Mission command and supply chain resilience in wartime military logistics: empirical evidence from the Ukrainian Armed Forces

Supply chain resilience (SCRES) has become an increasingly important research field (Han et al., 2020; Katsaliaki et al., 2021). As a concept that goes beyond traditional risk management, resilience is the adaptive capability of supply chains to prepare for, respond to and recover from unexpected events (Ponomarov and Holcomb, 2009).

Although SCRES has been studied extensively in commercial contexts, less attention has been paid to the military context (Adobor, 2019; Lucas et al., 2024). Defense organizations operate within national political structures, which adds complexity to resource coordination. Integrating both military and civilian resources is paramount to meet strategic goals of deterring an adversary and sustaining defense if deterrence fails (Ekström, 2023; Keenan et al., 2024). In military operations, deliberate lethal actions by an adversary increase the complexity of disruptions (Ekström et al., 2020; Sani et al., 2022); the closer the logistics flow is to the frontline, the higher the risk of destruction. These challenges are prevalent in the war in Ukraine, where supply chains themselves are strategic targets. Supply chain resilience thus becomes a decisive factor in operational endurance (Lucas et al., 2024). Despite its importance, the command and leadership dimensions enabling SCRES in wartime operations remain underexplored.

To address this gap, we examine command and control (C2) as the organizational mechanism through which dynamic capabilities (DC) – sensing, seizing and reconfiguring resources – are enacted to enable resilience (Nair and Reed-Tsochas, 2019; Stadtfeld and Gruchmann, 2024; Teece, 2018), and promote adaptability under pressure (Knevelsrud et al., 2024). This study examines C2 as an organizational and leadership practice rather than physical infrastructure or frontline logistics flows.

In NATO doctrine, mission command (MC) emphasizes decentralized decision-making, initiative and trust-based leadership (Sjøgren and Nilsson, 2025), qualities that align with SCRES enablers such as flexibility and agility (Pettit et al., 2010). Since 2014, Ukraine has increasingly adopted NATO-style MC (Junius, 2024; Milczanowski, 2024), marking a cultural shift from Soviet traditions (Kliuchnikov, 2014; Moreno, 2024). Complementing MC, leader–member exchange (LMX) captures the relational trust essential for effective decentralized leadership, while visibility (VIS) provides situational awareness critical for adaptive logistics and planning (PLAN) supports the proactive and preparatory dimensions of SCRES.

With this background, we developed and tested a partial least squares structural equation model (PLS-SEM) combined with necessary condition analysis (NCA) to examine how these factors shape SCRES under wartime conditions. This study makes three principal contributions. First, it shows that command-and-control practices and MC specifically help explain resilience in a wartime military supply chain, a context rarely accessible to academic research. Second, it demonstrates that MC affects recovery primarily indirectly, through responsiveness, rather than through a direct effect. Third, by combining PLS-SEM with NCA, it distinguishes factors that improve resilience on average from factors whose minimum presence is required before high resilience becomes achievable.

The article is structured as follows. Section 2 develops the theoretical underpinnings and research model. Section 3 presents the research design and model development. Section 4 reports the measurement model analysis, followed by the results in Section 5. Section 6 discusses theoretical contributions and managerial implications, and Section 7 concludes.

This study is grounded in the resource-based view (RBV) and DC theory, which together provide the overarching theoretical lens. RBV identifies organizational capabilities as the foundation for resilience, while DC theory extends this by explaining how organizations sense, seize and reconfigure those capabilities under pressure (Teece, 2018). Antecedents of resilience are organizational capabilities (Stadtfeld and Gruchmann, 2024; Teece, 2018) enacted through mechanisms such as command and control, which together give rise to supply chain resilience (SCRES) as an emergent organizational property (Summers, 2018) that arises from a variety of widely recognized enabling capabilities and antecedents such as flexibility, visibility, collaboration, trust and redundancy (Chowdhury et al., 2019; Dubey et al., 2019; Leoni et al., 2024; Pettit et al., 2010; Ponis and Koronis, 2012; Wieland and Wallenburg, 2013).

