Digital Maturity and Supply Chain Resilience in Emerging Markets: Dynamic Capabilities as Mediators in the Industry 4.0 Transition-Evidence from Morocco

Abstract

Background: Digital transformation is viewed as a lever of supply chain resilience, yet the intermediate pathways through which digital maturity relates to resilience remain underspecified, particularly in emerging-market contexts. Drawing on the Resource-Based View and the Dynamic Capabilities Framework, this study examines whether four dynamic capabilities (visibility, flexibility, risk management, and collaboration) mediate the relationship between digital maturity and supply chain resilience. Methods: Using a cross-sectional survey of 250 Moroccan firms and partial least squares structural equation modeling (PLS-SEM), we estimate a multi-mediator model and decompose the total association using variance accounted for (VAF). Results: The findings indicate that digital maturity is positively associated with resilience both directly (β = 0.219, p < 0.01) and indirectly through the four mediators, with the four capabilities jointly accounting for 63.7% of the total association (R² = 0.523, SRMR = 0.027). Visibility (18.9%) and flexibility (15.9%) emerge as the strongest indirect channels. Conclusions: The study contributes by simultaneously testing four dynamic capabilities as mediators within a single specification, documenting evidence from an under-represented emerging-market context, and providing empirically grounded managerial recommendations and policy implications. Because the data are cross-sectional, all reported coefficients describe statistical associations.

1. Introduction

Global supply chains have recently endured a striking series of shocks, from pandemic lockdowns and geopolitical confrontations to climate disasters and chip shortages [1,2,3,4]. What were once treated as rare tail-risk scenarios now occur with increasing frequency, forcing firms to redesign their supply chain strategies at a structural rather than a purely tactical level [5,6,7]. Consequently, supply chain resilience, understood as the ability to prepare for, absorb, and bounce back from unexpected disruptions while maintaining operational continuity [5,8], has moved to the forefront of both academic inquiry and managerial attention.

The connection between digital transformation and supply chain resilience, however, proves less straightforward than popular accounts suggest. A body of recent work describes a “digitalization paradox” [9,10]: digital technologies hold great promise for monitoring, prediction, and coordination, yet an estimated 70% of digital transformation initiatives fail to deliver the outcomes organizations expect [1,11]. That such a wide gap persists between technological possibility and organizational reality implies that the pathways from digital maturity to resilience-related outcomes remain poorly mapped.

The issue becomes even more urgent in emerging economies. The institutional and infrastructural conditions that shape both digitalization and supply chain resilience in these settings differ markedly from those in high-income countries, where most existing research has been conducted [12,13,14]. Morocco provides an instructive case. Its Digital Morocco 2030 strategy [15] channels substantial public investment into digital infrastructure, but the economy still depends heavily on manufacturing, agriculture, and logistics, sectors that are vulnerable to both climate variability and the demands of integration into European production networks. Moroccan firms thus face a distinctive situation: large-scale digital ambitions colliding with limited resources, evolving regulatory frameworks, and supply chains that straddle developed and developing worlds.

Research on digitalization and supply chain resilience has expanded rapidly [13,14,16], yet three substantive gaps persist. The first is geographical: most empirical evidence originates from China, India, Western Europe, or North America [12,17,18], leaving it uncertain whether the reported relationships transfer to contexts with weaker infrastructure, different regulations, and different firm-level capabilities. The second is conceptual: many studies test a direct link between digital capabilities and resilience without examining the intervening processes. When mediators are included, researchers typically focus on a single one, whether visibility [19], flexibility [20], risk management [21], or collaboration [22], rather than comparing how several channels operate simultaneously. The third is methodological: variance-based structural equation modeling with rigorous mediation decomposition remains underused, restricting inferences about the relative importance of different pathways.

Three considerations motivate the present study. Academically, the existing literature on digital maturity and supply chain resilience remains tilted toward advanced economies and rarely models several dynamic capabilities simultaneously, leaving a structural gap that this study addresses through a multi-mediator specification with VAF decomposition. Practically, Moroccan firms operate in a context where ambitious public digital strategies (Digital Morocco 2030, Plan d’Accélération Industrielle) coexist with substantial heterogeneity in firm-level digital adoption, creating a unique field setting in which the digital-resilience association can be tested under variance-rich conditions. Contextually, North African supply chains straddle developed and developing markets and face climate, geopolitical, and infrastructural shocks, making resilience a strategic priority. These three considerations (academic, practical, and contextual) converge to justify the empirical investigation reported in this paper.

Our study responds to these gaps. We examine how digital maturity is associated with supply chain resilience among Moroccan firms, focusing on four mediating dynamic capabilities: visibility, flexibility, risk management, and partner collaboration. Rather than estimating a single direct effect, we use a mediation framework that separates the total association into the specific channels through which digital maturity and resilience are connected.

Grounded in the Resource-Based View and the Dynamic Capabilities framework [23,24,25], this study develops and tests a comprehensive structural model using Partial Least Squares Structural Equation Modeling (PLS-SEM) on survey data from 250 Moroccan firms in manufacturing, logistics, distribution, and agrifood sectors. The analysis employs bootstrapped mediation tests and a Variance Accounted For (VAF) decomposition to assess whether the four capabilities partially or fully mediate the relationship between digital maturity and resilience [26,27,28].

Three contributions emerge from this work. First, we apply supply chain resilience theory to an emerging-market setting and show empirically that the digital maturity–resilience link operates through multiple mediating pathways simultaneously. Second, the VAF decomposition we employ quantifies the extent to which each capability pathway contributes, a step beyond the binary mediation tests common in the literature. Third, the results yield a set of empirically grounded investment priorities tailored to the conditions facing Moroccan firms and firms in similar emerging economies.

The paper is structured as follows. Section 3 lays out the theoretical background and develops the nine hypotheses. Section 4 explains the methodology, including survey design, measurement, and data analysis. Section 5 presents the empirical findings. Section 6 discusses their theoretical and practical implications. Section 7 concludes.

2. Literature Review

This section reviews the empirical literature on digital maturity, dynamic capabilities, and supply chain resilience to position our contribution within the most recent body of work. Three observations structure the review. First, the association between digital maturity/supply chain resilience has been documented predominantly in advanced economies, with evidence from emerging markets still underrepresented. Second, prior empirical models typically test one mediator at a time (visibility, flexibility, risk management, or collaboration) and rarely compare several mediators within a single specification [23]. Third, mediation decomposition is usually treated dichotomously (full vs. partial) rather than quantitatively (VAF).

Table 1. Comparative summary of prior empirical studies on digital maturity/dynamic capabilities/supply chain resilience.

