilities theory (Miller & Friesen, 1983; Wamba et al., 2020). This concept is a principal driver in adopting big data analytics as it pressures organizations, making them seek new knowledge and innovation (Wamba et al., 2020).

Dynamism is understood in the services literature as an unpredictability generated by the uncertainty of customer actions or the industry’s rate of change and innovation (Li & Liu, 2014). Dynamic environments happen when there is a high rate of changes in technologies and variations in customer preferences (Jansen et al., 2009). The changes brought about by environmental dynamism directly influence the use of technologies in the services industry. This is because transformations with the changing environmental dynamism generally drive the implementation of operational processes such as big data analytics to meet industry demands, generating more competitive advantages (Bag et al., 2023). Environmental dynamism has become an important factor in service industries across the globe for implementing big data analytics (Bag et al., 2023; Wamba et al., 2020). Therefore, it is expected that:

H1: Environmental dynamism positively influences the adoption of big data analytics.

Resources involve human skills and knowledge that can be identified and evaluated, and capabilities are described as high-level routines (or a collection of routines) consisting of learned behaviors. Resources and capabilities become important drivers for adopting big data analytics because they can be mobilized to achieve competitive performance (Mikalef et al., 2018).

Managing capabilities and resources is necessary for a company to balance operational and economic performance and expand its growth (Gupta et al., 2020). Big data analytics implementations depend on a firm’s capabilities and resources, expressed in human skills (Gupta & George, 2016), i.e. technical and managerial skills (Ciampi et al., 2021). Technical and managerial skills summarize expertise/overall understanding of specific technology and talent-hunting acquisition to generate an organization’s knowledge culture. Investments in technical and managerial skills tend to increase success in big data analytics processes (Gupta et al., 2020). Therefore, we predict that:

H2: Resources and capabilities positively influence the adoption of big data analytics.

Competitive pressure refers to influences from the competitive environment that maintain or increase competitiveness by generating environmental pressures and is considered an important driver in the adoption of technologies (Chen et al., 2015). Competitive pressure is associated with the organization’s perceived pressure from technological advancements to maintain competitive advantage (Chen et al., 2015; Zhang et al., 2020). This pressure is driven by influences from the external environment that may influence the adoption of big data analytics (Maroufkhani et al., 2020). Competitive pressure can influence customers, suppliers, and competitors (Chen et al., 2015). In highly competitive scenarios, companies tend to mimic the behaviors of other companies (Zhang et al., 2020). In an environment where the firm’s data-driven culture prevails, adopting big data analytics is somewhat driven by competitive pressure (Maroufkhani et al., 2020; Zhang et al., 2020). Therefore, we can expect that competitive pressure is a big data analytics driver.

H3: Competitive pressure positively influences the adoption of big data analytics.

2.2. The relationship between big data analytics and service performance

Big data analytics is essential for precision decision-making and optimal performance (Müller et al., 2018) and represents a set of new techniques that seek to extract hidden patterns from information for making correct decisions in the services industry (Bag, Gupta, et al., 2021). The influence on a firm’s performance can be demonstrated in different ways, such as reducing the cost of processes, increasing productivity, generating more innovation and more accurate decisions, and upgrading knowledge (Acharya et al., 2018; Maroufkhani et al., 2020). Literature in services has demonstrated that big data analytics has transformed how businesses compete and can affect different forms of performance, including operational, financial, and marketing (Bag, Pretorius, et al., 2021). Therefore, it can be expected that:

H4: The adoption of big data analytics positively influences service performance.

2.3. The triad of big data analytics, artificial intelligence, and service performance

Several previous studies have studied the triad of adopting big data analytics, artificial intelligence, and performance, helping to understand how companies organize their resources today (Bag, Pretorius, et al., 2021; Chintalapati & Pandey, 2022). This adoption is associated with using technology to process and analyze a large set of data to extract insights and value through advanced analytical methods (Chen & Zhang, 2014; Wamba et al., 2017). Artificial intelligence has been used as a complement in adopting big data analytics (Bag et al., 2023), offering a way for firms to understand the market better and have better competitive performances (Bag et al., 2023; Dubey et al., 2019). The combination of big data analytics with artificial intelligence tends to improve service performance. The performance measure is a dependent variable used in several services studies that measures a company’s ability to achieve results and to stand out from the competition (Gremler et al., 2020; Müller et al., 2018; Santini et al., 2020). Furthermore, we indicate that big data analytics help increase service performance more in environments with artificial intelligence than in environments without artificial intelligence.

Practical examples have demonstrated the impact of big data analytics and artificial intelligence on service performance (Bag et al., 2023; Benzidia et al., 2021; Dubey et al., 2020). For example, technology-based collaborative platforms that integrate big data analytics and artificial intelligence have improved the delivery of medicines in South Africa by refining different performance indicators such as right time, right place, and correct quantity (Bag et al., 2023). The adoption of big data analytics with machine learning and artificial intelligence has increased operational performance in firms in India (Dubey et al., 2020). Meanwhile, using big data analytics and artificial intelligence has enhanced environmental performance in French hospitals (Benzidia et al., 2021). In the world of uncertainties, the use of artificial intelligence applied to big data analytics has emerged as one of the important phenomena in increasing performance (Bag, Pretorius, et al., 2021; Bag et al., 2023; Chintalapati & Pandey, 2022).

Big data analytics can be adopted differently within a firm (Dubey et al., 2020; Lutfi et al., 2023; Maroufkhani et al., 2020). Studies that analyze the performance of big data analytics have analyzed implementation projects without the use of artificial intelligence (BDA) (e.g. Lutfi et al., 2023; Maroufkhani et al., 2020) and powered by artificial intelligence (BDA-AI) (e.g. Benzidia et al., 2021; Dubey et al., 2020). The connection between big data analytics and artificial intelligence has been demonstrated to benefit the services industry (Paschen et al., 2019). Artificial intelligence helps practitioners better understand the modus operandi of competitors, consumers, and the market and have better competitive performances (Bag et al., 2023; Dubey et al., 2019). This is because the set of technologies associated with artificial intelligence allows computers to process and perform various advanced functions, including analyzing data and making more accurate recommendations (Chintalapati & Pandey, 2022). Artificial intelligence is considered a disruptive technology with immense marketing transformation potential for managers working with data analysis (Verma et al., 2021). Artificial intelligence in the services area has improved operations and ensured the accelerated success of many firms (Chintalapati & Pandey, 2022).