A well-defined research area (Katsaliaki et al., 2021), SCRES research spans multiple phenomena and modeling techniques (Joshi and Luong, 2022). Influential models converge on three core phases or dimensions through which this emergent resilience capacity is realized (Chowdhury and Quaddus, 2017; Han et al., 2020; Pettit et al., 2010; Ponomarov and Holcomb, 2009):

1. Planning (PLAN): The proactive capability to anticipate disruptions and prepare contingency options (absorptive capacity).

2. Responsiveness (RESP): The ability to react swiftly and effectively when disruptions occur (adaptive capacity).

3. Recovery (REC): Restoring operations to a desired performance level, ideally with improved performance (restorative capacity).

This dynamic and multidimensional conceptualization of SCRES reflects adaptive capacity rather than static robustness (Adobor and McMullen, 2018). How these phases manifest depends on the operational context in which resilience is required.

Military supply chains operate within a distinct context that demands resilience frameworks tailored to their specific needs (Ekström et al., 2020; Sani et al., 2022, 2023). Military operations are inherently complex, requiring coordination across organizational, technological, strategic and political dimensions, and integration with commercial partners to sustain resource flow (Ekström, 2025). However, commercial and military supply chains are intertwined in complex networks, and researchers in military logistics can draw on a broad spectrum of SCRES literature (see, e.g. Cabrera et al., 2023; Dubey et al., 2019; Lucas et al., 2024; Pettit et al., 2010; Sani et al., 2023; Sokri, 2014; Solis et al., 2023).

Although the conduct of war is ever-evolving, its fundamental nature can be summarized by the Clausewitzian term friction, which describes the brutal gap between planning and conducting war (Clausewitz, 1989). For defense supply chains, this friction manifests as disruption risks that SCRES mechanisms must overcome. Since logistics are integral to military strategy (Kinsey and Ti, 2023; Skoglund et al., 2022), and friction cannot be eliminated, operational endurance depends on resilient supply chains tailored to wartime conditions. This situation underscores why SCRES in military operations must be examined not only as a technical, structural and engineering issue focusing on the physical means of restoring supply, such as redundancy and safety stocks, but also as a social-ecological capacity reflecting how the organization structures decision-making and action (Wieland and Durach, 2021). The two perspectives complement one another, and this study adopts this social-ecological approach, focusing on aspects of C2, including decentralized decision making and leadership (Ti, 2025).

Similarly, Ti (2025) emphasizes the importance of maintaining control and oversight of logistics networks. For military practitioners, command-and-control enables the sensing, seizing and reconfiguration of resources. Applied to the military context, SCRES is here understood as the capacity to absorb, adapt and recover from disruptions, while maintaining an appropriate ability to plan and conduct military operations (Summers, 2018, p. 98).

In wartime logistics, command-and-control is central to managing friction and enabling resilience. MC, NATO's joint command philosophy, is particularly relevant (Sjøgren and Nilsson, 2025). MC is generally acknowledged as effective in demanding situations, relying on core psychological needs like autonomy, competence and relationships (Knevelsrud et al., 2024). Derived from the German Auftragstaktik, MC is a normative approach that promotes efficient command by enabling subordinates' rapid decision-making and establishing decentralized execution in uncertain environments (Knevelsrud et al., 2024). MC is mostly discussed as a general leadership philosophy, and its effects on adaptive supply networks remain underexplored (Ben-Shalom and Shamir, 2011; Knevelsrud et al., 2024; Ploumis, 2020). As the organizational mechanism through which C2 is enacted, MC allows subordinates to sense disruptions, seize adaptive options and reconfigure resources in a decentralized manner with direct implications for responsiveness (RESP) and recovery (REC).