Author/Year	Method	Context	Constructs	Key Findings	Gap Addressed
Brusset & Teller [31]	PLS-SEM	European mfg	Information sharing → SC resilience	Visibility positively associated with resilience	Single-mediator focus; advanced-economy sample
Dubey et al. [14]	PLS-SEM	India (mfg.)	BDA + ambidexterity → SC agility	Mediation via ambidexterity	Single mediator; emerging-market specific
El Baz & Ruel [4]	PLS-SEM	Morocco	SCRM practices → resilience	RM mitigates COVID disruptions	Pre-Industry-4.0 framework
Belhadi et al. [32]	Systematic review	Emerging mkts.	Industry 4.0 → resilience	Industry 4.0 → resilience but heterogeneous	Need for empirical multi-mediator tests
Aslam et al. [6]	PLS-SEM	Pakistan	Digital transformation → SC resilience (adaptability)	Adaptability and innovation are key channels	Single mediator; no VAF
Wamba & Queiroz [33]	Systematic review	Global	Industry 4.0 → SC digitisation	Conceptual framework for digital supply chains	Mostly conceptual, limited empirical mediation
Eltantawy et al. [23]	Survey + SEM	Multi-country	Coordination → SC productivity	Coordination identified as a key dynamic capabilityis	Productivity outcome, not resilience
Ali & Govindan [29]	Conceptual	Industry 4.0 SC	Context-sensitive RBV	RBV remains relevant if dynamic	No empirical test
Ivanov & Dolgui [30]	Conceptual	Global	Shortage economy → resilience	Viability framework	No empirical test at the firm level
This study (2026)	PLS-SEM (B = 5000); MICOM; PLSpredict	Morocco (N = 250)	Digital maturity → 4 mediators (VIS, FLEX, RISK, COL) → resilience	Partial mediation with VAF = 63.7%; visibility as the strongest mediating channel	Addresses the three identified gaps within the present empirical context

summarizes the most relevant prior studies along five dimensions: methodology, geographic context, constructs studied, key findings, and identified gaps. The table situates the three observations that the present study addresses within their specific empirical contexts: emerging-market evidence, multiple parallel mediators, and a fine-grained VAF decomposition. Further methodological and contextual antecedents to this comparative review are drawn from the additional studies summarized in Table 1 [29,30].

3. Theoretical Background and Hypotheses Development

3.1. Supply Chain Resilience

Over the past twenty years, supply chain resilience has drawn sustained attention from scholars, yet the concept’s interpretation continues to evolve as disruptions grow more frequent and varied. Broadly, resilience denotes a firm’s capacity to foresee, prepare for, cope with, and recover from disruptions while sustaining acceptable performance or moving toward an improved state [5,34,35]. Whereas early definitions stressed the idea of snapping back to a prior equilibrium, newer work gives equal weight to adaptive and transformative capacities [6,7,29,36].

Christopher and Peck [37] were among the first to identify vulnerability, robustness, agility, and supply chain re-engineering as pillars of resilient supply chains. Ponomarov and Holcomb [5] built on this by proposing a definition that unifies readiness, response, and recovery within a single framework. A further advance came when scholars recognized that resilience is not a single attribute but a bundle of interacting capabilities [38,39]. Hosseini et al. [39] surveyed the quantitative modeling literature and noted that most studies addressed individual resilience dimensions without examining their interrelationships.

In a major synthesis of more than a decade of empirical work, Tukamuhabwa et al. [38] pinpointed four recurring capability dimensions behind resilience: flexible processes and sourcing, supply chain-wide visibility, speed in decision-making and recovery, and collaboration among supply chain partners. Subsequent studies have validated these dimensions across a range of settings and provide the conceptual foundation for our mediating constructs.

The COVID-19 pandemic served as a real-world natural experiment in supply chain resilience. Firms that already possessed digital capabilities and maintained diversified supplier networks recovered faster [4,7,9,13]; those relying on single sources or manual coordination suffered prolonged disruptions. Using Moroccan data, El Baz and Ruel [4] found that proactive risk management practices substantially attenuated the impact of pandemic-related disruptions, a finding we draw on to develop our hypotheses.

Four capabilities surface repeatedly in the literature as core ingredients of supply chain resilience; we treat them as mediating constructs. Visibility denotes the extent to which supply chain actors can obtain timely and accurate information on upstream and downstream activities [19]. Flexibility refers to the capacity to adjust operations rapidly when circumstances change [20,35,40]. Risk management encompasses the systematic identification, evaluation, and mitigation of possible disruptions [21,31]. Collaboration reflects the depth and quality of inter-firm relationships that enable coordinated responses [22,41].

3.2. Digital Maturity and Industry 4.0

In this paper, we treat digital maturity as an organizational capability that captures the extent to which digital technologies are woven into value-creating activities [42]. The notion goes beyond simple technology adoption: it addresses how deeply digital tools, workflows, and mindsets permeate strategic and operational routines. This perspective is consistent with recent scholarship that distinguishes genuine digital capability from the mere presence of digital tools [11,43].

Drawing on Capability Maturity Models and digital transformation frameworks, researchers argue that digital maturity develops across several organizational domains [31,39,40,44]: (1) embedding technologies in core processes, (2) building sophisticated data analytics and decision support, (3) enabling interoperability within the firm and with outside partners, and (4) making use of real-time data for both day-to-day and strategic decisions. We relied on these four domains when designing our digital maturity measure.

Supply chain digital maturity rarely rests on a single technology. Rather, it emerges from a layering process in which various Industry 4.0 technologies are stacked and combined [45,46,47]. IoT sensors and cloud platforms form the data backbone; machine learning and AI enable pattern recognition and anticipation of disruptions; blockchain and distributed ledger technologies provide traceability and trust [48,49,50]. Ivanov et al. [51] demonstrated that these technologies, when combined, reshape supply chain risk analytics by enabling both proactive detection and real-time response to ripple effects. Recent evidence from emerging markets further suggests that AI-driven tools are associated with stronger supply chain resilience, specifically through the development of dynamic capabilities [50].

This point is especially relevant for emerging markets, where information environments are typically more opaque, less standardized, and less digitally connected than those in high-income countries. In Morocco, digital maturity differs markedly across sectors and firm sizes, a heterogeneity that the Digital Morocco 2030 strategy [15] seeks to reduce. Analytically, this variation is advantageous: by sampling firms at different maturity levels, we can trace how differences in digital capability relate to differences in dynamic capabilities and, in turn, in resilience.

3.3. Dynamic Capabilities and the Resource-Based View

Our study rests on two complementary theoretical foundations. The Resource-Based View (RBV) provides the logic for why digital capabilities can confer competitive advantage. According to Barney [52], valuable, rare, inimitable, and non-substitutable resources can generate sustained competitive advantage. Digital maturity qualifies on all four counts, especially when it is deeply embedded in organizational routines and in coordination with supply chain partners. The difficulty in replicating arises from the complementary investments it requires in people, process redesign, and organizational culture, all of which accumulate over time [53].

Yet the RBV has a static quality that limits its usefulness in environments characterized by fast technological change and pronounced uncertainty, which is precisely the situation facing emerging markets in the throes of digital transformation. The Dynamic Capabilities framework (DCF) addresses this limitation [24,49,50,54].

Teece and colleagues define dynamic capabilities as the firm’s capacity to integrate, build, and reconfigure internal and external competencies to address rapidly changing environments [24,55]. The framework identifies three core processes: sensing (identifying opportunities and threats), seizing (mobilizing resources to capture opportunities), and transforming (reconfiguring assets and structures). These processes are not routines in themselves but meta-capabilities that shape how operational capabilities evolve [25,56].

The Dynamic Capabilities framework is particularly well-suited to this study for four reasons. First, the sensing–seizing–transforming logic maps naturally onto our four mediating capabilities: visibility and risk management primarily relate to sensing; flexibility primarily relates to seizing; and collaboration primarily relates to transforming through relational reconfiguration. Second, the DCF explicitly addresses how firms in turbulent environments convert resource endowments into performance outcomes, directly relevant to understanding how digital maturity is associated with resilience. Third, recent applications of the DCF to supply chain contexts have demonstrated its explanatory power for understanding both direct and mediated capability effects [25,50]. Fourth, the four selected capabilities form a parsimonious yet exhaustive set of operational mediators. Systematic reviews of supply-chain resilience [38,39] consistently identify visibility, flexibility, risk management, and partner collaboration as the four most frequently reported and most empirically supported capability dimensions. In contrast, alternative dimensions (e.g., agility, redundancy, learning) are typically subsumed under one of these four headings. This empirical convergence in the literature, combined with the theoretical alignment described above, justifies retaining these four capabilities as the mediating set in the present model.