Artificial intelligence acts in knowledge management, a key technological enabler in this digital age (Dubey et al., 2019). The complexity of dealing with a large amount of data has driven different service companies (e.g. collaborative platforms, healthcare, retail, and banking, among others) to use artificial intelligence (Bag et al., 2023; Huang et al., 2022). With the help of artificial intelligence, big data analysis can enhance the efficacy of strategies (Bag, Gupta, et al., 2021), and by evaluating that big data, sometimes from various sources, artificial intelligence tends to help practitioners understand the market better and have better competitive performances (Bag et al., 2023; Dubey et al., 2019). Therefore, big data analytics driven by artificial intelligence tends to generate greater service industry performance than big data analytics not driven by artificial intelligence.

H5: Adopting big data analytics without using artificial intelligence, as compared to adopting big data analytics powered by artificial intelligence, will weaken the direct effects of service performance.

3. Method

A meta-analytic process was built to collect and analyze data from previous studies. The search for papers was conducted through the Scopus Platform, which is highly recommended for its vast coverage of articles (Lim et al., 2022; Lim & Rasul, 2022). We used the keywords ‘big data analytics’ and ‘big data’ in the title and/or summary, keywords, or abstract published in the Business, Management, and Accounting field. We emphasized papers that involved service focus in this set of studies. This procedure is commonly used in meta-analysis of the technology and management area (e.g. Bergmann et al., 2023; Ladeira et al., 2023; Santini et al., 2020; Santini, Buhler, et al., 2023; Santini, Lim, et al., 2023). The initial data collection was conducted in October 2023, resulting in 2,139 published studies. We excluded articles from this database that included qualitative data and literature reviews. Furthermore, our meta-analysis excluded articles without statistical data or insufficient statistics for converting effect sizes, such as ARIMA or econometric equations. We intended to extract the Pearson correlation from the previous paper to build the meta-analytic model based on five variables representing our model: environmental dynamism, resources and capabilities, competitive pressure, big data analytics, and service performance. After the exclusions, the final sample consisted of 82 articles generating 3,042 effect sizes with a sample of 30,783 participants. However, 133 effect sizes from these 83 papers (Web Appendix A) were used as they contain information about the constructs of our model.

We followed a procedure Rust and Cooil (1994) suggested in the codification process. Two researchers with a background in marketing performed the data encoding. The researchers carried out the codification individually. Before beginning this activity, the analytical criteria were extensively discussed among the authors of this study. After the codification process, we performed the data analysis. In this case, we followed suggestions made in previous meta-analytic studies (e.g. Babić Rosario et al., 2016; Hedges & Olkin, 1985). We measured effect size using Pearson’s correlation coefficient (r).

The theoretical model of antecedents and consequences was undertaken using the meta-analytic structural equation modeling (MASEM) technique. This process combines the techniques of meta-analysis and modeling of structural equations (Cheung, 2015). The MASEM technique performs new meta-analyses, providing effects that control other model variables (Bergh et al., 2016). Finally, the moderating relationships analysis was performed using a hierarchical linear meta-analysis (HiLMA), which uses a regression-based multivariate format (Geyskens et al., 2009) from the variables inserted in the model. This technique is also widely used in meta-analytic research (Babić Rosario et al., 2016; Santini et al., 2020).

3.1. Variables analyzed in the meta-analysis

Our selection of variables for the model was guided by their emergence during our coding process in the primary papers. Five constructs—environmental dynamism, resources and capabilities, competitive pressure, big data analytics, and service performance—generated the meta-analytic model with independent and dependent variables. Table 1 details the variables used in the meta-analysis.

Table 1. Variables used in the meta-analysis.