A cornerstone of MC is trust, defined as the shared confidence allowing subordinates to depend on each other while executing mission orders (Ploumis, 2020). Trust is a recurring theme in military leadership effectiveness (Bergh and Boe, 2018; Bjørnstad and Ulleberg, 2021; Fors Brandebo et al., 2013; Nazri and Rudi, 2019), and is emphasized in both training and combat contexts (Tollefson, 2016, p. 17; US Department of the Army, 2019). To capture the relational dimension, LMX offers a theoretical lens that conceptualizes leadership in terms of the quality of dyadic relationships (Graen and Uhl-Bien, 1995). Conversely, lack of trust is a real-world barrier to collaboration and resilience in military supply chains (Ekström, 2025). High-quality LMX relationships provide the interpersonal trust and mutual understanding that allow MC principles to function in practice: subordinates act on delegated authority when relational confidence between leader and subordinate is sufficiently established (Ploumis, 2020). LMX therefore operates as a relational foundation for MC and may also directly support adaptive action by reducing friction and enabling coordinated response under disruption. Whether this influence operates through MC, independently or both, is examined empirically in this study.

Although LMX has been criticized due to methodological concerns (Gottfredson et al., 2020), its continued use is justified by broad empirical support regarding team performance and citizenship behavior (Bauer and Erdogan, 2016; Dulebohn et al., 2012). In this study, LMX provides a well-validated measure of the relational capital required for MC to function effectively. Empirical evidence supports the proposition that LMX-based leadership strengthens strategic resilience capabilities (Shin and Park, 2021), supporting its inclusion as a relational antecedent of MC and, through MC, of responsiveness (RESP).

Military supply chains depend on commercial partners whose business logics differ from those of defense organizations, creating inherent visibility challenges (Ekström, 2025). Adversaries deliberately target logistics assets and infrastructure, making defense organizations reluctant to share sensitive logistics-related information. These constraints make supply chain-wide visibility (VIS) particularly challenging while simultaneously making it essential for resilience.

VIS refers to the timely and accurate monitoring of resources and material flows across the network, providing the situational awareness that underpins SCRES (Han et al., 2020; Pettit et al., 2010; Sá et al., 2019). In military contexts, VIS encompasses both tactical-level awareness of resource flows and strategic-level control over supply chain structures. Although the information age has introduced tension between centralized VIS and the decentralized nature of MC (Storr, 2003), NATO doctrine reconciles this by framing VIS as a command enabler rather than an obstacle. Timely, accurate VIS empowers autonomous action within the commander's intent, reducing the need for centralized control (Ti, 2025). For defense planners, improving VIS is critical for enabling SCRES.

Planning is a proactive organizational capability that prepares organizations to respond effectively under disruptions. By developing contingency options and resource allocation strategies, PLAN enables the sensing and preparation dimensions of resilience, creating the preconditions for decision-making and adaptive action (Han et al., 2020; Sani et al., 2022). Together, VIS and PLAN function as structural enablers that support the adaptive capacity established through MC and LMX.

Building on these theoretical foundations, we derive hypotheses examining how MC, LMX, VIS and PLAN interact to enable RESP and REC, the core variables representing SCRES in military operations. The following section presents the research model and the PLS-SEM approach used to test these relationships empirically.

All items were drawn from a broader questionnaire developed and validated for a related Norwegian Armed Forces (NAF) study (Elvemo, 2025), which combined a literature-based review of SCRES antecedents, principally Pettit et al. (2010, 2013) and Han et al. (2020), with iterative subject matter expert (SME) interviews and pilot testing. Figure 1 illustrates this process in full, spanning the original NAF instrument development summarized above and its subsequent adaptation for the present study, described below (see supplementary Material, Survey Instrument). LMX items captured subordinates' perceptions of the relationship with their superior, consistent with established practice using the LMX-7 tool (Graen and Uhl-Bien, 1995). The instrument was shortened from the NAF training-exercise version to reflect active wartime conditions, and professionally translated and adapted to the Ukrainian context, and verified through back-translation into Norwegian. An interim reliability check was conducted after the first ten responses were collected, confirming acceptable performance before continuing data collection with the instrument unchanged. Unlike the original NAF instrument, the shortened and translated version was not subject to a formal pretest prior to full deployment; measurement quality was instead assessed through this interim check and the post hoc item-loading analysis reported in Section 4.