From this integrated theoretical perspective, digital maturity can be understood as an enabling resource (in RBV terms) that facilitates the development of dynamic capabilities (in DCF terms), which in turn contribute to the desired performance outcome of supply chain resilience. This logic directly informs the mediation model tested in this study.

The integration of RBV and DCF is not without tension: RBV originally describes resources as relatively stable stocks, whereas DCF was developed to explain firm performance in turbulent environments. We resolve this tension by adopting the asset-orchestration view (Teece, 2007 [55]), in which the value of a resource depends not on its stock but on how it is configured, redeployed, and recombined through firm routines. Under this view, digital maturity is not a static technological endowment but a substrate for routinised dynamic capabilities (Eisenhardt and Martin, 2000 [56]): the same digital assets are mobilized differently across sensing, seizing and transforming cycles, which is precisely what allows them to support resilience in volatile, uncertain, complex and ambiguous environments (Helfat et al., 2007 [40]; Teece, 2007 [55]).

3.4. Hypotheses Development

The theoretical framework outlined above generates nine hypotheses organized in three groups: the association of digital maturity with each dynamic capability (H1–H4), the direct association of digital maturity with resilience (H5), and the association of each dynamic capability with resilience (H6–H9).

Digital maturity and supply chain visibility.

The deployment of digital technologies has been associated with fundamental changes in how supply chains are monitored and coordinated. Several studies have linked enhanced monitoring to the circulation of data via digital platforms, enabling firms to track activities at successive stages of the chain [17,19,57]. Research across multiple industries has described how digital tools can expand the scope, timeliness, and accuracy of supply chain information [16,32,41,51]. Additional evidence suggests that organizations with more mature digital infrastructures maintain broader visibility into partner activities, inventory positions, and potential bottlenecks [51,58].

H1. 

Digital maturity is positively related to supply chain visibility.

Digital maturity and supply chain flexibility.

Supply chain flexibility has been linked to the extent to which operational systems can adjust to changing conditions. Digital infrastructures facilitate flexible responses by enabling modifications to production schedules, logistics routing, and sourcing arrangements in shorter timeframes [20,59]. Studies have noted that real-time analytics and integrated information systems can reduce the latency between disruption detection and organizational response, enabling supply chains to shift suppliers, reconfigure routes, or adjust volumes more rapidly [14,42,43].

H2. 

Digital maturity is positively related to supply chain flexibility.

Digital maturity and risk management capability.

Variation in supply chain risk management is frequently discussed in terms of how organizations process and act upon information to manage uncertainty. Research identifies proactive risk-related practices that benefit from digital capabilities: continuous monitoring of supplier financial health, scenario simulation via digital twins, and automated alerting systems that flag emerging threats before they propagate through the supply chain [21,31,51]. By integrating data from multiple sources, digital systems enable broader scope and faster risk identification and assessment [48,50].

H3. 

Digital maturity is positively related to risk management capability.

Digital maturity and partner collaboration.

The quality of interaction between supply chain partners is partly dependent on the communication and coordination infrastructure available to them. Prior research suggests that digital platforms facilitate collaboration by reducing information asymmetries and transaction costs [22,41]. Studies have linked collaborative planning platforms, shared dashboards, and integrated communication systems to improvements in the quality and speed of joint decision-making [17,49]. However, the effectiveness of these tools depends partly on relational factors (trust, commitment, and shared norms) that digital technologies can support but not wholly create [60,61].

H4. 

Digital maturity is positively related to partner collaboration.

Direct association of digital maturity with resilience.

Beyond operating through intermediate capabilities, digital maturity may be directly associated with resilience through three complementary theoretical pathways.

First, automation and resilience-as-routine: Wieland and Durach [35] argue that digitally enabled routines are associated with more stable decision-making patterns, reflecting the codification of responses to common types of disruption and shorter observed reaction times, without requiring full mediation by higher-order capabilities.

Second, RBV-grounded direct technology associations: Wu et al. [10] document that digital infrastructure exhibits the characteristics of a valuable, rare, and inimitable resource, with direct associations to operational performance observable even when intermediate capabilities are not fully developed.

Third, AI-driven direct paths in emerging-market settings: Al-Banna et al. [50] report a direct association between AI-powered analytics and resilience, observable in contexts characterized by real-time anomaly detection and automated response selection, particularly in resource-constrained contexts.

Taken together, these three theoretical considerations are consistent with the hypothesis of a direct path from digital maturity to resilience, as well as the four mediated pathways [33]. The existence of both direct and mediated pathways is consistent with partial mediation, a configuration frequently observed in organizational capability research [27].

H5. 

Digital maturity is positively and directly related to supply chain resilience.

Dynamic capabilities and supply chain resilience.

The literature has examined supply chain resilience as the emergent outcome of multiple complementary organizational capabilities. Visibility, flexibility, risk management, and collaboration operate in concert, each contributing differentially to the overall resilience posture [31,38,62]. Several studies have demonstrated positive associations between individual capabilities and resilience outcomes [4,6,19,20,21,22], while others have argued that the combined effect of multiple capabilities exceeds the sum of their individual contributions [14,25]. These findings underpin the following hypotheses:

H6–H9. 

Supply chain visibility (H6), flexibility (H7), risk management capability (H8), and partner collaboration (H9) are positively related to supply chain resilience.

Figure 1 presents the conceptual model with all hypothesized relationships. The model posits that digital maturity is both directly associated with supply chain resilience (H5) and indirectly associated with it through four mediating dynamic capabilities: visibility (H1 → H6), flexibility (H2 → H7), risk management (H3 → H8), and collaboration (H4 → H9).

Triple Bottom Line (TBL) and sustainability considerations. While the present model focuses on the three classical resilience dimensions (visibility, flexibility, risk management, partner collaboration), recent scholarship has emphasized that supply chain resilience increasingly intersects with the three pillars of the Triple Bottom Line: economic, environmental, and social sustainability [63,64]. Although the incorporation of explicit sustainability constructs is beyond the scope of this study, we acknowledge that resilience pathways may also operate through environmental (e.g., low-carbon logistics, circular sourcing) and social (e.g., labor-relations robustness, community engagement) channels. We treat this as a complementary research stream rather than a competing one and discuss its implications for future research in Section 6.3.

4. Materials and Methods

4.1. Research Design and Sample

The hypothesized relationships were tested through a cross-sectional survey. We targeted Moroccan firms in sectors that are particularly prone to supply chain shocks: manufacturing, logistics and transportation, distribution and retail, and agrifood. Eligible firms had at least 10 employees, were active in supply chain operations, and had begun some form of digital transformation. Firms with fewer than 10 employees were excluded to ensure a minimum level of organizational complexity.

Respondents occupied positions that entail decision-making or cross-functional coordination in supply chain management, logistics, information systems, or general management. A total of 250 usable responses were collected through a combination of professional association channels (AMLOG, CGEM), LinkedIn outreach, and face-to-face contacts at industry events. Relying on several channels helped minimize single-source sampling bias and yielded a sample that reflects firms of varying sizes, sectors, and digital maturity levels.

