Supply Chain Resilience in Military Operations: A Case Study Exploring Command and Control

Developing a Contextual Model for SCRES

The theoretical foundations of this research – CAS and DC – are summarized in Table 1. PLS-SEM is well-suited to modeling complex systems with multiple interdependent variables, as it aligns with both CAS and DC theories. In my research model, CAS underpins the system-level perspective, while DC represents the adaptive capabilities.

The development of this research model was supported by semi-structured interviews with four experts on logistics from the Norwegian Armed Forces (NAF) and one associate professor specializing in management science and logistics. Each expert has more than 20 years of combined military and academic experience. All the military experts have relevant experience within logistics scenario training. Their contributions included refining the survey’s scenarios, variables, and items, as well as translation from English to Norwegian. The process of model development, up to data collection and analysis, is illustrated in Figure 1.

Figure 1

The antecedents of resilience that formed the basis for expert evaluation were those presented in the previous section by Han et. al. (2020) and Pettit et. al. (2010), which I have combined in Table 2.

To ensure shared situational awareness, the survey was built around a scenario commonly used in live and simulated training at both combat unit and C2 levels within the NAF. The scenario depicts full NAF deployment to northern Norway under conditions of high tension without active combat. The organizational structure is fully manned and equipped, reflecting the NAF’s 2023 political mandate, including staffing, capabilities, and equipment. The scenario targets the tactical and operational levels, with the operational level being the highest resource-allocating authority.

The strategic level, concerned with policy and capability-building, is indirectly addressed through the scenario’s assumption of full readiness. The abstraction level focuses on sustaining highly active military forces but excludes detailed modeling of ammunition, supplies, combat effects, or corrective maintenance.

The scenario reflects the operational framework based on political tasking, balanced with the need to ensure unclassified data collection that can be used for open publication. While such trade-offs are inherent in scenario-based research, the author managed them carefully to ensure conceptual clarity and practical relevance, and to preserve validity and reliability.

The SMEs and author reviewed the literature revealing the capabilities of SCRES in relation to the scenario. They then adjusted them to fit the NAF context, ensuring practical validity. Some variables were discarded, while others were combined to ensure a valid model. Key variables include recovery, which plays an important role in SCRES and distinguishes SCRES from risk management. Additionally, we (experts and I) identified responsiveness – encompassing flexibility and agility – as a primary variable. Since there is no single Norwegian term for agility, responsiveness was used to capture both agility and flexibility. Further elaboration on these variables is provided in the hypothesis development section.

Research Model Hypotheses

Figure 2 illustrates the variables and their relationships in the research model. All constructs are reflective (i.e., the items reflect the construct), consistent with experts’ feedback and the study’s exploratory approach. Expert involvement ensured alignment between practitioner perspectives and literature (Chowdhury & Quaddus, 2016; Han et al., 2020; Summers, 2018) addressing mitigation, preparation, readiness (H5, H6, H7), response (H4), and recovery (H3). To prevent construct misspecification, the model adhered to the recommendations made by Jarvis and colleagues (2003).

Figure 2

Instrument Design, Research Approach, and Data Collection

The survey indicators are developed from both existing literature (e.g., Pettit et al., 2013) and expert input. Indicators adapted to the NAF context were translated from English to Norwegian, and adjusted to ensure both academic rigor and practical relevance. The experts provided iterative feedback to ensure the survey’s real-world applicability, particularly in aligning theoretical constructs with operational realities. The survey covered 8 variables and presented 65 questions related to SCRES, along with additional items assessing variable relevance.

I conducted a pilot test with nine experts, four of whom were involved in the model’s development. Cronbach’s α values from the pilot ranged from 0.63 to 0.9, with an average of 0.74. Factor loadings averaged 0.68. Although some loadings were below the 0.7 threshold, all were above 0.4, which is acceptable for exploratory studies (Hulland, 1999). Convergent validity assessed through average variance extracted (AVE) was higher than 0.592 for all items. Based on these findings, I proceeded with the main data collection activities without altering the survey.

The survey was distributed during two large, combined NATO and NAF exercises that involved both live and simulated components. Of the estimated 150 respondents, 74 completed the survey (50 online and 24 on paper, later transcribed by researchers). Respondents averaged 19.6 years of experience across various commands (air, sea, land, and special forces). The importance of these exercises for operational training and certification ensured high respondent motivation. While the comprehensiveness of the survey may have affected response quality, the survey aligns with the methods of conducting and assessing military exercises for both training and certifying military units, including headquarters. The developed instruments thus bring SCRES into an already established practice.