Data were analyzed in SmartPLS4 (Ringle et al., 2024), following the recommended procedures for PLS-SEM (Hair et al., 2019), selected for its suitability in models with predictive orientation, multiple latent constructs and modest sample sizes. NCA was applied as a complementary method to identify threshold prerequisites, following Richter et al. (2020) and Dul (2016). Items with loadings below 0.7 were removed during measurement refinement (Section 4) with one exception: LMX1 (loading 0.690) was retained to preserve the integrity of the validated LMX-7 instrument; removing it also reduced AVE for the construct.

The target population comprised military logistics experts within the Ukrainian Armed Forces (UAF), yielding a sample size of n = 82. Inclusion criteria targeted UAF personnel with command-and-control experience in logistics roles spanning tactical, operational and strategic levels, with no restrictions on age or defense branch. A purposive sampling approach was used to recruit only Ukrainian personnel. Approximately 60 responses were obtained through the networks of the Ukrainian co-author, while Norwegian liaison officers assisted in further distribution of the survey. Respondents' average service length was 15.4 years. All 82 responses were complete and useable. The absence of a defined sampling frame precludes a formal response rate and prevents assessment of nonresponse bias. Findings should be interpreted with this in mind.

All constructs were measured on a 7-point Likert scale (e.g. 1 = strongly disagree to 7 = strongly agree), with higher scores indicating greater realized capabilities for each variable.

Figure 2 presents the structural model. RESP and REC represent the core resilience capabilities, with REC emphasized as the ultimate outcome. PLAN, MC, LMX and VIS are treated as antecedent variables hypothesized to positively affect these capabilities. Taken together, the model specifies a hypothesized chain of relationships in which LMX supports MC; MC, PLAN and VIS enhance RESP; and RESP affects REC.

MC is the organizational mechanism through which decentralized adaptation is enacted in the face of friction and uncertainty (Ben-Shalom and Shamir, 2011; Knevelsrud et al., 2024). While RESP captures the speed and agility of initial adaptive action, REC requires the sustained reconfiguration of resources and capabilities over time. MC's emphasis on decentralized initiatives and trust enables this reconfiguration dimension, suggesting a direct effect on REC beyond mediation through RESP alone.

MC's attributes are expected to enhance RESP by empowering actors to interpret and respond to disruptions without waiting for centralized directions. Empirical evidence from NATO military supply networks shows that decentralized command structures improve adaptability and RESP under operational stress (Ti, 2025).

While MC captures the doctrinal side of trust-based leadership, LMX theory provides a relational lens reflecting the quality of the relationship between leaders and subordinates (Dulebohn et al., 2012; Graen and Uhl-Bien, 1995), which also affects resilience (Shin and Park, 2021). Since MC relies on these same qualities to function effectively in decentralized settings, LMX is expected to promote MC by fostering the interpersonal trust required for intent-based leadership.

High-quality LMX relationships may also directly enhance RESP by reducing the friction of acting under disruption. Where relational trust and mutual obligation are strong, coordinated adaptive action becomes more fluid and less dependent on formal authorization. While LMX is expected to operate primarily through MC, this relational mechanism may support RESP independently and is examined here as a separate hypothesis (Shin and Park, 2021).

We also explore whether LMX relationships moderate the effect of MC on RESP. While H2.1 captures LMX as a relational foundation for MC, the strength of that foundation may affect how MC principles translate into responsive action. When LMX quality is high, subordinates are more likely to act, amplifying MC's effect on RESP. When LMX quality is low, MC principles may be constrained. This interaction effect is distinct from the sequential LMX → MC pathway, justifying a moderation hypothesis.