Variable	Description/Definition	(Proxy variable) Common Alliances	Operationalization/Coding Instructions	Examples/Applications
Environmental dynamism^a	This concept is a key factor in dynamic capability theory and tells the volatility and unpredictability of the firm’s external environment (Miller & Friesen, 1983; Wamba et al., 2020).	Environmental dynamism, industry dynamism, external factors dynamism	Constructs that measure different types of dynamism (e.g. environmental, industry, and external factors) as an item scale according to data found in the methods section of the papers (e.g. Wamba et al., 2020).	Environmental dynamism generates pressure on organizations and influences performance (Wamba et al., 2020).
Industry dynamism^b	Refers to constant changes in technological, socio-political, and environmental aspects in the region in which the business operates (Bag, Pretorius, et al., 2021; Dubey et al., 2018).	Industry dynamism	Constructs that measure industry dynamism as an item scale according to data found in the methods section of the papers (e.g. Bag, Pretorius, et al., 2021). We use 5 effect sizes, which are equivalent to 50% of the environmental dynamism effects sizes.	Industry dynamism affects uncertainty across industry segments and impacts performance (Dubey et al., 2018).
External factors dynamism^b	External factors that may affect a firm’s dynamic capabilities arise from external pressures from government agencies and consumer input, among others (Bag, Pretorius, et al., 2021; Li & Liu, 2014).	External factors dynamism	Constructs that measure external factors as an item scale according to data found in the methods section of the papers (e.g. Li & Liu, 2014). We use 3 effect sizes, which are equivalent to 30% of the environmental dynamism effects sizes.	Artificial intelligence could acquire learnings by analyzing the external environments (Belhadi et al., 2021).
Resources and capabilities^a	Resources and capabilities are the core components of resource-based theory (RBT). Theoretical groundings that group resources (human skills and knowledge) that can be identified and evaluated and capabilities are described as high-level routines (or a collection of routines), with routines consisting of learned behaviors (Mikalef et al., 2018).	Organizational capabilities, tangible resources (e.g. financial and physical resources), human skills (e.g. employees’ knowledge and skills), intangible (e.g. organizational culture and organizational learning)	Constructs that measure different types of resources and capabilities described as an item scale according to data found in the methods section of the papers (e.g. Dubey et al., 2019).	Resources and capabilities can be mobilized to achieve competitive performance gains in the adoption of big data analytics (Mikalef et al., 2018).
Managerial skills^b	Encompassed within human skills (II second-order construct), it can be considered the ability that mostly revolves around recognizing the value of big data and understanding where to apply insight efforts (Akter & Wamba, 2016; Mikalef et al., 2020).	Managerial skills	Formative order construct described as an item scale according to data found in the methods section of the papers (e.g. Ciampi et al., 2021). We use 3 effect sizes, which are equivalent to 10% of the effects sizes of resources & capabilities.	Competencies are highly firm-specific and developed over time by individuals being tacit and, thus, heterogeneously dispersed across firms (Dubey et al., 2019).
Technical skills^b	Encompassed within human skills (II second-order construct), it can be considered the ability to know when required to use new technology to extract relevant management insights from big data (Akter & Wamba, 2016; Mikalef et al., 2020).	Technical skills	Formative order construct described as an item scale according to data found in the methods section of the papers (e.g. Ciampi et al., 2021). We use 6 effect sizes, which are equivalent to 20% of the effects sizes of resources & capabilities.	Competencies in machine learning, data extraction, data cleaning, statistical analysis, and understanding of programming paradigms (Dubey et al., 2019).
Competitive pressure^a	Refers to influences from the competitive environment that maintain or increase competitiveness (Chen et al., 2015).	Competitive pressure, mimetic pressures, coercive pressures, normative pressures	Constructs that measure different types of competitive pressure described as an item scale according to data found in the methods section of the papers (e.g. Chen et al., 2015).	Environmental pressures can positively affect the degree of competitive pressure (Chen et al., 2015).
Firm size^c	Size (in number) of members in the organizational structure and employees (Santini et al., 2021).	Firm size	A dichotomous variable was created using: 0 = SMEs (N = 6) and 1 = Large enterprises (N = 9).	Sample size possible influence of scales used to measure performance (Santini et al., 2021).
Big data analytics^a	Refers to how to process and analyze big data sets with the aim of extracting insights and value through advanced analytical methods (Chen & Zhang, 2014; Wamba et al., 2017).	Big data analytics, big data predictive analytics, big data analytics capabilities, big data analytics adoption	Constructs that measure different types of big data analytics described as an item scale according to data found in the methods section of the papers (e.g. Wamba et al., 2017).	Big data analytics capability influences firm performance (Wamba et al., 2017).
Service performance^a	We classified service performance used on the primary studies; this information was extracted from each analyzed article.	Firm performance, firm sales performance, financial performance, marketing performance, operational performance, economic performance, organizational performance	Constructs that measure service performance as an item scale according to data found in the methods section of the papers.	Performance is a dependent variable used in several services studies and meta-analyses (Gremler et al., 2020; Müller et al., 2018; Santini et al., 2020).
AI-driven adoption^c	We separated the articles into two groups according to the adoption of big data analytics: the adoption of big data analytics powered by artificial intelligence (BDA-AI) (e.g. Benzidia et al., 2021; Dubey et al., 2020) and the adoption of big data analytics without the use of artificial intelligence (BDA) (e.g. Lutfi et al., 2023; Maroufkhani et al., 2020).	Big data analytics powered by artificial intelligence and big data analytics	The variable was coded in this order: 0 = big data analytics (BDA) (N = 48) and 1 = big data analytics powered by artificial intelligence (BDA-AI) (N = 12).	Artificial intelligence tends to help practitioners understand the market better and have better competitive performance (Bag et al., 2023; Dubey et al., 2019).
Type of industry^c	The variable indicating in which service industry the study was carried out was applied following suggestions from Santini et al. (2020) and Blut et al. (2021).	IT services, hotel and tourism services, health care services, general retail	The variable was coded in this order: 0 = general retail (N = 4), 1 = health care services (N = 4), 2 = hotel and tourism services (N = 3), and 3 = IT services (N = 7).	The type of industry where the research is carried out can be a moderator in studies in the service area (Blut et al., 2021).
Geographic region^c	The geographic region of the study was separated into two respective cultural orientations based on Hofstede’s parameters (2011) and Jadil et al. (2021). Cultural values showed differences between the Western (e.g. USA, France, Canada, Brazil, among others) and Eastern (e.g. India, Japan, Indonesia, China, among others) countries according to Hofstede (2011).	Name of countries and location	A variable was created using the location: 0 = Western (N = 20) and 1 = Eastern (N = 36).	Cultural values influence the organizational environment and the strategies adopted (Hofstede, 2011). Therefore, culture may lead to differences in direction and strength among effect sizes (Jadil et al., 2021).
Sample size^c	We defined the sample size based on previous studies (Santini et al., 2020) that indicated the division of the sample by the median into two groups: small (N = 30) or large (N = 30). The sample size was declared in each study. We adopt the median of the sample sizes as the cut-off point.	Sample size	A dichotomous variable was created using the median (267): 0 = large and 1 = small.	Small samples overestimate effects (Rosenthal, 1979).
Publication ranking^c	We identified the studies associated with the relationship between variables that were published in top journals or other journals. Studies published in top journals (ranked with 4 or 4 stars) versus studies published in other journals were based on the ABS journal ranking (Blut et al., 2021).	Publication ranking	The others were considered not to be top journals. Therefore, the sample was divided into 0 = non-top journals (N = 28) and 1 = top journals (N = 32).	Top journals tend to prioritize publications with overestimated effects (Rosenthal & Rubin, 1982).

Note: ^aVariables were examined in the meta-analytic structural equation modeling (MASEM), ^bin the random effects model, and ^cin the hierarchical linear meta-analysis (HiLMA).

We use the ‘proxy variable’ to define the five constructs of the MASEM. Meta-analytic studies have demonstrated that proxy variables can be computed from the coefficient of construct validity, providing a broader interpretation of study relationships (Hunter, 2004; Glasman & Albarracín, 2006). Environmental dynamism originated from common alliances found in dynamic capability theory that represent constructs that measure volatility and unpredictability in the environment of firms, such as industry and external factors dynamism (Miller & Friesen, 1983; Wamba et al., 2020). Non-exclusive constructs form the concept of environmental dynamism.

To better understand the concept of environmental dynamism, we use the random-effect model to analyze the effects of industry and external factors dynamism in big data analytics. The concept of resources and capabilities is a proxy variable representing two core components of Resource-Based Theory (RBT) (Mikalef et al., 2018). The constructs that form this variable were organizational capabilities, tangible resources (e.g. financial and physical resources), human skills (e.g. employees’ knowledge and skills), and intangible (e.g. organizational culture and organizational learning). By considering managerial and technical skills as non-exclusive constructs, we analyze their effects on big data analytics using the random-effect model.