Data collection took place between October 2024 and January 2025, a period of four months. A total of 412 questionnaires were distributed through the multi-channel strategy described above. The questionnaire was administered electronically using Google Forms (Google LLC, Mountain View, CA, USA; https://forms.google.com). Data screening, preparation, and statistical analyses were conducted using SmartPLS 4 (SmartPLS GmbH, Oststeinbek, Germany). SmartPLS 4 was used to estimate the measurement and structural models and to test the proposed hypotheses. Several design decisions contributed to the final response rate of 60.7% (250 usable responses from 412 distributed). First, the multi-channel strategy enabled direct engagement with respondents, rather than relying on passive distribution. Second, a clear and concise questionnaire (28 items, approximately 12 min to complete) reduced abandonment. Third, personalized follow-up communications were sent at two-week intervals, including a summary of the study’s aims. Fourth, distribution through professional associations and industry events enabled targeted recruitment of respondents with relevant supply chain responsibilities, rather than broad-based distribution to general populations. These factors, combined with the topic’s timeliness in the context of Morocco’s Digital Morocco 2030 strategy [15], likely contributed to the relatively high response rate.

4.2. Measures

Each construct was measured using multi-item scales adapted from well-established instruments in the supply chain and information systems literature. The wording was adjusted for the Moroccan setting using Brislin’s back-translation procedure. Appendix A contains the full survey instrument.

Digital maturity (MAT) was measured using four items from Kane et al. [11] and Ross et al. [65], covering the integration of digital tools into core operations, data analytics sophistication, system interoperability, and the use of real-time data. Visibility (VIS) was measured using four items from Brandon-Jones et al. [62]. Flexibility (FLEX) used four items from Stevenson and Spring [59]. Risk management (RISK) drew on four items from Wieland and Wallenburg [60]. Collaboration (COL) was captured with four items from Cao and Zhang [41]. Resilience (RCL) used four items adapted from Ponomarov and Holcomb [5] and Blackhurst et al. [66].

The structural model also included two control variables. Firm size was coded categorically by employee count (1 = small, 1–49 employees; 2 = medium, 50–249 employees; 3 = large, 250+ employees). The industry sector reflected the five industry groups in the sample (manufacturing, logistics, distribution, agrifood, and other). Supplier diversification (DIVERS), measured with three items on the breadth and geographic spread of the supplier base, served as an additional control for alternative resilience strategies.

All reflective items used five-point Likert scales ranging from 1 (strongly disagree) to 5 (strongly agree). Because French is the principal business language in Morocco, the questionnaire was administered in French. A bilingual researcher performed back-translation following Brislin’s protocol to verify equivalence.

Following Hair et al. [26] for categorical control variables in PLS-SEM, the industry-sector variable was recoded into k − 1 = 4 dummy variables (D_Manufacturing, D_Logistics, D_Distribution, and D_Agrifood), with the residual category ‘Other’ (n = 21, 8.4%) serving as the reference. Each dummy was entered as a single-item construct in the structural model in line with the standard procedure for nominal controls in PLS-SEM. This recoding replaces the previous single-coefficient specification of the SECTOR control, which was mathematically inappropriate for a five-category nominal variable. The structural model was re-estimated with B = 5000 bootstrap samples; the substantive paths and their standard errors remained stable, and the individual sector dummies were retained in the model.

4.3. Data Analysis

The skewness and excess kurtosis of the 24 reflective indicators were examined as a transparency diagnostic. The results, summarized in

Table 2. Normality diagnostics, skewness and excess kurtosis of the 24 reflective indicators (N = 250).

Indicator	Mean	SD	Skewness	Excess Kurtosis
MAT1	3.272	0.977	0.005	−0.74
MAT2	3.256	1.005	−0.077	−0.394
MAT3	3.256	0.952	−0.11	−0.41
MAT4	3.284	0.946	−0.079	−0.528
VIS1	3.336	0.868	−0.042	−0.094
VIS2	3.352	0.907	−0.365	0.062
VIS3	3.36	0.904	−0.282	0.161
VIS4	3.376	0.889	−0.364	0.282
FLEX1	3.348	0.773	−0.007	−0.172
FLEX2	3.376	0.818	0.316	−0.36
FLEX3	3.396	0.831	−0.004	−0.58
FLEX4	3.392	0.859	0.071	−0.452
RISK1	3.196	0.868	−0.019	−0.161
RISK2	3.192	0.861	0.075	−0.17
RISK3	3.172	0.868	−0.192	−0.146
RISK4	3.188	0.851	−0.133	−0.037
COL1	3.284	0.814	−0.114	−0.325
COL2	3.304	0.824	0.123	−0.107
COL3	3.292	0.821	−0.103	−0.116
COL4	3.348	0.842	−0.285	0.111
RCL1	3.352	0.967	−0.085	−0.417
RCL2	3.352	0.929	−0.03	−0.418
RCL3	3.36	0.926	−0.165	−0.34
RCL4	3.364	0.965	−0.136	−0.53

, show skewness values in the range [−0.36, +0.32] and excess kurtosis in [−0.74, +0.28], with no indicator exceeding the |1| skewness or |3| excess kurtosis thresholds suggested by Hair et al. [26]. The choice of variance-based PLS-SEM is therefore justified primarily on grounds of model complexity (four parallel mediators), an explanatory-predictive orientation that supports out-of-sample diagnostics, and dominant practice in the digital maturity/supply chain literature [11,62].

We chose PLS-SEM, implemented in SmartPLS 4.0, as the primary analytical technique. Several reasons guided this decision. The study’s goal is to predict and explain variance in the endogenous construct (resilience), a task well suited to PLS-SEM’s variance-based estimation logic [26,28]. The model contains multiple mediators, creating a path structure for which PLS-SEM has computational advantages [67]. Moreover, PLS-SEM performs well with samples of moderate size; our n of 250 comfortably exceeds [68] the standard minimum of ten times the largest [69] number of structural paths directed at any single construct [26,70]. PLS-SEM was preferred over covariance-based SEM (CB-SEM) for three reasons: (i) the model has multiple parallel mediators, which favors the variance-based estimator [26]; (ii) our research orientation is explanatory-predictive, supporting the use of out-of-sample diagnostics (PLSpredict, [70]); and (iii) PLS-SEM is the dominant practice in the digital-maturity/supply-chain literature [11,32].

The analysis followed the two-stage protocol recommended by Hair et al. [26]. The first stage assessed measurement adequacy: for reflective constructs, this included indicator loadings (≥0.70), internal consistency (Cronbach’s α ≥ 0.70, composite reliability ≥ 0.70), convergent validity (AVE ≥ 0.50), and discriminant validity (Fornell-Larcker criterion and HTMT ratios < 0.85). The second stage evaluated the structural model: path coefficients with bootstrap significance (5000 subsamples), R² values, Q² predictive relevance via blindfolding, f² effect sizes, and model fit (SRMR).

Mediation effects were assessed following Hair et al. [26] and Zhao et al. [27], computing specific indirect effects via bootstrapping and calculating Variance Accounted For (VAF) ratios to determine the extent of mediation. Common method bias was evaluated using Harman’s single-factor test and Kock’s [71] full collinearity VIF approach (all VIFs < 3.3).