Measurement Model Assessment with PLS-SEM

The sample size of 74 is about half of the estimated population, helping to minimize the risk of Type II errors. This sample size is sufficient to detect R-square of minimum 0.25 at 5% significance level and 80% statistical power. According to the inverse square root method of calculating recommended sample size using similar conditions, the recommended minimum sample size is 79 for a path coefficient of 0.28, which means that the model is less suited to detecting significant results with weaker effects. Additionally, using the analytical tool G*power, an effect size of 0.23 (between a medium size of 0.15 and a large size of 0.35) was calculated for a sample size of 74, indicating conservative estimation of detecting weak effects (Faul et al., 2007). However, these sample size calculations do not account for the sample-size-to-population ratio, which may mitigate the potential limitations. I will further discuss this aspect in the context of the model’s theoretical foundation and path coefficients in the discussion section. R-squared values are presented in Table 3.

The measurement model was tested for internal consistency, convergent validity, and discriminant validity. I measured internal consistency using ρA (rho_A [composite reliability]) with a threshold of ρA > 0.7. I evaluated the convergent validity by checking the AVE (Average Variance Extracted), with the recommended threshold being AVE > 0.5. I discarded indicators with loadings below 0.7, ensuring robustness. The model contains 46 items. The internal consistency of the model is assessed to be adequate based on these values, including the conventional threshold of Cronbach’s α and the composite reliability test. Table 4 shows descriptive statistics.

Discriminant validity refers to how distinct each construct is from others in the model. The Fornell-Larcker criterion is commonly used, which involves utilizing the square root of AVE to establish discriminant validity if this value exceeds the highest correlation with any other construct (Table 6). However, in variance-based modeling, the heterotrait-monotrait (HTMT) (Table 5) ratio as part of a confirmatory composite analysis (Hair et al., 2020) is recommended to better determine discriminant validity (Hair et al., 2019). Discriminant validity is generally established when the HTMT ratio is less than 0.85 for non-similar constructs (Voorhees et al., 2016). However, 0.90 is recommended as the threshold for closely related constructs, and values above 0.90 do not represent a definite violation of discriminant validity (Henseler et al., 2015).

All HTMT values are below 0.9; however, three relationships exceed 0.85: 0.898 (RESP-CAP), 0.854 (SCRES-CAP), and 0.867 (SCRES-RESP). This indicates some similarities between constructs within the context of this study, which is expected as the dependent variable SCRES is based on theoretical definitions that contain aspects of the constructs used in this model.

I also assessed the results for potential common method variance using variance inflation factor (VIF) values (Table 7). VIF values greater than three are proposed as an indication of pathological collinearity (Becker et al., 2015; Kock, 2015), which was not the case in this study.

I conducted hypothesis testing by analyzing the model in Figure 2, following the recommendations made by Hair et al. (2019). The predictive relevance of the model was confirmed using Stone-Geisser’s Q2 values; all values were above zero, indicating that the model is useful for understanding and explaining the constructs within the sample data. To further assess predictive accuracy, I compared mean absolute error (MEA) values between the PLS-SEM and a linear model (LM). The lower MEA in PLS-SEM relative to LM supports the model’s superior predictive accuracy (Table 8), in line with Shmueli et al. (2019).

The approximation of model-fit followed recommendations made by Henseler et al. (2016) and used the standardized root mean residual (SRMR) criterion, with a recommended threshold of SRMR < 0.08. The model’s SRMR value of 0.168 indicates a poor fit between the predicted and actual data. However, the use of SRMR as a fit criterion should be approached with caution (Hair et al., 2019). Although the sample size provides good coverage, its relatively small size may affect predictive accuracy due to increased variability and potential overfitting. I will evaluate the theoretical explanatory power in relation to the hypotheses, considering the possible predictive limitation.

Identifying Critical Factors: Necessary Condition Analysis (NCA)

This section applies necessary condition analysis (NCA) to identify, or rather indicate, essential factors for achieving SCRES. NCA evaluates effect sizes, consistency values, and statistical significance through permutation tests to determine which factors are necessary for resilience (Table 9). The NCA followed guidelines provided by Richter et al. (2020). Threshold values of effect sizes (d) follow recommendations made by Dul (2016):

NCA values denominations in Table 9:

• Original sample (O): This shows the effect size of the relationship between the exogenous and dependent variables.

• T statistics (|O/STDEV|): T-value derived from the effect size divided by its standard deviation, used to test the statistical significance of the relationship.