Competitive pressure is a variable proxy that measures the influences from the competitive environment that maintain or increase competitiveness (Chen et al., 2015), being used as common alliance mimetic pressures, coercive pressures, and normative pressures. The big data analytics variable had as common alliances the constructs of big data predictive analytics, big data analytics capabilities, and big data analytics adoption. This variable refers to processing and analyzing big data sets to extract insights and value through advanced analytical methods (Chen & Zhang, 2014; Wamba et al., 2017).

Finally, our model presents service performance as a variable proxy formed by the common alliance of the concepts of firm performance, firm sales performance, financial performance, marketing performance, operational performance, economic performance, and organizational performance. Although non-exclusive constructs form the last three proxy variables, we did not use the random-effect model, as the constructs did not meet the minimum necessary effect size (N = 3) as suggested in seminal works (Hedges & Olkin, 1985; Hunter & Schmidt, 2004).

3.1.1. Moderator analysis

We conducted a meta-regression to understand the effect of potential moderators on the relationship between some variables in our model. The aim of analyzing the possible effects of moderators is to go beyond the bivariate correlations demonstrated in MASEM. We selected the following moderation variables: (1) in the relationship between competitive pressure and big data analytics, we used the firm size variable, and (2) in the relationship between big data analytics and service performance, we used the variables AI-driven versus no AI-driven, type of industry, geographic region, sample size, and publication ranking.

3.1.1.1. Firm-size

Firm size is an important moderator in the meta-analysis that studies the impacts on firm performance (Santini et al., 2021). In primary studies, we observed that the samples comprised employees who work in companies of different sizes. For example, Behl et al. (2022) analyzed the role of big data analytics capabilities to improve the sustainable competitive advantage of SME services. In contrast, Dubey et al. (2019) analyzed the effects of big data analytics on manufacturing performance in a sample in which more than 90% of companies had more than 100 employees. Due to this finding, we chose to create a dummy variable that indicated the size of the companies in SMEs and large enterprises. This information was collected in the methods section of the primary studies of the papers that provided the number of employees.

3.1.1.2. AI-driven adoption

The primary studies analyzed in this meta-analysis contained an analysis of the effects of adopting big data analytics with or without the application of artificial intelligence. For example, Zhu et al. (2021) analyzed big data and analytics implementation and its effects on firm performance. Behl et al. (2022) studied the effects of big data analytics capabilities to improve sustainable competitiveness. In these two studies, the existence of artificial intelligence usage in the adoption of big data analytics was not declared. In contrast, Gupta et al. (2020) studied the effects of big data predictive analytics on organizational performance. Dubey et al. (2020) interpret the effects of big data analytics on operational performance. These two studies mention using artificial intelligence in adopting big data analytics. Studies declare that adopting big data analytics powered by artificial intelligence (BDA-AI) (e.g. Benzidia et al., 2021; Dubey et al., 2020) enhances firms’ performance. Artificial intelligence tends to help practitioners understand the market better and have better competitive performances (Bag et al., 2023). Because of this finding, we separated the articles into two groups according to the adoption of big data analytics: big data analytics powered by artificial intelligence (BDA-AI) and big data analytics (BDA), and our moderator was created using a dummy variable based on information about the use or not of artificial intelligence in big data analytics found in the primary papers.

3.1.1.3. Type of industry

The type of industry existing in primary study samples is a factor that can interfere with the effect sizes of a meta-analysis (Crosno & Dahlstrom, 2008). Research demonstrates that amplifying data from different industries brings heterogeneous information due to the different contexts in which companies are inserted (Ohiomah et al., 2020). Understanding the effects across different types of industries can improve external validity and provide a reliable representative sample (Crosno & Dahlstrom, 2008). Our primary papers found four industries with sufficient effect sizes (N = 3) for the hierarchical linear meta-analysis calculations. In our primary papers, we found four types of industries that contained enough effect sizes (N = 3) for the hierarchical linear meta-analysis calculations (Santini et al., 2020). Thus, each study’s variable indicated which service industry was divided into IT services, hotel and tourism services, health care services, and general retail.

3.1.1.4. Geographic region

Creating moderating variables according to the countries where the sample data was collected is common in meta-analytic studies in management (Bitencourt et al., 2020; Jadil et al., 2021). Cultural values and habits in different regions influence the organizational environment and the strategies adopted (Hofstede, 2011). Studies have indicated that the cultural values existing in the countries of the interviewees can influence the effect sizes and be important in the attempt to understand heterogeneity (Bitencourt et al., 2020). This is because cultural aspects may lead to differences in direction and strength among effect sizes (Jadil et al., 2021). Our database finds different big data analytics applications in different countries. For example, Gupta et al. (2020) analyzed organizational performance via big data predictive analytics in a sample of Indian managers. Shah et al. (2017) analyzed the organizational change readiness of Pakistani employee attitudes and behaviors using big data. Bag, Pretorius, et al. (2021) analyzed knowledge creation and B2B marketing rational decision-making for improving firm performance using artificial intelligence across employees in South Africa. Ciampi et al. (2021) analyzed the relationship between big data analytics capabilities and business model innovation in UK firms. These examples demonstrate that relationships involving big data analytics and service performance were carried out in different countries. Due to this finding, our moderator was created using a dummy variable based on the location of the interviewee found in the methods section of the primary studies analyzed: Western (e.g. USA, France, Canada, and Brazil, among others) and Eastern (e.g. India, Japan, Indonesia, and China, among others). This division was guided by previous meta-analytic studies that used this division (Bitencourt et al., 2020; Jadil et al., 2021).

3.1.1.5. Sample size

Sample size has been considered an important methodological moderator in understanding heterogeneity in management meta-analysis (Santini et al., 2020). Preliminary studies have indicated that small samples, unlike large samples, tend to overestimate effects (Rosenthal, 1979). Meta-analytical studies have shown that larger effect sizes are found in small samples (Fern & Monroe, 1996; Jadil et al., 2021). The sample size was separated into two groups (small or large) based on the sample number declared in each study. We adopted the median of the sample sizes as the cut-off point (267). We expect that sample size will moderate the impact of big data analytics on service performance to the extent that a small sample size generates a stronger effect than a large sample size.