To address potential common method bias (CMB), the protocol is reported in full detail. We applied the two-level approach recommended by Podsakoff et al. [54]. At the procedural (ex-ante) level, we guaranteed full respondent anonymity and separated predictor and criterion items across distinct sections of the questionnaire. At the statistical (ex post) level, three diagnostics were conducted. (i) Harman’s single-factor test: the largest extracted factor explains 38.8% of the total variance, well below the conventional 50% threshold. (ii) Kock’s [71] full-collinearity VIF: all inner-model VIFs are below the conservative 3.3 threshold (maximum VIF = 2.41 for MAT). (iii) An unmeasured latent method construct (CLF/marker variable) was added to the structural model; the magnitudes and significances of the substantive paths remained stable, confirming that the structural estimates are not driven by common method variance. Together, these diagnostics indicate that CMB is unlikely to bias the reported relationships.

5. Results

5.1. Sample Characteristics

Table 3. Respondent demographics, hierarchical level and functional area (N = 250).

Characteristic	Category	n	%
Hierarchical level	Middle/Functional management	200	80.0%
Hierarchical level	Other	30	12.0%
Hierarchical level	Senior management (General Management)	20	8.0%
Functional area	Supply chain/Logistics	105	42.0%
Functional area	Operations/Procurement	95	38.0%
Functional area	IT/Information Systems	30	12.0%
Functional area	General management	20	8.0%
Job title	Supply Chain Manager	60	24.0%
Job title	Operations Manager	55	22.0%
Job title	Logistics Manager	45	18.0%
Job title	Procurement Manager	40	16.0%
Job title	IT/Information Systems	30	12.0%
Job title	General Management	20	8.0%

reports the demographic profile of the 250 respondents. To assess firm-level constructs reliably, single informants must possess holistic operational and strategic knowledge of the firm. In our sample, all respondents hold a managerial-level position: 8.0% are senior managers in General Management roles, while 92.0% hold middle/functional management positions (Supply Chain Manager, Operations Manager, Logistics Manager, Procurement Manager, IT Manager). 80.0% work directly in supply-chain, logistics, operations, or procurement functions, 12.0% in IT/Information Systems, and 8.0% in General Management. This profile supports the assumption that respondents are competent key informants for the constructs of digital maturity and resilience.

The final sample spans multiple sectors, firm sizes, and geographic regions, mirroring the diversity of Morocco’s business landscape.

Table 4. Sample characteristics (N = 250).

Characteristic	Category	n	%
Industry Sector	Manufacturing	99	39.6%
	Logistics	49	19.6%
	Distribution	49	19.6%
	Agrifood	32	12.8%
	Other	21	8.4%
Company Size	Small (10–49)	68	27.2%
	Medium (50–249)	110	44.0%
	Large (250+)	72	28.8%
Geographic Region	Casablanca-Settat	120	48.0%
	Tanger-Tétouan-Al Hoceïma	59	23.6%
	Other regions	71	28.4%

provides the full sample profile.

Medium-sized firms (50–249 employees) represent the largest group (44.0%, n = 110), followed by large firms (28.8%, n = 72) and small firms (27.2%, n = 68). The predominance of mid-sized firms is noteworthy: these are the organizations where the twin pressures of digital transformation and resource scarcity are felt most acutely. Manufacturing accounts for 39.6% of the sample (n = 99), with logistics and distribution each contributing 19.6%, agrifood 12.8%, and other sectors 8.4%.

The composition of the sample reflects the structure of the Moroccan economy rather than a convenience sampling artifact. Manufacturing (39.6%) is the dominant non-agricultural sector in national GDP; logistics and distribution (19.6% each) are over-represented because Morocco’s industrial strategy positions the country as a North-African logistics hub, anchored by the Tanger Med port and the dedicated industrial zones; agrifood (12.8%) reflects its weight in industrial value-added. The geographic concentration in the Casablanca-Settat region (48.0%) is consistent with this region accounting for approximately half of national industrial activity. We acknowledge that this composition is specific to Morocco and discuss its implications for external validity in Section 6.3.

Descriptive statistics for the main constructs reveal moderate levels across all six constructs (

Table 5. Descriptive statistics and inter-construct correlations.

Construct	Mean	SD	1	2	3	4	5	6	7
1. MAT	3.29	0.89	1.000
2. VIS	3.42	0.91	0.568	1.000
3. FLEX	3.28	0.82	0.478	0.512	1.000
4. RISK	3.19	0.88	0.480	0.537	0.498	1.000
5. COL	3.31	0.85	0.406	0.467	0.453	0.471	1.000
6. RCL	3.36	0.78	0.502	0.521	0.489	0.476	0.483	1.000
7. DIVERS	3.14	0.91	0.213	0.198	0.187	0.204	0.176	0.312	1.000

Note: All correlations are significant at p < 0.01 (two-tailed). N = 250. All constructs were measured on 5-point Likert scales.

). Mean digital maturity was 3.29 (SD = 0.89) on the 5-point scale, suggesting that the majority of firms in the sample have initiated but not completed their digital transformation journeys. Supply chain visibility reported the highest mean (3.42), while risk management scored the lowest (3.19).

To address concerns about non-response bias, we performed an Armstrong and Overton [72] wave analysis, comparing early respondents to late respondents as proxies for non-respondents. The first quartile (early wave, n = 62) was compared with the last quartile (late wave, n = 62) on each of the six latent constructs: digital maturity (MAT), visibility (VIS), flexibility (FLEX), risk management (RISK), partner collaboration (COL), and supply chain resilience (RCL)and on the key categorical demographic variables. As reported in

Table 6. Non-response bias diagnostics: early-wave (Q1, n = 62) versus late-wave (Q4, n = 62) respondents.

Variable	Early Wave M ± SD	Late Wave M ± SD	Test	t/χ2	df	p
MAT (Digital maturity)	3.427 ± 0.833	3.254 ± 0.886	Welch t	+1.123	121.53	0.264
VIS (Visibility)	3.411 ± 0.650	3.359 ± 0.742	Welch t	+0.418	119.92	0.676
FLEX (Flexibility)	3.569 ± 0.675	3.359 ± 0.678	Welch t	+1.725	122.00	0.087
RISK (Risk management)	3.230 ± 0.729	3.194 ± 0.792	Welch t	+0.265	121.16	0.791
COL (Partner collaboration)	3.331 ± 0.704	3.383 ± 0.757	Welch t	−0.399	121.36	0.690
RCL (Supply chain resilience)	3.516 ± 0.873	3.286 ± 0.823	+1.508	121.59	0.134
Industry sector			χ2	3.646	4	0.456
Company size			χ2	1.381	2	0.501

, independent-samples Welch t-tests yielded no statistically significant difference between waves on any construct (all p > 0.05); the smallest gap was observed for FLEX (t = +1.725, df = 122.00, p = 0.087), and all remaining p-values exceeded 0.13. Chi-square tests on industry sector (χ² = 3.646, df = 4, p = 0.456) and company size (χ² = 1.381, df = 2, p = 0.501) were similarly non-significant. We therefore conclude that non-response bias is unlikely to affect the substantive findings.

5.2. Measurement Model Assessment

The psychometric properties of the measurement model are summarized in

Table 7. Measurement model summary.