VIS refers to the timely and accurate flow of information on resources, actors and material flows across the supply network. In DC terms, VIS enables the sensing dimension of resilience: without accurate situational awareness, organizations cannot detect disruptions early enough to initiate adaptive action, doing so is particularly critical in defense contexts, where supply chain opacity (Ekström, 2025) requires decentralized units to maintain situational awareness to act autonomously (Ti, 2025). Prior research consistently identifies VIS as a structural enabler of RESP (Han et al., 2020; Pettit et al., 2010; Sá et al., 2019).

Planning (PLAN) involves the development of readiness and contingency options that prepare organizations to respond effectively. By enabling structured anticipation and coordinated action, planning is a proactive capability that creates the necessary conditions for a swift response (Sani et al., 2022; Summers, 2018). It is therefore considered a key enabler of RESP (Han et al., 2020; Pettit et al., 2010).

RESP refers to the ability to convert available options into timely and effective action. In DC terms, RESP represents the seizing dimension of resilience, enabling actions that establish the organizational conditions required for reconfiguration and recovery (Teece, 2018). RESP is therefore recognized both theoretically and empirically as the critical pathway to REC (Han et al., 2020; Pettit et al., 2010).

LMX and MC are modeled on theoretical foundations and subject-matter expert input. While reciprocal effects may exist in practice (Gottfredson et al., 2020), this study adopts a causal-predictive rather than a strictly explanatory orientation (Hult et al., 2018; Wulff et al., 2023). In this context, endogeneity assessment serves model credibility rather than causal identification, and PLS-SEM's predictive orientation mitigates but does not eliminate this concern (Hair et al., 2019). The findings should therefore be interpreted as predictive rather than strictly causal.

We analyzed the survey data using SmartPLS (Ringle et al., 2024), following recommended procedures for PLS-SEM (Hair et al., 2019). Statistical power was assessed using G*Power (Faul et al., 2007) with an assumed effect size of 0.155, a significance level of 5%, and a desired power of 0.80. Under these conditions, the required sample size aligns with the 82 valid responses collected. This adequacy is supported by the inverse square root method (Kock and Hadaya, 2018), confirming the ability to detect path coefficients as small as 0.11. Although the sample is small relative to the population, this robust justification helps mitigate the risk of Type II errors. The resulting values (Table 1) offer an initial indication of the model's explanatory power.

We assessed reliability and validity according to recommended thresholds (Hair et al., 2019). Convergent validity (AVE) and internal consistency (composite reliability) were satisfied for all constructs (Table 2). Discriminant validity was assessed using the Fornell-Larcker criterion (Table 3) and heterotrait–monotrait (HTMT) ratios (Table 4), which were below conservative limits except for RESP vs MC (HTMT 0.896, Fornell–Larcker 0.787). HTMT remained below the more liberal 0.90 threshold recommended for closely related constructs (Henseler et al., 2015; Voorhees et al., 2016). A bootstrapped 95% confidence interval for the MC-RESP HTMT value [0.792, 1.008] includes 1.000, so discriminant validity cannot be statistically confirmed at this threshold (see Supplementary Table 1 for full HTMT confidence intervals). This is theoretically consistent with the sequential relationship, in which MC is hypothesized to precede RESP, and is treated as a measurement limitation affecting interpretation of findings.

Common method bias (collinearity) was assessed using variance inflation factors (VIFs). As shown in Table 5, all values were below the conservative cut-off of 3.0 (Becker et al., 2015; Kock, 2015).