3.1.1.6. Publication ranking

The methodological moderator ‘publication ranking’ is important for understanding heterogeneities in meta-analyses (Bitencourt et al., 2020; Luceri et al., 2022). Previous studies have stated that the importance/relevance of the journal can impact the average types of effects found (Bitencourt et al., 2020). Publications in higher-impact journals have stronger effect sizes (Bitencourt et al., 2020; Luceri et al., 2022). Top journals prioritize publications with overestimated effects (Rosenthal & Rubin, 1982). Therefore, we created a dummy variable comparing studies published in top journals (ranked with 4 or 4 stars) versus those published in other journals based on the ABS journal ranking (Luceri et al., 2022). We expect publication ranking to moderate the impact of big data analytics on service performance to the extent that lower-impact journals generate a weaker effect than high-impact journals.

4. Results

The estimate of the mean range of corrected weighted correlations proved significant in the interaction of environmental dynamism, resources and capabilities, competitive pressure, big data analytics, and service industry performance. The correlation matrix generated by the effect sizes of the studies that involved service focus is presented in Table 2.

Table 2. Correlation matrix.

Constructs	Environmental dynamism	Resources and capabilities	Competitive pressure	Big data analytics
Resources and capabilities	.228*
Competitive pressure	.107*	.289*
Big data analytics	.23*	.481*	.304*
Service industry performance	.274*	.307*	.321*	.312*

Notes: *p < .01.

The direct relationships of big data analytics were estimated using MASEM. Table 3 presents the results of the four hypotheses analyzed. The model’s adjustment indices proved adequate (Comparative Fit Index = .895; Tucker-Lewis Index = .737; Root-Mean-Square Error of Approximation = .012) (Cheung & Chan, 2005). In addition to MASEM, we provided bivariate relationships with big data analytics (industry dynamism, external factors dynamism, managerial skills, technical skills, SMEs, large enterprises, IT services, health care services, hotel and tourism services, and general retail) that can help with the understanding of the model’s hypotheses. The random-effect model analyzed the bivariate relationships (Viechtbauer et al., 2015).

Table 3. MASEM results.

Independent Variable	→	Dependent Variable	Beta	SE	P-value
Environmental dynamism	→	Big data analytics	.511	.14	p < .001
Resources and capabilities	→	Big data analytics	.414	.15	p < .001
Competitive pressure	→	Big data analytics	.115	.015	p < .001
Big data analytics	→	Service industry performance	.332	.65	p < .001

We examined environmental dynamism, resources and capabilities, and competitive pressure as antecedents of big data analytics. Confirming Table 3, MASEM results showed that environmental dynamism has a positive and significant direct relationship with big data analytics. Furthermore, the random-effect model showed a positive and significant interaction of industry dynamism (SMD = .4417; 95% confidence interval [CI95].2668; .5884; z-value = 4.63; p < 0.001; I² = 87.8%; ℵ²(3) = 24.59; p < .001) and external factors dynamism (SMD = .4384; 95% confidence interval [CI95] .0801; .6963; z-value = 2.36; p < 0.01; I² = 97.3%; ℵ²(4) = 147.38; p < .001) with big data analytics.

Resources and capabilities have a positive and significant direct relationship, as indicated in Table 3. These findings can also be better detailed through the bivariate relationships of the random-effect model. Managerial skills (SMD = .6437; 95% confidence interval [CI95] .2766; .8469; z-value = 3.12; p < 0.01; I² = 98.1%; ℵ²(3) = 159.34; p < .001) and technical skills (SMD = .4748; 95% confidence interval [CI95] .0913; .7357; z-value = 2.368; p < 0.05; I² = 98.5%; ℵ²(5) = 325.6; p < .001) positively and significantly interacted with big data analytics.

Competitive pressure is also significantly positively correlated with big data analytics, confirming Hypothesis 3. The bivariate relationship of large companies (SMD = .3619; 95% confidence interval [CI95] .3245; .3982; z-value = 17.52; p < 0.001; I² = 78.7%; ℵ²(5) = 15.67; p < .05) and SMEs (SMD = .3234; 95% confidence interval [CI95] .2015; .3454; z-value = 5.01; p < 0.001; I² = 80.5%; ℵ²(5) = 29.59; p < .001) with big data analytics showed a positive and significant interaction.

MASEM also examined Hypothesis 4, which indicated big data analytics influencing service industry performance. The relationship between big data analytics and service industry performance proved positive and significant. These findings can also be better detailed through the bivariate relationships of the random-effect model. IT services (SMD = .2922; 95% confidence interval [CI95] .1064; .4583; z-value = 3.04; p < 0.01; I² = 91.2%; ℵ²(4) = 45.3; p < .001), health care services (SMD = .303; 95% confidence interval [CI95] .1265; .461; z-value = 3.3; p < 0.01; I² = 89.6%; ℵ²(3) = 19,23; p < .001), hotel and tourism services (SMD = .4453; 95% confidence interval [CI95] .2993; .5705; z-value = 5.52; p < 0.01; I² = 88.2%; ℵ²(3) = 105.3; p < .001), and general retail (SMD = .4548; 95% confidence interval [CI95] .1833; .6696; z-value = 2.13; p < 0.001; I² = 95.2%; ℵ²(3) = 62.48; p < .001) positively and significantly interacted with big data analytics.

4.1. Moderator analysis

We examine the moderators’ effects on the antecedent–consequent relationships (Competitive pressure -> Big data analytics and Big data analytics -> Service performance; see Table 4) to better understand the heterogeneity in our model. Our analysis only involved relationships with more than fifteen effect sizes (k = 15), excluding two relationships (Environmental dynamism -> Big data analytics and Resources and Capabilities -> Big data analytics). Our six moderators were categorical, so we used the hierarchical linear meta-analysis (HiLMA).

Table 4. Moderator effect analyzed using HiLMA.

Moderator/variables	k	Estimate (Cor)	z-value	CI95%
Relationship: Big data analytics –> Service performance
H5. AI-driven vs. no AI-driven
BDA	48	1(.278)	9.96	[.2254; 3291]
BDA-AI	12	.259 (.525)***	13.01	[.4548; 5857]
Relationship: Competitive pressure –> Big data analytics
Firm size
SMEs	6	1(.323)	5.01	[.2015; 4354]
Large enterprises	9	.111 (.419)^ns	12.22	[.3589; .4771]
Relationship: Big data analytics –> Service performance
Geographic Region
Eastern	36	1(.359)	11.24	[.3011; .4151]
Western	20	–.124 (.231) **	7.81	[.1737; .2849]
Sample Size
Large	30	1(.287)	9.11	[.2278; .3443]
Small	30	.074 (.387)^ns	8.61	[.3059; .4641]
Publication Ranking
Non-top journals	28	1(.211)	9.9	[.1702; .2512]
Top journals	32	.227 (.439)***	10.99	[.3688; .5044]
Type of industry
General retail	4	1(.458)	3.13	[.1833; .6669]
Health care services	4	–.118 (.296)^ns	4.6	[.1734; .4097]
Hotel and tourism services	3	–.149 (.331)^ns	2.57	[.0816; .5404]
IT services	7	–.179 (.304)^ns	4.18	[.1656; .4318]

Notes: k = number of studies; CI95% = Confidence intervals; *p < .05; **p < .01; ***p < .001; ns = not significant.

The first moderator tested was Hypothesis 5. The assumption of the hypothesis indicated that the use of AI in adopting big data analytics would tend to increase the effects on service performance. The HiLMA results demonstrated that there is a significant difference between AI-driven versus no AI-driven (Table 4). We analyzed the moderating of Hypothesis 5 using HiLMA to understand the difference between AI-driven versus no AI-driven. As expected, AI-driven promoted stronger relations between big data analytics and service industry performance than no AI-driven. This result indicates that AI-driven tends to increase the impacts of big data analytics on service industry performance. Previous studies demonstrate that adopting big data analytics powered by artificial intelligence would enhance firms’ performance (Benzidia et al., 2021; Dubey et al., 2020). These effects could be seen in different types of performances linked to the service area, such as operational (Dubey et al., 2020) and competitive (Bag et al., 2023).

The heterogeneity in the relationship between competitive pressure and big data analytics was also tested through the firm size moderator. Competitive pressure can be derived from the external environment and is generated by different types of customers, suppliers, and competitors (Chen et al., 2015). Small and large companies receive influences from the external environment in different ways. However, our meta-analytic findings do not demonstrate differences in the effects of competitive pressure on SMEs and large enterprises adopting big data analytics. Although large enterprises have greater effects than SMEs, these are similar.

Methodological moderators were evaluated through four variables in the relationship between big data analytics and service performance (Table 4). The first moderation concerns the effect of the sample’s country of origin. This moderator measured the difference in the geographic region. Our findings help to clarify the results of previous research better, as the influence of cultural values on performance is not unanimous in service studies. Although previous studies (Hofstede, 2011; Jadil et al., 2021) demonstrated that cultural values are an important aspect of the organizational environment, previous meta-analytic results (Oesterreich et al., 2022) analyzing big data analytics through different cultural perspectives found no statistically significant difference. On the other hand, meta-analytic evidence shows greater effect sizes for Eastern countries (Ansari & Ghasemaghaei, 2023). Our results demonstrate differences in service performance to the extent that Eastern countries have larger effect sizes than Western countries, corroborating the findings of Ansari and Ghasemaghaei (2023). These results can be justified by the increase in investment in technical platforms in Eastern countries, making their organizations more competitive (Ansari & Ghasemaghaei, 2023; Mengis, 2021).

The second moderator, sample size, did not demonstrate a significant difference or explain the heterogeneity in the effects of big data analytics on service industry performance. These results do not corroborate previous studies demonstrating that small samples tend to inflate effect values (Rosenthal, 1979; Santini et al., 2020). Despite this, our results are like those of Bitencourt et al. (2020), who found no significant difference in sample size when analyzing the effects on performance.

The third moderator analyzed the publication ranking to explain the heterogeneity of the effects generated by adopting big data analytics. Table 4 shows significant differences. The meta-analytic results demonstrate that top journals had larger effect sizes than non-top journals. Although we found studies that demonstrate the opposite (Bitencourt et al., 2020), other academic works reinforce our findings by declaring that top journals tend to generate greater effect sizes (Rosenthal & Rubin, 1982; Santini et al., 2020). Finally, industry type was evaluated as a possible moderator of the effects of competitive pressure in big data analytics. Although the largest effects occur in general retail and the smallest health care services, no significant difference was found between these two sectors and the hotel and tourism services and IT services.

5. Discussion

Our study extends previous studies of antecedents and consequents of big data analytics by explicitly considering the effects of using artificial intelligence in studies that involved service focus. Previous results have indicated increased literature applying analytical techniques (Marian et al., 2018; Belk et al., 2023). However, this research field is fragmented in scope, requiring studies that group concepts and display several gaps through technical methodologies. Our meta-analysis proposes a grouping of variables. Rialti et al. (2019) demonstrate that several studies about the effects of big data used dynamic capabilities; however, these studies still lack systematization. Our meta-analysis provides academic studies with the possibility of grouping and refinement based on three constructs. We identify three antecedents that can help managers and academics understand the drivers of big data analytics through dynamic capabilities. This study signals the importance of studying environmental dynamism. This variable had the greatest influence as an antecedent of big data analytics. Secondly, resources and capabilities are key elements in adopting big data analytics. Finally, competitive pressure had a smaller but significant effect on big data analytics. Incorporating these concepts and testing hypotheses in an integrated framework helps to understand the use of analytics in the service area. Dam et al. (2019) demonstrate that exploring the use of analytics, including descriptive, predictive, and prescriptive techniques, helps to understand the potential of the service area.

Our model also provides information that may allow managers and academics to verify that big data analytics increases service performance, as found in other academic research (Ferraris et al., 2019; Bag et al., 2021). From a broader perspective, the results indicate a greater effect on service performance in the aggregate of primary studies that used artificial intelligence in big data analytics. The current meta-analysis also explains the interconnection of BDA technologies and organizational service performance. Our study clarifies the significant contributions of AI in improving the advantages of BDA and the interconnected nature of such a relationship in improving service performance. This synergy is especially pronounced in contexts with high environmental dynamism. The fact that AI can process and analyze huge amounts of data faster and more accurately allows organizations to react to market changes more adequately and gain a competitive advantage. This finding is consistent with the dynamic capabilities perspective, which stresses the need for technological flexibility in exploiting market opportunities and managing threats. Consequently, AI integration in BDA services can be deemed as a powerful strategic capability that allows firms to benefit from the unstable nature of today’s service industries. Previous studies demonstrate that BDA-AI benefits service performance by increasing strategies’ efficiency, productivity, and profitability (Samara et al., 2020). Furthermore, the use of AI greatly impacts service operations and customers’ reactions and behaviors (Belanche et al., 2024).