Construct	Item	Loading	Cronbach’s α	CR	AVE
Digital Maturity (MAT)	MAT1	0.873
	MAT2	0.901
	MAT3	0.927	0.929	0.950	0.825
	MAT4	0.944
SC Visibility (VIS)	VIS1	0.889
	VIS2	0.921
	VIS3	0.943	0.959	0.971	0.892
	VIS4	0.959
SC Flexibility (FLEX)	FLEX1	0.853
	FLEX2	0.867
	FLEX3	0.878	0.894	0.927	0.760
	FLEX4	0.892
Risk Management (RISK)	RISK1	0.839
	RISK2	0.862
	RISK3	0.881	0.903	0.933	0.776
	RISK4	0.906
Partner Collab. (COL)	COL1	0.820
	COL2	0.851
	COL3	0.874	0.874	0.914	0.726
	COL4	0.900
SC Resilience (RCL)	RCL1	0.867
	RCL2	0.879
	RCL3	0.895	0.911	0.938	0.791
	RCL4	0.909

Note: CR = Composite Reliability; AVE = Average Variance Extracted. All factor loadings are significant at p < 0.001.

. All quality criteria were satisfied. Indicator loadings ranged from 0.820 (partner collaboration) to 0.959 (visibility), all exceeding the 0.70 threshold. Cronbach’s α ranged from 0.874 (collaboration) to 0.959 (visibility), and composite reliability ranged from 0.914 to 0.971, indicating strong internal consistency. Average variance extracted (AVE) ranged from 0.726 to 0.892, exceeding the 0.50 threshold for convergent validity.

Discriminant validity was assessed using both the Fornell-Larcker criterion and the more rigorous Heterotrait-Monotrait (HTMT) ratio (

Table 8. Discriminant validity—Fornell-Larcker criterion.

	MAT	VIS	FLEX	RISK	COL	RCL
MAT	0.908
VIS	0.568	0.944
FLEX	0.478	0.512	0.872
RISK	0.480	0.537	0.498	0.881
COL	0.406	0.467	0.453	0.471	0.852
RCL	0.502	0.521	0.489	0.476	0.483	0.889

Note: Diagonal values represent the square root of AVE. Off-diagonal values represent inter-construct correlations. Discriminant validity is established when diagonal values exceed all off-diagonal values in their respective row and column.

and

Table 9. Discriminant validity—Heterotrait-Monotrait (HTMT) ratios.

	MAT	VIS	FLEX	RISK	COL	RCL
MAT	–
VIS	0.601	–
FLEX	0.518	0.548	–
RISK	0.519	0.571	0.546	–
COL	0.449	0.502	0.507	0.527	–
RCL	0.541	0.554	0.536	0.521	0.537	–

Note: All HTMT ratios are below the conservative threshold of 0.85 [26]. Maximum HTMT = 0.601 (MAT ↔ VIS).

). Under the Fornell-Larcker criterion, the square root of each construct’s AVE exceeds all inter-construct correlations in the corresponding row and column. All HTMT ratios are below the conservative 0.85 threshold [26], with the highest ratio being 0.601 (MAT ↔ VIS).

5.3. Structural Model Results

The structural model showed strong explanatory power in

Table 10. Structural Path Coefficients and Hypothesis Testing Results.

Hypothesis	Path	β	t-Value	p-Value	f2	Effect	Decision
H1	MAT → VIS	0.568	12.847	<0.001	0.476	Large	Supported
H2	MAT → FLEX	0.478	9.234	<0.001	0.296	Medium	Supported
H3	MAT → RISK	0.480	9.456	<0.001	0.300	Medium	Supported
H4	MAT → COL	0.406	7.123	<0.001	0.197	Medium	Supported
H5	MAT → RCL	0.219	2.891	0.004	0.058	Small	Supported
H6	VIS → RCL	0.201	3.156	0.002	0.047	Small	Supported
H7	FLEX → RCL	0.200	3.089	0.002	0.046	Small	Supported
H8	RISK → RCL	0.188	2.945	0.003	0.040	Small	Supported
H9	COL → RCL	0.204	3.234	0.001	0.049	Small	Supported
Control	DIVERS → RCL	0.127	2.278	0.023	0.024	Small	Significant
Control	SIZE → RCL	0.098	1.987	0.047	0.014	–	Significant
Control	D_Agrifood → RCL	0.180	2.332	0.021	0.011	Small	Significant
Control	D_Distribution → RCL	0.099	1.150	0.237	0.003		n.s.
Control	D_Logistics → RCL	0.052	0.609	0.506	0.001		n.s.
Control	D_Manufacturing → RCL	0.127	1.274	0.191	0.005		n.s.

Note: Bootstrap = 5000. R²(RCL) = 0.523; Q²(RCL) = 0.497; SRMR = 0.027. f² thresholds: 0.02 = small, 0.15 = medium, 0.35 = large [26]. Control variables: supplier diversification, firm size, and industry sector.

. The R² for supply chain resilience was 0.523, indicating that the model explains 52.3% of the variance in the dependent variable, a moderate-to-substantial result according to Hair et al. [26]. The Q² value of 0.497 confirms the model’s predictive relevance (Q² > 0). The SRMR of 0.027 indicates good model fit (threshold < 0.08).

Collinearity among predictors was assessed using the variance inflation factor (VIF) values. For the endogenous construct supply chain resilience, the VIF values of all predictors were well below the threshold of 5.0: visibility (VIF = 1.72), flexibility (VIF = 1.58), risk management (VIF = 1.63), collaboration (VIF = 1.41), direct digital maturity (VIF = 1.68), supplier diversification (VIF = 1.15), firm size (VIF = 1.09), and industry sector (VIF = 1.07). These values indicate that multicollinearity does not threaten the stability of the path estimates.

All nine hypotheses were supported at conventional significance levels (see Figure 2). Digital maturity’s strongest association was with visibility (β = 0.568, p < 0.001), with an effect size nearly three times that of the direct path to resilience (f² = 0.476, large effect). The three remaining dynamic capability paths showed medium effect sizes: flexibility (β = 0.478, f² = 0.296), risk management (β = 0.480, f² = 0.300), and collaboration (β = 0.406, f² = 0.197).

5.4. Mediation Analysis

The mediation analysis below decomposes the total association between digital maturity and supply chain resilience into a direct path and four indirect paths. Because the data are cross-sectional, these decompositions describe the structural pattern of associations among the constructs at one point in time.

Table 11. Mediation analysis results.

Mediating Path	Indirect Effect	t-Value	95% CI	% of Total
MAT → VIS → RCL	0.114	2.89	[0.041, 0.193]	18.9%
MAT → FLEX → RCL	0.096	2.67	[0.029, 0.169]	15.9%
MAT → RISK → RCL	0.090	2.54	[0.024, 0.163]	15.0%
MAT → COL → RCL	0.083	2.41	[0.019, 0.153]	13.8%
Total Indirect	0.383	5.12	[0.241, 0.531]	63.7%
Direct Effect	0.219	2.89	[0.071, 0.367]	36.3%
Total Effect	0.602	8.94	[0.473, 0.731]	100.0%

(visualized in Figure 3) breaks the total effect down into its direct and indirect parts. Digital maturity’s total effect on resilience reached 0.602. The direct component was 0.219 (36.3%); the indirect component, running through the four dynamic capabilities, was 0.383 (63.7%). This pattern is consistent with partial mediation (VAF = 63.7%).

Visibility emerged as the strongest indirect channel, accounting for 18.9% of the total association (indirect effect = 0.114). This figure results from combining the strong path from digital maturity to visibility (β = 0.568) with the path from visibility to resilience (β = 0.201). Flexibility mediated 15.9% (indirect = 0.096) and risk management 15.0% (0.090). Collaboration, though the smallest individual mediator, still carried 13.8% (0.083) and remained statistically significant. All four confidence intervals excluded zero.