• P values: Indicate the probability of obtaining the observed results assuming that the null hypothesis is true (i.e., no effect). A p-value below 0.05 typically suggests that the effect is statistically significant (indicated in the table by red values).

• Original effect size: Indicates the strength of the relationship between variables.

• Permutation p value: This value tests the hypothesis that the observed relationship could be due to random chance; a red DISR (disruption) value indicates not necessary.

• Interpretation: Summarizes whether the exogenous variable is both statistically significant and necessary to achieve the dependent variable’s outcome.

Notice that the analytical model’s NCA and bottleneck table (Table 10) can result in somewhat contradictory values: PLAN scores as NN (not necessary) up to a 10% SCRES score in the bottleneck table, while the NCA table shows PLAN as both a significant and necessary condition. This must be seen considering LV SCRES scores up to 10% correlate with low scores given in the survey, indicating less agreement with questions related to the dependent variable. Of more interest is the need to increase PLAN to 2.674 in order to achieve a 60% SCRES score.

The NCA and bottleneck table illustrate that while basic levels of capacity (CAP) and other factors are sufficient for achieving lower levels of resilience, significantly higher levels of these variables are required to attain greater resilience. This pattern is consistent across most variables, underscoring the increasing importance of resources, planning, and responsiveness as SCRES goals become more ambitious. The variable DISR (indicating disruptions) should be interpreted carefully due to its role as a negative antecedent of SCRES and the non-significance of its score. Although VIS (visibility) and COOP (collaboration) are non-significant in terms of their scores, they remain necessary factors, indicating that they are essential components for achieving SCRES.

Structural Model Evaluation and Hypotheses Testing

This section presents the results of my analysis of the structural model analysis. I estimated the model using 10,000 bootstrap samples, as recommended by Becker et al. (2023). The results are illustrated in Figure 3 and followed by an importance-performance map analysis (IPMA; Figure 4), which provides the foundation for discussing both theoretical and managerial implications.

Figure 3

Figure 4

All indicators exhibit significantly high t-values, confirming the statistical significance of the factor loadings. All p-values on item loadings are 0.000, meaning that the relationships between the indicators and their corresponding latent constructs are statistically significant.

Importance-Performance Map Analysis (IPMA)

The importance-performance map analysis (IPMA) plots key constructs across two dimensions: importance (x-axis), based on the total effects of each construct on the dependent variable (DV); and performance (y-axis), derived from the average scores of each construct (converted to a percentage).

The DV SCRES performance is measured at 59.858, indicating that respondents perceive the level of SCRES at approximately 60%. The lower-right segment of the IPMA identifies areas with room for improvement. For example, improving RESP one point from 49.715 to 50.715 would increase SCRES performance by 0.693, raising it from 59.858 to 60.551. DISR is rated highly in terms of impact, but its importance is perceived as low, reflecting its expected negative effect. The IPMA shows that the respondents perceive RESP and REC as key antecedents of SCRES, while the hypothesized importance of VIS is not confirmed by the data.

Contributions to Theory

Researchers recommend approaching supply chain research through theories of complex adaptive systems and dynamic capabilities (Chowdhury & Quaddus, 2017; Nair & Reed-Tsochas, 2019). This study builds on such recommendations and offers a refined view of how resilience can be developed and maintained in military operations. It complements existing theoretical work by integrating CAS theory with dynamic capabilities (Nair & Reed-Tsochas, 2019; Stadtfeld & Gruchmann, 2024), operationalized through a novel, scenario-based, PLS-SEM model tailored to military logistics. It also extends the integration of command and control and SCRES in military contexts as an emergent property of underlying constructs (Summers, 2018). Responsiveness emerges as a key antecedent, linking preparedness factors such as planning and capacity-building with both direct and indirect effects on SCRES through recovery. These findings align with previous work in both commercial (Han et al., 2020; Pettit et al., 2010) and defense supply chains (Cabrera et al., 2023; Sani et al., 2022; 2023), and support the idea of responsiveness as a dynamic capability that enhances recovery (Teece, 2018). This underscores the adaptability of commercial theory for defense SC research when contextualized appropriately.