Predictive validity was confirmed using Stone-Geisser's Q² via blindfolding, with all values above zero, supporting predictive relevance. Out-of-sample performance was further assessed using PLSpredict (Shmueli et al., 2019), comparing root mean square error (RMSE) values from the PLS-SEM model with those from a linear regression benchmark (LM). Across most indicators, the PLS-SEM model produced lower RMSE values than the LM benchmark, indicating stronger predictive performance, particularly for the RESP and REC items (Table 6). MC items showed comparatively weaker predictive relevance than RESP items, consistent with the measurement caution already noted for this construct (Section 4).

We estimated the model using 10,000 bootstrap samples, following Becker et al. (2023). Figure 3 illustrates the model with path coefficients and R² values, while Table 7 summarizes hypothesis testing results, including effect sizes (f²) assessed following thresholds of 0.02 (small), 0.15 (medium) and 0.35 (large) (Hair et al., 2019). All outer loadings were statistically significant (t > 1.96, p < 0.05), confirming indicator reliability and supporting the measurement model (Hair et al., 2019). Four of the eight hypothesized paths were supported (Table 7). MC, planning and LMX through MC were positively associated with the model's outcomes, with RESP showing the strongest association with REC. VIS and LMX had no significant direct association with RESP; MC had no significant direct association with REC; and the LMX × MC interaction was nonsignificant. These findings are consistent with MC's association with REC operating through RESP and LMX's association with RESP operating through MC.

The mediation analysis was added following reviewer feedback on earlier version of this manuscript, prompted by MC's nonsignificant direct effect on REC. Theoretically and contextually MC creates the doctrinal foundation, and RESP is the executed action through which REC is achieved. Similarly, MC is expected to mediate the relationship between LMX and RESP: LMX provides the relational trust that allows MC to be enacted (Section 2.2.1), but it is MC's decentralized doctrinal practices that translate into the adaptive, responsive action RESP captures. Given the cross-sectional, self-reported nature of the data, the results below are described as consistent with mediation rather than as evidence of a causal process. Bootstrapped indirect effects were estimated with bias-corrected 95% confidence intervals following Nitzl et al. (2016) and Preacher and Hayes (2008). The results are depicted in Table 8.

The MC → RESP → REC pathway (β = 0.259, CI [0.112, 0.458]) is significant, while the direct MC → REC path is nonsignificant, indicating that MC's association with REC is primarily through RESP. Given the discriminant validity limitation (Section 6.1), this pathway needs cautious interpretation, as the results may reflect closeness between MC and RESP. The indirect LMX → MC → RESP and LMX → MC → RESP → REC pathways, as well as PLAN → RESP → REC, were significant; the remaining indirect paths were nonsignificant.

We applied NCA to assess which antecedents act as prerequisites for achieving high levels of REC. It also complements the mediation analysis by assessing whether RESP is a necessary condition for REC. NCA identifies necessary conditions through effect sizes, consistency thresholds and permutation-based significance testing, following Richter et al. (2020). Effect sizes (d) were interpreted according to Dul's (2016) guidelines:

1. d < 0.1 = small

2. 0.1 ≤ d < 0.3 = medium

3. 0.3 ≤ d < 0.5 = large

4. d ≥ 0.5 = very large.

Table 9 summarizes both SEM significance and NCA necessity. For REC, RESP emerged as both significant and necessary, while MC was necessary despite a nonsignificant direct effect in SEM. VIS showed a small but significant necessity effect, whereas LMX was not necessary. Blank values in the β, t and p estimations indicate paths not tested in the PLS-SEM structural model but evaluated with NCA.

For RESP, PLAN and MC were both significant and necessary. VIS appeared necessary despite weak or nonsignificant direct effects, indicating it constrains RESP at higher levels. LMX was neither a significant predictor nor a necessary condition for RESP.

Taken together, these findings provide a nuanced picture: RESP was both positively associated with and a necessary condition for REC, while PLAN, MC and VIS exhibited sample-specific necessary-condition thresholds for RESP and/or REC.