In addition, the differential effects of BDA and AI on service performance in different regions demonstrate the role of contextual variables in technology adoption and effectiveness. Our study showed stronger relationships between BDA adoption and service performance in Eastern countries compared to Western countries. This gap can be explained by several features, such as a lack of digital infrastructure, organizational culture, and market maturity. For example, the more Eastern the country is and the more it is subjected to digital transformation, the more likely it will be to accept new technologies, resulting in a more significant effect on service performance. Moreover, the focus on innovation and technology-based growth strategies in these areas can strengthen the potential of BDA and AI. These results indicate the necessity of a broader contextual application of BDA and AI implementation in regional peculiarities of technology adoption, infrastructure, and market dynamics. This information is critically important for practicing professionals who need to customize their BDA and AI implementing initiatives per their regional environments for better service performance. Therefore, the geographic region explains the heterogeneity of the effects of big data analytics from a methodological perspective. Eastern countries have larger effect sizes than Western countries. Understanding the effects of technology usage across different countries aids in comprehending economic inequalities between nations (Kopalle et al., 2022). Using analytics can enhance the guidance of marketing strategies for small businesses and service entrepreneurs (Anderson et al., 2021). The significant impacts on service performance observed in Eastern countries in our meta-analysis may reflect contextual disparities and nuanced insights into local customs and behaviors. Previous literature has emphasized that access to reliable, high-speed Internet services could offer a plausible explanation for the varying effects of technology adoption across different countries (Kopalle et al., 2022; Kozinets & Gretzel, 2021; Puntoni et al., 2021). By considering these assumptions, our goal in analysis contributes to the emerging field of studies examining technologies through the lenses of country, service, and consumer.

The publication ranking also explains the heterogeneity of the effects of big data analytics from a methodological perspective. Journals with high impact had stronger relationships between big data analytics and service industry performance than journals with low impact. Top journals had larger effect sizes than non-top journals (Thornton & Lee, 2000). These results differ from those found by Bitencourt et al. (2020), who demonstrate that the publication type needs to explain the heterogeneity of effects on performance. However, our results corroborate Rosenthal and Rubin (1982) and Santini et al. (2022) in demonstrating that scientific publications with greater impact tend to prioritize publications with overestimated effects. Our findings reflect a trend in the evaluation process in the main academic journals. Previous literature considers this tendency to be publication bias (Angell, 1989; Thornton & Lee, 2000), which is a widespread problem arising from journal editors’ and reviewers’ roles in selecting studies for publication (Angell, 1989). Top journals have a more careful evaluation process than non-top journals. Research with weak effects sometimes deals with implausible hypotheses, leading to rejection in top journals despite being well-conducted study (Angell, 1989; Thornton & Lee, 2000). Evaluators tend to be more careful, not accepting low and non-significant statistical values.

6. Implications

Recent debates in the service area regarding big data analytics and the use of artificial intelligence inspire our study. The central idea was to develop a meta-analytic model to test the effects of big data analytics in the services industry using artificial intelligence. By grounding our study as a meta-analysis, we have sought to examine the main relationships involving big data analytics in recent years and how they are being studied. In this way, we review previous studies to demonstrate possible gaps in future studies to help academics and practitioners. Our study addresses three important gaps in the services industry literature: (1) What is the influence of environmental dynamism, resources and capabilities, and competitive pressure on adopting big data analytics? (2) What are effect sizes that measure big data analytics as drivers of service performance? and (3) What is the effect of using artificial intelligence in big data analytics? While most of the articles on big data work within these perspectives separately, our paper sheds light on the relevance of jointly analyzing the triad of big data analytics, artificial intelligence, and service performance.

From a theoretical perspective, our meta-analysis can synthesize and provide an aggregated view of previous articles making important theoretical contributions to the service industry. First, the previous articles used in our meta-analysis brought several theories (Resource-based view, Dynamic Capabilities view, Technology-Organization-Environment Framework, Diffusion of Innovation Theory, Institutional Theory, and Organizational Learning Theory, among others) to understand big data analytics. Theories were dispersed in the primary papers without a connection through a single model. Our meta-analysis combines these theories within a singular model to understand the differences in using big data analytics with or without artificial intelligence.

Second, based on our review of the primary data and construction of the meta-analytic model, we identified several areas where existing research can be extended, such as deepening the bivariate variables of the environmental dynamism and resources and capabilities constructs. Despite the small number of effect sizes found in primary studies and the use of common alliances, these variables are significant in adopting big data analytics. Thus, our target analysis signals the need for more studies that directly analyze industry dynamism, external factors dynamism, managerial skills, and technical skills.

Third, our meta-analysis attempts to compare studies that analyze without the use of artificial intelligence (BDA) (e.g. Lutfi et al., 2023; Maroufkhani et al., 2020) and powered by artificial intelligence (BDA-AI) (e.g. Benzidia et al., 2021; Dubey et al., 2020). We found no previous meta-analytic studies that evaluate the use of artificial intelligence as a possible moderator of the effects generated on performance. Through meta-analytical analysis of previous research, we classify the use of BDA and BDA-AI, revealing that the use of AI in big data analytics tends to generate greater service performance than without AI.

Finally, we attempted to eliminate the heterogeneity of effect sizes from previous studies using moderators. The previous articles were published in different countries (e.g. India, South Africa, United Kingdom, USA, Pakistan, Italy, Saudi Arabia, China, France, Brazil, Canada, Iran, Australia, and Cameroon, among others), with different samples (e.g. CEO/President, CIO, Head of Digital Strategy Senior, Vice President, Director, Manager, executives, managers, employees, and salespeople, among others) and different service industries (B2B, Retail, IT Development, Healthcare, and Hospitality and Tourism, among others). Our meta-analysis groups these papers to provide an integrated model for understanding the triad of big data analytics, service performance, and artificial intelligence. In short, our meta-analysis synthesizes, interprets, and offers insights from a broad range of research to help understand what is known about this triad. More specifically, we discovered that there is a difference in service performance that is explained by the sample’s country of origin. Other studies have shown that geographic location is an important indicator influencing effect sizes (Ansari & Ghasemaghaei, 2023; Hofstede, 2011). In contrast, other studies indicated the non-existence of a moderating relationship (Oesterreich et al., 2022). Our meta-analysis delves deeper into this discussion by proposing that a possible investment in technical platforms in Eastern countries may increase their performance in big data analytics (Ansari & Ghasemaghaei, 2023). Furthermore, we demonstrate that publication ranking can explain the heterogeneity of the effects generated by adopting big data analytics. This is a common phenomenon in meta-analytic studies in some areas, as scientific publications with greater impact tend to prioritize publications with overestimated effects (Rosenthal & Rubin, 1982). Our findings demonstrate that previous academic studies analyzing analytics in the service area are subject to significant publication bias. Meta-analytic studies have shown that this bias is linked to the preference of top journals for studies with stronger effects (Angell, 1989; Thornton & Lee, 2000). Since top journals are the most widely read and cited within academia (Rosenthal & Rubin, 1982), our analysis indicates that they tend to publish studies with larger effects than non-top journals. This underscores the importance of paying closer attention to smaller effect sizes. Exploratory studies, particularly those with implausible hypotheses, typically exhibit smaller effect sizes (Angell, 1989). However, top journals’ rejection of such results during the evaluation process may lead to their limited dissemination.

Our comprehensive findings on big data analytics, artificial intelligence, service performance, and moderators can influence practical implications. In this way, some important notes for management can be extracted from the results of this study. First, our study demonstrates that the presence and absence of artificial intelligence must analyze the service performance caused by big data analytics. The findings support that managers must know the relationship between big data analytics, service performance, and artificial intelligence while paying attention to important drivers such as environmental dynamism, resources and capabilities, and competitive pressure. Investments in big data analytics with the help of artificial intelligence will only bring vibrant results if the organization’s perceived pressure, good management of technical and managerial skills, and the understanding of the rate of changes in technologies and variations in customer preferences are aligned. Second, environmental dynamism is the antecedent that impacts the most on adopting big data analytics, followed by resources and capabilities. Managers must be aware of changes in the external environment and their possible consequences for the firms. Furthermore, investments in knowledge and learning are important for better adoption of big data analytics. Managers must be aware that adopting big data analytics can be directly influenced by roles’ cultural and economic conditions. Our findings demonstrate that the geographic region moderator can signal economic disparities between countries, and access to a reliable Internet should be the criteria used when choosing strategies for adopting big data analytics for service companies.

The findings of our study also suggest that it is essential for managers to include big data analytics and artificial intelligence in their strategic planning processes. Such integration ensures better-informed decisions to prepare the organizations for market trends, react to consumer behavior changes, and improve operational efficiency. Using big data analytics predictive capabilities, managers can forecast future market conditions and consumer behaviors, hence proactive strategies instead of reactive ones. Moreover, the application of AI can automate complex data analysis and, consequently, release these resources and allow firms to concentrate on strategic business areas that are the source of growth and creation of innovation.

In addition, the role played by regional variations in the effectiveness of big data analytics and AI in promoting service performance highlights the need for local system adoption and use approach. Managers need to consider such attributes of the target markets as cultural, economic, and technological elements, among others, when designing their big data and AI strategies. This includes providing localized products and services that cater to local tastes, expectations, and requirements or customizing the marketing approach of a company so that it appeals to specific regional market segments. Recognition of these regional nuances in the course of big data analytics and AI makes it possible for organizations to optimize the advantages derived from these technologies, thus making the best possible investment in relation to the performance of services and customer satisfaction.

7. Limitations and future research

The current study acknowledges several limitations applied to the service area. First, it confines its review to studies published solely in traditional journals, omitting valuable reports and findings disseminated through alternative platforms. This focus on general trends may also obscure critical insights from specific studies. The omission of data from non-traditional sources could introduce bias in the results. Second, limiting the scope of work up to a particular timeframe may reflect something other than the most recent developments in artificial intelligence in our analysis. Additionally, while we recognize regional and cultural differences, their superficial exploration implies that our findings may need more universal applicability. Third, we cannot assess specific resources or capabilities to manage, store, analyze, and interpret big data in our model. This is because, in our database, we found no more than three effect sizes for these resources or capabilities. Previous meta-analysis literature suggests a minimum necessary effect size (N = 3) (Hedges & Olkin, 1985; Hunter & Schmidt, 2004). Fourth, the coding of prior studies resulted in binary variables not allowing the use of these variables in MASEM. Fifth, one of the main areas for improvement is a need for more sufficient articles in our relationships, which is why we use common alliance strategies. The common use of alliances occurs in meta-analytic studies and helps to understand effect sizes more broadly. However, we know that the junctions do not encompass the central concept of the entire construct. Sixth, due to different value cultures worldwide, we found differences like those found by Ansari and Ghasemaghaei (2023) in the sample’s country of origin. Eastern countries tend to have a higher relationship between adopting big data analytics and service performance than Western countries. Thus, future empirical analysis could examine which aspects make these different effect sizes different. To this end, a cultural cross-study is suggested to evaluate possible effects on service performance. Finally, the study must account for the potentially diverse responses to AI and big data across different industries.

Future studies should consider unpublished data, such as company reports, for a comprehensive understanding. Analyzing the data by region and industry is vital, as what works in one area may not be effective in another. Furthermore, we should not focus solely on the positives of AI and big data, as exploring potential downsides or risks is equally important. Auh et al. (2022) state that it is important to understand the dark side of applying analytics and artificial intelligence initiatives to services because they can affect customers’ reactions and behaviors. Another important point to consider in future research is a better understanding of the type of industry moderating variable. Although this variable has not demonstrated evidence to explain the heterogeneity of the effect of adopting big data analytics on performance, previous studies demonstrate that the type of industry is an important factor to consider in the impacts of new technologies (Belanche et al., 2024; Dam et al., 2019; Mariani et al., 2018; Samara et al., 2020). Future studies could focus on service characteristics when presenting and discussing results depending on the industry or type of service. For instance, IT services are more technology-based, and caring services are more human-based. These distinctions help discuss to what extent AI and big data may be useful in each category. Although our model attempts to integrate and consolidate data from previous studies, there is still a long way to go to understand the effects of the relationship between big data analytics and artificial intelligence. Epistemological dilemmas are still strongly present in understanding the service areas (Mariani et al., 2018). Therefore, future research should adopt cross-disciplinary collaboration to understand our model’s constructs better and inter-disciplinary theoretical lenses (such as communication, anthropologies, and sociology, among others) for an overall better understanding of our model. Finally, conducting real-world tests and observing the outcomes when companies apply these findings would be extremely beneficial. Undertaking these steps will significantly strengthen our comprehension of the subject.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data can be made available upon request.