5.5. Post-Hoc Analyses

Following Shmueli et al. (2019) [70], we complemented the in-sample blindfolding diagnostic (Q² = 0.497) with the PLSpredict procedure to assess out-of-sample predictive validity. With k = 10 folds and 10 repetitions, all four indicators of the focal endogenous construct (RCL) yielded positive Q²_predict values ranging from 0.289 (RCL1) to 0.411 (RCL3), indicating that the model retains predictive power on hold-out data. The PLS-SEM RMSE was lower than the linear model (LM) benchmark RMSE for three of the four indicators (RCL2, RCL3, RCL4), and the same pattern held for MAE, supporting medium out-of-sample predictive power. Taken together, the in-sample R² (0.523) and SRMR (0.027) indicate that the model balances explanatory and predictive performance. We acknowledge that Q²_blindfolding is now considered an in-sample diagnostic only, and we therefore rely on PLSpredict for the predictive-relevance claim (

Table 12. PLSpredict results out-of-sample predictive validity of the four RCL indicators (k = 10 folds × 10 reps).

RCL Indicator	Q2_Predict	RMSE PLS	RMSE LM	MAE PLS	MAE LM
RCL1	0.2889	0.8181	0.808	0.6635	0.6459
RCL2	0.3594	0.7458	0.7718	0.6138	0.6243
RCL3	0.4111	0.7121	0.7288	0.577	0.5926
RCL4	0.3967	0.7513	0.7809	0.6	0.6253

).

Before performing the multi-group comparison between manufacturing (n = 99) and service firms (n = 151), we tested the measurement invariance of the composite model using the three-step MICOM procedure [44]. Step 1 (configural invariance) is satisfied by design, since both groups use the identical measurement instrument and the same algorithmic settings. Step 2 (compositional invariance) was tested via a permutation test with 1000 permutations of the group label: the original cross-group correlation between composite scores reaches 1.000 for all six constructs, well above the 5% lower-bound permutation quantile, supporting compositional invariance for every construct. Step 3 (equality of composite means and variances) is satisfied for the variance of all six constructs (all Levene p > 0.09) and for the mean of five out of six constructs (t-test p > 0.13 for MAT, VIS, FLEX, COL, RCL); only RISK shows a non-trivial mean difference (services 0.27 points higher; p = 0.007). Partial measurement invariance is therefore established, the threshold required to interpret group differences in structural path coefficients [44]. The subsequent multi-group comparison shows seven of the nine structural paths display |Δβ| < 0.20 between Services (n = 151) and Manufacturing (n = 99).

A series of supplementary checks tested the robustness of the main findings. Adding quadratic terms for digital maturity did not yield significant nonlinear effects (p > 0.10), supporting the linearity assumption. A multi-group comparison between manufacturing firms (n = 99) and service firms (n = 151) revealed no significant path coefficient differences for eight of the nine structural paths (all Henseler p > 0.10), pointing to structural stability across sectors, except H7 (FLEX → RCL), which exhibits a significant inter-group difference (Δβ = +0.296, p = 0.009).

Before performing the multi-group comparison, we tested the measurement invariance of the composite model using the three-step MICOM procedure [44]. Step 1 (configural invariance) is satisfied by design, since both groups use the identical measurement instrument and the same algorithmic settings. Step 2 (compositional invariance) was tested via a permutation test with 1000 permutations of the group label: the original cross-group correlation between composite scores reaches 1.000 for all six constructs, well above the 5% lower-bound permutation quantile, supporting compositional invariance for every construct. Step 3 (equality of composite means and variances) is satisfied for the variance of all six constructs (all Levene p > 0.09) and for the mean of five out of six constructs (t-test p > 0.13 for MAT, VIS, FLEX, COL, RCL); only RISK shows a non-trivial mean difference (services 0.27 points higher; p = 0.007). Partial measurement invariance is therefore established, the threshold required to interpret group differences in structural path coefficients [44]. Detailed MICOM and MGA results are reported in

Table 13. Measurement invariance of composites.

Construct	5% Perm. Quantile	M Services	M Mfg	ΔMean	95% CI of ΔMean	t (Welch)	p (Mean)	Var Services	Var Mfg	Levene F	p (Var)
MAT	1.000	3.305	3.210	+0.095	[−0.127, +0.317]	+0.845	0.399	0.777	0.745	0.024	0.876
VIS	1.000	3.336	3.386	−0.050	[−0.255, +0.154]	−0.485	0.628	0.709	0.599	0.917	0.339
FLEX	1.000	3.434	3.293	+0.141	[−0.041, +0.323]	+1.525	0.129	0.513	0.508	0.002	0.969
RISK	1.000	3.293	3.025	+0.268	[+0.074, +0.462]	+2.720	0.007	0.511	0.624	0.050	0.824
COL	1.000	3.285	3.341	−0.056	[−0.237, +0.125]	−0.613	0.541	0.495	0.507	0.024	0.877
RCL	1.000	3.374	3.331	+0.043	[−0.163, +0.250]	+0.413	0.680	0.791	0.571	2.761	0.098

and

Table 14. Multi-group analysis.

Hypothesis	Path	β Services	β Manufacturing	Δβ (Services − Mfg)	95% CI of Δβ	t (Param.)	p (Henseler)
H1	MAT → VIS	+0.603	+0.640	−0.036	[−0.189, +0.119]	−0.458	0.644
H2	MAT → FLEX	+0.564	+0.565	−0.001	[−0.177, +0.180]	−0.009	0.992
H3	MAT → RISK	+0.433	+0.565	−0.132	[−0.310, +0.049]	−1.465	0.142
H4	MAT → COL	+0.455	+0.485	−0.030	[−0.247, +0.196]	−0.268	0.779
H5	MAT → RCL (direct)	+0.134	+0.409	−0.275	[−0.560, +0.025]	−1.841	0.071
H6	VIS → RCL	+0.227	+0.132	+0.095	[−0.139, +0.326]	+0.790	0.427
H7	FLEX → RCL	+0.297	+0.001	+0.296	[+0.076, +0.502]	+2.670	0.009
H8	RISK → RCL	+0.095	+0.172	−0.077	[−0.298, +0.149]	−0.679	0.494
H9	COL → RCL	+0.210	+0.161	+0.048	[−0.162, +0.259]	+0.447	0.664

, respectively.

6. Discussion

6.1. Theoretical Implications

What do these findings mean for existing theory? We highlight five contributions that advance our understanding of digital transformation, dynamic capabilities, and supply chain resilience, with particular reference to emerging-market contexts.

First, and perhaps most fundamentally, our findings speak directly to the digitalization paradox that has preoccupied recent scholarship [9,10]. The observation that digital investments often fail to deliver expected returns has led some researchers to question whether digital maturity is genuinely beneficial to organizational resilience. Our results suggest that the apparent paradox may, in part, reflect a pattern of mediation. The substantial total association (β = 0.602) is consistent with the view that digital maturity is meaningfully related to resilience, but 63.7% of this association operates through dynamic capabilities rather than through the direct path. Organizations that invest in digital technologies without developing the complementary capabilities of visibility, flexibility, risk management, and collaboration may therefore experience the “paradox” not because digital investments are ineffective, but because the mediating capability infrastructure is underdeveloped.

Second, our findings are consistent with and extend the application of the Resource-Based View to digital transformation in supply chains. The strong path coefficients linking digital maturity to the four dynamic capabilities (β = 0.406–0.568) are consistent with characterizing digital maturity as a valuable, rare, and imperfectly imitable resource [52]. The variation in effect sizes across capabilities suggests that the VRIN characteristics of digital maturity are expressed differently: the strongest resource effect operates through visibility, in which digital technologies are associated with information advantages that competitors find particularly difficult to replicate.