Building on Pettit et al.’s (2010; 2013) framework, this study examines capacity building and planning as distinct constructs. While planning in mentioned in resilience studies (e.g., Das, 2018; Helgeson & Roa-Henriquez, 2022), its role as a necessary condition in military operations represents a novel contribution. The study also adds contextual insights into visibility and cooperation (e.g., Liu & Lee, 2018; Scholten & Schilder, 2015), showing how these constructs may behave differently in military settings, although limitations remain due to scenario abstraction. Construct refinement is also highlighted: for instance, one item on mobility – initially part of capacity – was removed due to low loading (0.345), suggesting a distinction between physical mobility and other forms of resource capacity. This supports the argument that SCRES constructs are often operationalized too broadly. Scholten et al. (2019) similarly advocate for a more granular modeling of capabilities for non-routine events, distinguishing between learning routines that build different resilience types across disruption phases. Another low-loading item (0.455), related to learning and development, touches on resilience mechanisms beyond the traditional absorb–adapt–recover cycle. Its rejection may reflect conceptual gaps in operationalizing transformative capabilities in high-pressure settings, or suggest that such capacities operate at different organizational levels (Brusset & Teller, 2017; Scholten, et al., 2019).

While refining constructs may add complexity unsuited to exploratory modeling, it is nonetheless valuable. This is also true for disruptions (DISR), which scored highly in performance but low in importance and effect, representing that disruptions are recognized as affecting the supply chain. The effect score is (as expected) negative, due to the design of the model. However, the low importance score and low effect implies that there are other factors at play that the model does not detect. While Pettit et al. (2010) include disruption sensitivity in their framework, the weak effect observed here echoes the findings of Blackhurst and her colleagues (2011) that the impact of disruptions on a system varies depending on the level of resilience within the supply chain, and that disruption perception varies across contexts and often requires more granular, context-sensitive modeling approaches to capture operational realities.

Testing the model with reversed DISR values also showed that the absence of disruptions was not necessary for SCRES, according to the NCA. Disruptions can be elusive, especially in military supply networks where non-hub nodes may degrade system performance when compromised (Xiong et al., 2020). These findings support the need for more context-sensitive modeling. For example, Ivanov & Dolgui (2021) propose the use of digital twins, which could be combined with PLS-SEM. Transformative capabilities also warrant further exploration beyond the prepare–respond–recover framework (Brusset & Teller, 2017; Wieland & Durach, 2021).

Despite such challenges, the research shows that using antecedent constructs in a simulated scenario helps to shed light on SCRES’s overarching concepts, demonstrating that significance, necessity, and importance vary among the capabilities. Achieving certain levels of SCRES (60–70% and above) emerges from correspondingly increasing levels of most other capabilities. Despite the limitations, the findings offer valuable contributions to both practitioners and scholars of military logistics. However, this study also shows how SCRES theory offers broad perspectives, while its practical application often proves highly contextual and challenging to generalize.

Practical Implications for Management

The IPMA indicates that managerial priorities should focus on RESP, REC, and CAP, as these areas offer the greatest return on managerial effort. The role of VIS, COOP, and DISR may need to be further explored to better understand their relationship with SCRES. Managers at different levels should ensure sufficient contextual adaptation when focusing on these antecedents. Planning emerges as a necessary condition, implying that a balanced focus on antecedents throughout the planning process is crucial for achieving SCRES – at least above a certain level, as indicated by the bottleneck table. The model suggests that while management capabilities enhancing resilience are easier to assess, pinpointing meaningful aspects of disruptions remains more challenging, which might align with the elusiveness of black swans those highly unexpected events with significant impact.

Similar to commercial supply chains, military logistics must allocate limited resources to mitigate risks and strive to operate within a “Zone of Balanced Resilience,” as described by Pettit et al. (2010; 2013). As such, this research contributes to improved decision-making regarding military logistics, particularly within the Norwegian Armed Forces.

This study confirms the value of scenario-based training exercises, while also highlighting the potential benefit of integrating more SCRES-related challenges into these scenarios. Scenario design could play a crucial role: incorporating specific disruptions with clear consequences may help assess and train mitigation strategies across the three key phases of SCRES – i.e., preparation, response, and recovery. These exercises should delve deeply into potential problems and their solutions, enabling the development of strategies that enhance SCRES through its underlying factors.

This research addresses two core functions of military organizations: planning processes that anticipate hypothetical disruptions to the society they are tasked with defending, whether at home or abroad; and conducting exercises to ensure readiness for future missions.

Finally, this study also touches on the ever-present challenge commanders face in balancing attention to details with the broader picture. The proverbial nail for the want of which the battle was lost reminds managers of the butterfly or ripple effect and the importance of connecting strategy to even the smallest details. As Kane notes when he links military strategy and logistics: