From dynamic capabilities theory to dynamic service analytics capabilities

The general framework for this study is provided by the dynamic capabilities theory. The basic proposition of this theory is that in a rapidly changing environment, a firm’s sustained competitive advantage does not depend primarily on a combination of resources at a particular point in time, but more critically comes from a firm’s ability to continually integrate, structure, and reorganize its resources [10]. In terms of process, this dynamic process is often described as a cycle of “sensing” market opportunities, ‘seizing’ them, and “reorganizing” corporate assets [11].

The situation is further different in the fast-paced, experience-heavy service industry environment. Due to high competitive pressure, real-time perception of subtle changes in customer needs becomes particularly important, and it is clear that service firms need a finer-grained, dynamic capabilities to guide their operations, and it is this real-world context that has given rise to the proposal of DSAC [4]. According to Akter et al.‘s understanding, DSAC focuses on real-time analysis of service elements and dynamic perception of market changes, which can essentially be regarded as the embodiment of the “perception” dimension of dynamic capabilities in service scenarios. However, it is puzzling that despite the forward-looking nature of the DSAC concept, academic progress has been surprisingly slow since its inception, and the research topics around DSAC proposed by Akter and his team in 2018 and 2020 have not been pursued, making DSAC almost a theoretically shelved tool. Interestingly, in stark contrast to the theoretical stagnation, there has been a rapid spread of dynamic service analytics in practice, whether it is real-time patient flow optimization in healthcare services [12], personalized recommendations with response times of less than 500 milliseconds in e-commerce services [13], or dynamic risk assessment in fintech services [14], and in practice, numerous companies are already quietly investing in the DSAC described by Those core competencies described by DSAC are already being benefited by many organizations in practice.

Therefore, it is the recovery and extension of the DSAC framework at the empirical level that this study aims to address. Not only do we want to verify the applicability of DSAC in other cultural contexts, but we also want to delve into the “black box” impact of DSAC: that is, the specific intermediary mechanisms through which these capabilities, which is so powerful in practice and so theoretically stagnant, ultimately affects the performance of service firms.

Adaptive capability

Adaptive capability (AC) has long been understood as an organization’s ability to coordinate internal resources to match external demands in response to environmental changes [15]. AC is naturally coupled with the “sense-making” function of DSAC, which translates data-driven insights into operationally meaningful organizational change. Most existing studies recognize AC as a key mediator connecting strategic resources and performance outcomes, thus demonstrating the importance of AC as a mediator [8,16–18]. However, little of the current discussion on AC has been systematically embedded in the DSAC framework. Can AC be a mediating pathway for translating DSAC into performance? This potential issue deserves further clarification. In addition, the question of how AC affects other CLC in a DSAC-driven context also remains unexplored.

Customer-linking capability

Customer-linking capability (CLC), based on CRM theory, is the ability of an organization to build and maintain strong customer relationships [19], and prior research has demonstrated CLC as a key driver of performance [20,21]. In the digital era, the expectation of instant service has brought the importance of CLC to an unprecedented level [22]. Also existing studies have shown that perceptual capabilities based on big data analytics [23], digital marketing capabilities [24], and strategic synergies of big data analytics initiatives [25] significantly enhance CLC, which can be seen as a driver of data analytics for CLC. However, existing studies have not yet placed CLC in a real-time dynamic environment to test whether it serves as a mediating mechanism for DSAC to influence FP. In addition, the potential interaction between AC and CLC has not been explored, which becomes a key missing link in understanding how data-driven adaptation translates into market success.

Dynamic service analysis capabilities and their impact on firm performance

Dynamic capabilities theory posits that in rapidly changing environments, companies possessing robust dynamic capabilities can achieve superior performance by continuously sensing external changes, capturing key information, and reconfiguring their resources [26,27]. Akter et al. argue that DSAC, as a domain-specific dynamic capabilities, enables service companies to analyze business environments, predict trends, and make informed decisions in near real-time. Previous research by Akter et al. confirmed that DSAC enhances performance by increasing organizational agility [3,4]. Within China’s highly dynamic digital markets—such as e-commerce, logistics, and fintech—intensifying competition and the growing importance of customer experience as a competitive advantage are driving organizations to invest in rapid service analytics [28,29]. Consequently, DSAC’s role may become even more critical. Based on the discussion, this study proposes the following hypothesis.

1. H1: DSAC positively influences FP in China’s digital service context

The mediating role of adaptive capability

DSAC’s ability to generate real-time insights based on consumer big data that reflect changes in the marketplace has been shown to have a direct positive impact on firm performance (FP). However, this is not the only way to realize its value, and the effectiveness of DSAC may largely depend on a key intermediary mechanism, the organization’s adaptive capability (AC), which translates insights into actionable operational measures [18]. AC is defined as the ability of an organization to reconfigure its operational processes and resource allocations based on insights [30], and in China’s fast-changing economy, the ability of DSAC to generate real-time insights based on consumer big data has been proven to have a direct positive impact on FP ([31]), which is particularly critical in China’s fast-changing digital marketplace. A typical example is that at the beginning of the COVID-19 pandemic, JD.com Logistics found a significant rise in the number of masks in Wuhan through data monitoring, and then based on this information, stocked masks for Wuhan couriers, as well as conducted rapid replenishment and internal alerts [32].The correlation between adaptive capability and performance has also been demonstrated to a considerable extent in the Chinese context [33].

DSAC insights into customer needs (e.g., the need to capture those subtle personalized recommendations [34]) enable firms to implement real-time, high-quality adaptations for their customers, and often such adaptations promote trust building between firms and their customers [35], and given that trust is a core driver of customer loyalty [36,37], it may be prudent to infer that these insight-driven adaptations are ultimately more likely to strengthen the firm-customer link. Therefore, the following hypotheses are proposed:

1. H2: DSAC significantly influences FP through AC.
2. H3: DSAC significantly influences CLC through AC.

The mediating role of a customer-linking capability

DSAC enables Chinese service firms to capture customer data in real time across a variety of contexts and identify unmet needs. This deep understanding is beneficial to service firms because they can build and maintain long-term customer relationships by creating contextualized interactions [38]. In a relationship-oriented market such as China, CLC directly drives FP because strong customer relationships enhance repeat purchase rates and word-of-mouth referrals, both of which are critical to growing market share and gaining performance [39]. One of the most prominent examples is the high performance of a Chinese fintech application service company that has cultivated a large group of loyal customers through full-cycle service accompaniment.

Adaptability (AC) enables companies to adjust processes, but these adjustments can only be translated into performance (FP) if they are communicated and validated through customer contact (CLC). For example, Chinese e-commerce firms use AC to optimize their return policies, and it is only through personalized messages that CLC can ensure that customers understand the policies, thus reducing complaints and increasing satisfaction [40]. Established research confirms that consumers in Eastern contexts value relational trust more than just functional adjustment [41], and thus CLC may play a mediating role between AC and FP. Therefore, the following hypotheses are proposed:

• H4: DSAC significantly influences FP through CLC.
• H5: AC significantly influences FP through CLC.

Chain mediating role of adaptive capability and customer-linking capability

Service organizations use DSAC to sense key changes in market and customer needs, and then use AC to translate these changes into operational-level adjustments, such as improving service processes or updating content delivery. It is worth noting that some modern service firms have leveraged artificial intelligence tools to make this adjustment process real-time and adaptive. Taking adaptive recommender systems as an example [42], most online platform recommender systems can recommend services of interest to customers in real time based on changes in user preferences. Similarly, adaptive educational services [43], adaptive bit rate adjustment for web video services [44], adaptive navigation for maps, adaptive fitness programs for health-focused apps and adaptive AI customer service for customer service platforms are facts that reflect the process from DSAC to AC.

Under the Service-Dominant Logic any adaptive behavior of the enterprise does not create value directly, but its value must be “determined” by the customer’s perceptions and interactions [45], therefore, the improvement of AC should ultimately affect the performance through the enhancement of CLC. Specifically, when a service firm responds quickly to customer needs through AC (e.g., real-time adjustment of services or personalization of content), it conveys a high degree of “caring” to the customer, and such positive interactions may significantly enhance the customer’s sense of trust and the quality of the relationship, which strengthens the link between the firm and the customer, a phenomenon which is not only a positive influence on the performance of the service firm, but also on the performance of the service firm in terms of “relationship”. This phenomenon is particularly salient in cultural contexts where “relationships” are important social capital. According to CRM, strong CLC is the cornerstone of business performance, and strong customer relationships mean higher customer loyalty, repurchase rates, and broader word-of-mouth referrals, which translate directly into superior performance. Therefore, the following hypotheses are proposed:

1. H6: DSAC significantly affects FP through the sequential mediating effects of AC and CLC.

Conceptual model

The central claim of this study is that the mechanism by which DSAC affects FP may be more convoluted than a direct causal association-it actually weaves through a synergistic mediating system consisting of AC and CLC. The model in Fig 1 outlines three possible pathways. DSAC has a direct enhancement on FP which is the starting point. Furthermore, it also acts indirectly on FP by enhancing AC and optimizing CLC, respectively, but the key here lies in the third path: AC and CLC are not isolated from each other, but form a coherent conduction sequence (DSAC → AC → CLC → FP). In other words, with DSAC, firms increase their internal resilience, which in turn makes them more capable of fostering customer-centered external linking capabilities that ultimately lead to performance. In this way, this chain model may be a useful extension of the traditional intermediation framework, which attempts to reveal that the realization of the value of technology investments actually follows a complete trajectory from “intra-organizational adaptation” to “market-side linkage”.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Conceptual model.

https://doi.org/10.1371/journal.pone.0338542.g001

Scale development

The measurement instrument was mainly based on the reliability-tested scales from established studies (see Appendix 1 for the complete items). To ensure cross-cultural validity, we implemented a rigorous translation and back-translation process for the original English questionnaire: first, the Chinese forward translation was completed by a bilingual expert, followed by the English back-translation by an independent translator, and finally, semantic and contextual differences were resolved through multiple rounds of comparison and discussion involving data analysts from three Chinese data service companies.

On this basis, we further organized a pre-test with a sample of 100 middle managers in the finance, logistics and retail industries. All participants were purposively sampled based on two criteria: first, they have no less than three years of work experience in their respective companies, and second, they have been directly involved in DSAC-related decision making. It is worth stating that the feedback from this pretest played a key role in the optimization of the scale, specifically in the improvement of item presentation clarity, contextual relevance and response format.

In terms of measurement, all constructs were measured on a 7-point Likert scale (1 = “strongly disagree”, 7 = “strongly agree”), which was designed to balance measurement efficiency and response consistency.

Survey

The data collection of this study relies on the Chinese research platform Credamo.com, which covers about 3 million users and provides the basis for the broad representativeness of the sample. The data collection period is from May 17 to November 6, 2023, and the study targets middle- and senior-level professionals in enterprises that have deployed real-time service analytics systems, and all data are collected in the form of anonymous questionnaires. On the ethical level, the project was ethically reviewed by the School of Tourism and Sports of Hezhou University (Project No. 2023005), and written informed consent was obtained from all participants.

To ensure the validity of the study, we set three core entry criteria for respondents: (1) technical literacy in interpreting real-time analytics outputs; (2) authority to deploy resources related to optimizing DSAC; (3) no less than three years of experience in an enterprise that has implemented DSAC. We believe that these three conditions, while seemingly stringent, are necessary prerequisites for accurately capturing the subtle correlation between competence and performance.

The study sample covers a wide range of service industries, with the following composition: information transmission, software and information technology services (34.06%), scientific research and technology services (6.5%), financial services (16.27%), tourism, accommodation and food services (16.57%), wholesale and retail (7.1%), transportation, warehousing and postal services (7.4%), leasing and business services (7.1%), and residential services (5%). These companies were included in the study in large part because they have all realized visible operational improvements through real-time data analytics, making them ideal candidates for our exploration of this topic. In terms of geographic distribution, the respondents cover 27 provinces and cities in China, with Shanghai, Guangdong, Zhejiang, Jiangsu and other economically active regions especially represented, which to some extent also enhances the regional representativeness of the sample.

In the data cleaning stage, this study initially excluded about 30% of invalid questionnaires (N = 460) by using the built-in mechanism of Credamo.com, based on the following behaviors: completion time less than 150 seconds, contradictory answers to reverse questions, and mechanical linear responses. Based on this, the manual review was further supplemented to exclude another 30 responses with logical inconsistencies. After such rounds of cleaning, the final number of valid samples retained was 338, with an effective recovery rate of 73.5%.

Data analysis

In this study, a reflective-reflective structure was used in the measurement model, and all first-, second-, and third-order constructs were modeled with reflective metrics. The DSAC was considered a higher-order constructs, which contains 12 dimensions in the first-order constructs, 4 dimensions in the second-order constructs, and 3 dimensions in the third-order constructs. In terms of method selection, Partial Least Squares Structural Equation Modeling (PLS-SEM) is used in this study for the following reasons:(1) Because DSAC is a complex hierarchical constructs, PLS-SEM is able to better estimate the complex hierarchical constituent model through techniques such as the repetition indicator method, whereas CB-SEM is susceptible to model identification problems [46]. (2) In this study its dual purpose of both exploratory and predictive, exploring chain mediation models with complex constructs involved, in which PLS-SEM demonstrates a unique advantage [47]. (3) PLS-SEM prioritizes variance interpretation and predictive power over model fit, which is consistent with our goal [48]. (4) Given the limited sample size (N = 338) and the need to reduce model complexity, PLS-SEM does have a better tolerance for small samples and non-normal distributions of the data relative to CB-SEM, which is suitable for dealing with complex models and sample sizes that are not very large [47]. (5) The method has been widely recognized in the field of data analysis capabilities for assessing complex models [49,50]. In this study, the bootstrap method with 5000 iterations was used to estimate the standard errors. In addition, the repeated indicator method with path weighting scheme was used to determine the optimal model parameters. Following the hierarchical component modeling methodology, lower-order constructs were always used to calculate higher-order construct scores. Accordingly, DSAC represents the aggregation of all corresponding indicators under its first-order latent variables. In the specific estimation process, we use the Bootstrap method with 5000 repeated samples to calculate the standard errors, and at the same time, we combine the repeated indicator method with the path weighting scheme in order to determine the optimal model parameters [51]. Following the typical logic of hierarchical component model construction, the scores of all higher-order constructs (including DSAC) are calculated based on their subordinate lower-order constructs. Therefore, the DSAC reported in this study essentially represents a systematic aggregation of all the indicators corresponding to its next order latent variables.

Common method bias

In questionnaire studies, common method bias (CMB) may arise when a single respondent completes all items. To mitigate this potential effect, the study included reverse coded items at the questionnaire design stage (e.g., reverse items were set in the initial “culture of analysis” construct, although this construct was excluded from the final model in subsequent analyses), which helped to identify possible habitual patterns of respondent answering, thereby reducing false correlation generation. The study also performed a Harman’s one-way test by analyzing the unrotated exploratory factor analysis. The results showed that: the first principal component explained 34.791% of the variance, a value below the recognized empirical threshold of 40% [52]. Further, the study followed the recommendation of [53] to implement a full covariance test, and the variance inflation factor (VIF) for all constructs ranged from 1.000 to 1.836, which is well below the commonly used threshold value of 3.3. Taken together, the results of these tests give reason to believe that common method bias does not appear to pose a serious problem in this study.

Measurement model

The preliminary evaluation of the measurement model involved assessing the factor loadings, composite reliability , and average variance extracted for the first-order constructs, as presented in Table 1. According to standard guidelines for PLS modeling, acceptable thresholds require item loadings greater than 0.70, CR values above 0.70, and AVE values exceeding 0.50 [48]. In this study, all item loadings exceeded 0.80, indicating strong convergent validity when measuring their respective constructs. Additionally, both and values surpassed their recommended thresholds, thus confirming the reliability and internal consistency of the measurement scale.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Measurement model.

https://doi.org/10.1371/journal.pone.0338542.t001

The correlation matrix in Table 2 confirms discriminant validity by displaying the square roots of the values along the diagonal, which are higher than the intercorrelations between the first-order constructs [54]. This test supports the theoretical distinctiveness of the constructs, indicating that they measure different concepts and are not redundant. Additionally, cross-loading analysis further validates this finding by demonstrating that each item loads more strongly on its corresponding construct than on any other construct. The first-order measurement model demonstrated robustness, as evidenced by its satisfactory reliability, convergent validity, and discriminant validity. Consequently, the measurement model was deemed adequate, providing a solid foundation for subsequent research.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Correlations and average variance extracted.

https://doi.org/10.1371/journal.pone.0338542.t002

Higher-order measurement model

Considering the hierarchical structure of the conceptual model, this study evaluates the measurement attributes of the higher-order constructs, specifically the third-order DSAC and the second-order constructs: dynamic information capability, dynamic technology capability, and dynamic talent capability, as detailed in Table 3. The third-order DSAC comprises 41 items, including 13 items for dynamic information capability, 12 items for dynamic technology capability, and 16 items for dynamic talent capability. As the DSAC was modeled reflectively, this study confirmed that the higher-order path coefficients were statistically significant. For instance, dynamic information capability accounts for approximately 86% of the variance, dynamic technology capability explains 88%, and dynamic talent capability explains 93% of the variance. As shown in Table 4, dynamic information capability is primarily reflected in information completeness (74%), currency (71%), format (76%), and accuracy (74%). Dynamic technology capability is measured through connectivity (79%), compatibility (77%), modularity (78%), and privacy (78%). Dynamic talent capability is explained by technical knowledge (72%), technical management knowledge (77%), business knowledge (75%), and relational knowledge (79%). These indicators are statistically significant at .