Table 10 (bottleneck table) shows the minimum levels of each antecedent required to achieve increasing levels of REC. While lower REC levels can be achieved without strong antecedents, higher levels require progressively greater thresholds of RESP, PLAN and MC, with VIS contributing at advanced levels. Achieving 90% REC performance requires minimum scores of 4.847 for MC and 5.088 for planning on the original 1–7 Likert scale. NN indicates no necessary level at this point.

To complement the structural model, we applied an importance-performance map analysis (IPMA) following Ringle and Sarstedt (2016). IPMA evaluates each antecedent along two dimensions: importance, reflecting total effect on the outcome and performance, reflecting realized capability on a 0–100 scale. Figure 4 presents the IPMA results. For context, the overall performance of REC was 77.076.

The IPMA results highlight clear differences among the antecedents of REC. RESP shows the highest importance (0.542) and a strong performance level (74.079), consistent with its role affecting REC. MC also emerges as an important factor (0.382), but with lower performance (69.127), indicating potential for improvement. Planning (importance 0.189, performance 69.276) and LMX (importance 0.179, performance 71.171) contribute moderately, with similar performance levels. VIS ranks lowest in importance (0.073) and shows the weakest performance (55.443), suggesting it is less central for REC but represents a structural weakness that could constrain resilience if left unaddressed.

Returning to the study's three contributions, PLS-SEM and NCA together reveal a distinction between predictors and constraints in SCRES that recurs throughout the constructs discussed below. While PLS-SEM identifies variables that improve resilience on average, NCA highlights conditions that do not raise performance themselves, but whose absence prevents high performance. In our model, VIS illustrates this pattern for RESP: it lacks a significant direct effect, yet NCA shows that insufficient VIS imposes upper bounds on achievable RESP. LMX, in contrast, is neither a significant predictor nor a necessary condition for RESP, indicating that its influence on responsiveness operates through MC.

The structural model supports the multi-phase conceptualization of resilience. RESP emerges as the adaptive capacity that converts other antecedents into REC, representing the strongest and most significant pathway to resilience outcomes: a significant predictor of REC, and, per the NCA, a necessary condition for it. This dual role highlights the dynamic nature of resilience as an emergent property of organizations. Although planning significantly improves RESP (H4 supported), SCRES is not just about preparation for a static state of robustness but also about converting preparedness into adaptive action. The explanatory power of the model (R² = 0.705 for RESP; R² = 0.415 for REC) indicates that the selected leadership and structural antecedents account for considerable variance in resilience capabilities within this study's context.

MC does not show a significant direct effect on REC. However, it emerges as a systemic prerequisite for REC through its substantial influence on RESP. This finding underscores MC's indirect role in resilience: while it does not explain additional variance in REC beyond its mediated effect, high levels of REC cannot be achieved without meeting a minimum threshold of MC. Without sufficient MC, the organization's capacity to adapt and recover can be constrained regardless of other antecedents.

The model provides a theoretically grounded yet narrowed phase-perspective on resilience, and its limitations must be acknowledged. Most notably, unconfirmed discriminant validity between MC and RESP is a measurement limitation that calls for careful interpretation of findings involving these two constructs. Future research might explore MC and RESP further.

An important theoretical contribution is the empirical support of the theoretical linkage between LMX and MC. LMX was strongly and significantly associated with MC and was also identified by NCA as a necessary condition for MC, suggesting that trust-based leadership is a prerequisite for the effective enactment of MC principles. LMX's role in RESP is more limited: neither the direct path nor the hypothesized moderating effect was significant. The nonsignificant LMX × MC interaction (H2.3) suggests LMX may function as a threshold for MC rather than a continuous moderator of its effect. NCA indicates LMX is neither a necessary condition for RESP nor a significant predictor, indicating LMX influence on RESP is primarily as a foundational enabler through MC. This reading should take the MC-RESP limitation noted above into consideration.

Integrating sufficiency and necessity perspectives clarifies the distinct functional roles of PLAN and VIS. Planning shows a strong and necessary effect on RESP, supporting its role as a structural enabler of adaptive response.