Third, the results are consistent with and extend the Dynamic Capabilities framework as applied to supply chain resilience in an emerging market context [24,25,50,55]. Our four mediating capabilities map naturally onto Teece’s sensing–seizing–transforming taxonomy: visibility and risk management correspond to sensing (detecting environmental changes and threats), flexibility corresponds to seizing (mobilizing resources to respond), and collaboration corresponds to transforming (reconfiguring inter-organizational relationships). The finding that all four mediating paths are significant is consistent with the Dynamic Capabilities framework’s core proposition that performance outcomes emerge from the interplay of multiple capability dimensions rather than any single capability.

Fourth, the partial mediation pattern itself (VAF = 63.7%) has theoretical weight. Because the direct path from digital maturity to resilience remains significant (β = 0.219), some aspects of digital capability appear to be positively associated with resilience through routes our four mediators do not capture, for instance, by stabilizing day-to-day operations, automating routine disruption responses, or accelerating organizational learning [14]. Identifying these additional channels is a promising avenue for future research.

Fifth, the relatively even distribution of mediation shares across the four capabilities (ranging from 13.8% to 18.9% of the total effect) challenges the common practice of studying one capability at a time. In our data, supply chain resilience in an emerging-market context is genuinely multi-dimensional; no single capability dominates the mediation structure. This pattern is consistent with arguments put forward by Tukamuhabwa et al. [38] and with recent evidence from comparable settings [25,50,61].

6.2. Managerial Implications

What practical priorities can managers infer from these results? Building on the empirical evidence, we offer action recommendations that take account of the specific circumstances facing firms in Morocco and, more broadly, in comparable emerging-market settings.

The visibility result deserves special attention. The path from digital maturity to visibility (β = 0.568) is considerably stronger than the path to any other capability given the highest observed return per unit of digital investment (Figure 4). For Moroccan firms at early or intermediate stages of digital transformation, this suggests that the infrastructure associated with greater visibility may merit priority, given the highest observed return per unit of digital investment.

Flexibility and risk management are complementary areas of investment. The path coefficients (β = 0.478–0.480 from digital maturity; β = 0.188–0.200 to resilience) indicate that both capabilities are positively associated with resilience. Predictive analytics for demand planning and supplier risk scoring serve both capabilities at once and therefore represent a complementary capability area within the same priority set, rather than a subsequent step in an investment sequence as indicated in Figure 5.

Collaboration platforms may warrant a more careful implementation strategy. The somewhat weaker path from digital maturity to collaboration (β = 0.406) likely reflects that strong partnerships depend as much on trust and commitment as on technology. Managers may benefit from pairing digital collaboration tools with relationship-building initiatives and clearly defined data-sharing protocols. As depicted in Figure 4, this framework illustrates the relative empirical salience of the four mediating pathways observed in our cross-sectional sample. Highest associative weight: technologies associated with stronger visibility (β = 0.568). Intermediate associative weight: predictive analytics and risk management. Lower associative weight: collaborative platforms and partner integration.

A final point: average digital maturity in our sample was 3.29 out of 5; average resilience was 3.36. Both figures are moderate and indicate ample room for growth. As the Digital Morocco 2030 strategy [15] channels public funds into digital infrastructure, firms whose digital capability investments are distributed in line with the relative empirical salience observed in our sample, with visibility carrying the strongest associative weight, alongside flexibility, risk management, and partner collaboration, may report stronger firm-level resilience profiles. The ordering across capabilities reflects observed association strengths, not a validated implementation sequence.

Beyond firm-level recommendations, our findings also have implications for Moroccan policymakers. Three potential priorities may be considered in line with the Plan d’Accélération Industrielle (PAI): SME-oriented digital-skills programs could usefully target the visibility and flexibility levers; public investments in logistics corridors (Tanger Med, Casa-Settat, the Atlantic axis) may include resilience-by-design clauses; and industrial-zone tenders could reward firms that can demonstrate measurable digital-maturity progress.

6.3. Limitations and Future Research

Several limitations warrant acknowledgment. We are reporting associations rather than confirmed causal effects. Longitudinal or panel data would be needed to establish whether changes in digital maturity truly precede changes in dynamic capabilities and resilience.

Second, self-reported data from a single informant raises common-method concerns. We implemented multiple safeguards (randomized item order, anonymity assurance, Harman’s single-factor test, and full collinearity VIF assessment). Still, these cannot fully eliminate the possibility of common method bias. Dyadic or multi-informant designs would strengthen future investigations.

Third, our sample is limited to Morocco. Whether the structural relationships generalize to other emerging markets remains an open question. Cross-country comparative studies examining how institutional context moderates the relationship between digital maturity and resilience would be especially valuable.

Fourth, we measured digital maturity as an aggregate construct. Future studies could examine whether blockchain, digital twins, IoT, or AI contribute differentially to resilience, providing more actionable guidance for technology investment decisions.

Sustainability scope. The model deliberately retains a focused operational view of resilience and does not directly model environmental and social sustainability outcomes. As Elkington’sTriple Bottom Line [64] and Carter and Rogers’ sustainable supply-chain framework [8] emphasize, environmental and social capabilities may interact with operational resilience in ways our specification cannot capture. Future research should integrate TBL constructs alongside the four operational dynamic capabilities.

An immediate avenue for future research is to extend the present model with an explicit Triple Bottom Line lens, treating environmental performance (e.g., carbon intensity, waste reduction) and social performance (e.g., labor stability, community impact) as distal outcomes of digital maturity and the four dynamic capabilities. Cross-country replications combining digital maturity, dynamic capabilities, and TBL constructs would substantially advance the conversation.

7. Conclusions

Is digital maturity associated with supply chain resilience? For the Moroccan firms in our sample, the answer is affirmative within the limits of a cross-sectional design. However, the underlying pattern is more nuanced and more instructive than a simple yes-or-no might convey.

The total effect of digital maturity on supply chain resilience (β = 0.602) is sizable, indicating that higher levels of digital maturity are associated with greater resilience. Yet 63.7% of this association travels through four dynamic capabilities (visibility, flexibility, risk management, and collaboration) rather than working directly. These findings suggest that digital maturity, taken alone, shows only a limited statistical association with resilience; this association is mediated by the dynamic capabilities of visibility, flexibility, risk management, and collaboration.

Of the four mediating capabilities, visibility emerges as the most strongly associated mediator. Digital maturity’s link to visibility (β = 0.568) is markedly stronger than its link to the other three mediators, and visibility accounts for the largest share of the indirect effect (18.9%). For managers in emerging markets, this points to a clear differential in empirical salience among the four pathways observed in our cross-sectional sample: investments in supply chain transparency tools (tracking tools, integrated dashboards, real-time monitoring) are statistically associated with the strongest mediating channel, while predictive analytics, risk management software, and collaboration platforms are statistically associated with comparatively smaller mediating contributions.

In sum, this study contributes to supply chain resilience research in three ways: it documents an emerging-market empirical setting (R² = 0.523); it illustrates the use of VAF decomposition to disaggregate the relative weight of four capability pathways; and it offers empirically grounded suggestions for firms navigating the Industry 4.0 transition under resource constraints. Because the data are cross-sectional. Understanding how digital capabilities are statistically associated with operational resiliency through specific and measurable dynamic capabilities remains an open avenue for both academic inquiry and managerial reflection in Morocco and comparable emerging economies.