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Evaluation of the higher-order measurement model.

https://doi.org/10.1371/journal.pone.0338542.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Results of structural model analysis.

https://doi.org/10.1371/journal.pone.0338542.t004

Structural model

The research developed a structural model to estimate the hypothesized relationships among DSAC, AC, CLC, and FP. Table 4 presents the key path coefficients of the main model. The study confirms Hypothesis 1 (H1) by demonstrating that higher-order DSAC significantly affects FP. Furthermore, the study investigates the interaction model, examining AC’s mediating role between DSAC and both FP and CLC. The mediating effects of CLC were also analyzed, Specifically, its role between DSAC and FP, as well as between AC and FP. The results confirmed that the mediating roles of AC and CLC were statistically significant, with t-values exceeding 1.96 , thereby supporting Hypotheses 2, 3, 4, and 5. Additionally, this study validates Hypothesis 6 (H6) by confirming the chain mediation effect of AC and CLC in the relationship between DSAC and FP.To assess the mediation effects, this study employed values, which are calculated as the ratio of the indirect effect to the total effect [48]. Path analysis revealed a chain mediation pattern: approximately 24% of DSAC’s effect on FP is mediated through AC, and 15% through CLC. Furthermore, AC partially mediates the relationship between DSAC and CLC, accounting for 60% of this effect. CLC also partially mediates the relationship between AC and FP, accounting for 48% of the effect. Approximately 22% of DSAC’s influence on FP is attributed to the combined effects of both mediators. These findings indicate that AC and CLC jointly and partially mediate the DSAC-FP relationship, with each playing a distinct and complementary role. Additionally, the values for AC, CLC, and FP were 0.332, 0.352, and 0.449, respectively.

Theoretical contributions

The theoretical contributions of this study are mainly in two aspects. First, by empirically testing the higher-order structure of DSAC in an Eastern cultural context, the cross-cultural robustness and theoretical generalizability of the concept are enhanced. This is not simply a repetitive validation, but a conscious expansion of the research landscape of dynamic capabilities in service industries. Second, the chain mediation model “DSAC→AC→CLC→FP” is constructed and validated, which systematically reveals the micro-mechanism by which dynamic service analytic capabilities are ultimately transformed into firm performance. This path not only echoes the core principle of dynamic capabilities theory, “capabilities are transformed through intermediary mechanisms” [10], but also poses a challenge to the mainstream perspective of [8,25], which tends to examine the roles of AC and CLC in isolation. It is worth pointing out that our model reveals the synergistic and sequential effects of the two, which enriches the theoretical connotation of DSAC at a deeper level, and expands our understanding of the path of “capabilities-performance” transformation in the context of digital services.

Managerial contributions

This study provides specific guidance for managers of service companies. First, managers should systematically invest in DSAC construction and strengthen the monitoring of the development status of AC and CLC on the path during practice. The path can be used to diagnose the bottleneck points encountered in the process of capabilities transformation, so as to optimize the allocation of resources. Second, managers should promote the in-depth integration of DSAC with organizational processes and customer management system to ensure that the results of data analysis can be quickly transformed into adaptive actions and accurate customer interactions. Finally, industry associations and government agencies can set standards and incentives to encourage service companies to invest in DSAC infrastructure and talent development, and work together to build a dynamic data-driven service ecosystem.

Limitations and future directions of this research

Like other similar studies, this study also has some limitations, firstly, the sample focuses on Chinese service firms and the generalizability of the findings to other cultures or industries needs to be further verified. Secondly, the cross-sectional data is difficult to capture the dynamic evolution of the variables, and future research could adopt a longitudinal design to track the trajectory of DSAC formation. In addition, other potential mediating variables (e.g., digital leadership, adaptive innovation) or moderating variables (e.g., environmental uncertainty) can be explored to improve the theoretical model; finally, due to the rapid development of AI, what changes in the dimensions of the DSAC framework (e.g., information competence, technological competence, and talent competence) are generated by AI is a direction worthy of further investigation.