In contrast, VIS shows no significant direct association with RESP in the SEM yet emerges as a necessary condition for RESP. This discrepancy may reflect model limitations in capturing VIS's full theoretical importance or contextual factors specific to this sample. VIS exhibits the lowest realized performance in the IPMA, suggesting an underdeveloped structural capability. In this context, VIS represents a potential bottleneck whose improvement could unlock greater resilience. The relatively high REC performance observed may reflect compensatory factors such as physical resource buffers and redundancies that mitigate low information quality.

The overall picture of our findings is that RESP and MC together represent the central managerial priority for REC. RESP combines high importance with strong realized performance, while MC, closely linked to RESP in this study, plays a necessary role for REC, yet shows only moderate realized performance, indicating room for improvement.

LMX further reinforces this dynamic, serving as a prerequisite for MC, though it is not itself a necessary condition for REC. This finding underscores the need for development strategies that consider the principles of MC and LMX.

The high scores for MC and LMX are consistent with respondents' perception that logistics units have moved toward NATO command-and-control doctrine, suggesting a shift from a purely technical, flow-based function to one that also prioritizes adaptability and trust. The most critical managerial observation in the study is VIS, which ranks lowest in performance and importance for REC, yet NCA identifies it as a necessary constraint. This observation suggests a managerial priority to increase awareness of VIS as an important factor for SCRES.

Managers can use the NCA bottleneck analysis as a diagnostic tool, indicating the minimum levels of MC, planning and visibility associated with higher REC in this sample. These sample-specific reference points can help prioritize where to focus improvement efforts, shifting the discussion from conceptual recommendations to measurable, context-bound targets.

This study investigated the relationships between key antecedents of SCRES in the context of war, using data from the UAF analyzed through PLS-SEM and NCA. Three contributions follow. First, the findings show that command-and-control practices, and MC specifically, help explain resilience in a wartime military supply chain, a context rarely accessible to empirical research. Second, MC's effect on recovery operates primarily indirectly, through its close empirical and doctrinal relationship with RESP, rather than through a direct path; LMX supports MC, which in turn relates to RESP capability. Third, combining PLS-SEM with NCA distinguishes factors that improve resilience on average from factors whose minimum presence is required before high resilience becomes achievable.

The study's contextually adapted design and cross-sectional nature impose limitations on generalizability and on capturing the sequential logic of sense, seize and reconfigure. Future research should employ longitudinal designs and incorporate qualitative assessments to better explain findings such as the MC-RESP relation and VIS's low performance despite its role as a necessary constraint. Given the discriminant validity concerns for MC and RESP, an exploratory construct-development approach (e.g. EFA/CFA) may help clarify whether these constructs can be modeled otherwise. Additionally, planning as a preparatory activity warrants deeper investigation, beyond the scope of this study. This study applies a DC approach to the three-phase SCRES framework. Resilience can also be conceptualized through viability models, network-centric views and dynamic capability extensions emphasizing collaboration, redundancy and absorptive capacity; future research could pursue these avenues, potentially using qualitative methods to capture contextual nuances beyond quantitative models' reach. The single-source measurement of LMX from the subordinate perspective only, while standard LMX practice, is a further limitation.

Despite these limitations, respondents' reports are consistent with practices aligned with NATO's MC doctrine; this reflects perceived practice rather than evidence of doctrinal or cultural change. Whether these dynamics extend to humanitarian, commercial or other nonmilitary supply chains facing adversarial or resource-intensive disruption remains a question for future research rather than an implication demonstrated by the present data. Similarly, whether operational context shapes which behaviors are perceived as part of MC warrants further investigation. Whether UAF's battle-tested MC practices should be regarded as convergence toward an established NATO standard, or as an independently refined variant meriting NATO's own attention, remains an open question for future research.

The supplementary material for this article can be found online.