The methodological process to prioritize labor protection measures in Industry 4.0 settings begins with the construction of a decision matrix X = [x_ij]_n×m, where x_ij denotes the performance of the alternative A_i (i = 1, 2, …, n) under criterion G_j (j = 1, 2, …, m). This matrix provides a unified quantitative representation that enables systematic comparison of all labor protection alternatives across all evaluation criteria.

Each criterion is first classified according to its optimization direction using a directional indicator δ_j as per Equation 1, which guides the choice of the appropriate normalization formula.

In this study, all criteria are modeled as benefit-oriented, reflecting a safety-first perspective in which higher scores indicate better labor protection performance.

The labor protection criteria may differ in scale and magnitude; the original decision matrix is transformed into a dimensionless, comparable form through normalization, yielding the matrix R = [r_ij]. For benefit-type criteria, the normalized value is obtained using Equation 2, whereas for cost-type criteria, it is obtained using Equation 3. It maps all entries to the interval [0, 1]. It also preserves preference orientation. Normalization is retained even when the original scores are defined on a uniform scale, ensuring consistency when benchmarking alternatives against ideal and anti-ideal reference solutions in the MARCOS procedure.

Ideal and anti-ideal reference profiles are then appended to the normalized matrix as defined by Equation 4. It presents the best attainable and worst observed performance across all criteria. These reference vectors serve as benchmarks for evaluating each alternative's relative position in the subsequent ranking phase.

To eliminate subjectivity in assigning criterion weights, four objective weighting methods grounded in different theoretical principles are employed: Entropy, CRITIC, Method based on the Removal Effects of Criteria (MEREC), and Criteria Impact Loss on System (CILOS) (38). All weighting procedures are applied to the original decision matrix X, thereby preserving the data's intrinsic variability.

To obtain a robust composite weight vector, the four objective weight sets are fused using a Bonferroni-type aggregation operator (40).

Let wj(k) denote the weight of the criterion G_j obtained from the k-th objective method, with k = 1, 2, …, K. The fused weight w~j is expressed as in Equation 15.

where L denotes a paired method index within the Bonferroni formulation, followed by normalization of w~j over all criteria. This operator captures interaction effects between weighting schemes and reduces the sensitivity of the final weights to outlier values produced by any single method.

The MARCOS method is employed as the primary MCDM ranking procedure, using the normalized matrix R and the fused weight vector w~j (40, 41). First, the weighted normalized values are computed as in Equation 16, and the overall performance score for each alternative is calculated using Equation 17. Relative indicators are then defined as in Equation 18, and the final utility value of the alternative A_i is given by Equation 19.

To validate the robustness of the MARCOS-based ranking, comparative analyses are conducted using other well-established MCDM methods, namely VIKOR, Preference Ranking Organization Method for Enrichment Evaluation II (PROMETHEE-II), TOPSIS, Evaluation based on Distance from Average Solution (EDAS), and Weighted Aggregated Sum Product Assessment (WASPAS), all applied with the same fused weights w~j.

VIKOR is used for compromise-based ranking. VIKOR focuses on identifying an alternative that provides the closest balance between group utility and individual regret. Its results differ because it emphasizes reaching a “compromise solution” rather than simply minimizing distance (42).

PROMETHEE-II is based on outranking logic; it evaluates pairwise preferences using preference functions and provides a complete ranking via net flow values. Its impact varies as it captures preference intensity rather than pure distance or aggregation (43).

TOPSIS, a distance-based technique, ranks alternatives relative to the ideal and negative-ideal solutions. Its results differ since it prioritizes closeness to the best scenario and remoteness from the worst (44).

EDAS, this deviation-based method measures how far each alternative's performance deviates from the average solution. The variation arises because it centers decisions on positive and negative distances from the mean rather than on absolute ideal points (40, 45).

WASPAS combines weighted-sum and weighted-product models to aggregate criterion scores. Its differing impact comes from integrating both additive and multiplicative principles, improving ranking stability and sensitivity (42, 46).

To assess the consistency of alternative rankings across different MCDM methods, non-parametric rank correlation measures are employed (47).

Kendall's tau is given by Equation 20.

where N_c and N_d are the numbers of concordant and discordant pairs of alternatives, respectively.

Spearman's rank correlation coefficient between rankings a and b is computed as with Equation 21.

where ri(a) and ri(b) denote the positions of the alternative A_i in rankings a and b.

Sensitivity analysis is conducted to examine the stability of rankings under perturbations of the criteria weights (47). For a selected criterion G_h, the adjusted weights are defined as

where ϵ is a small perturbation parameter. The utility scores and rankings are recomputed under these modified weights to observe how changes in the importance of individual criteria affect the prioritization of alternatives.

To obtain a final consensus ranking that integrates information from all MCDM methods and sensitivity scenarios, Borda count and Copeland aggregation rules are applied. These procedures combine multiple rank orders into a single collective ranking that reflects the overall preference structure (47).

The evaluation criteria were defined using a structured, multi-stage process. First, a targeted review of occupational safety and health literature and international regulatory frameworks (e.g., ILO-OSH, ISO 45001, EU-OSHA guidelines) was conducted to identify commonly used dimensions for assessing workplace safety interventions (9, 52–54, 56). This review highlighted recurring evaluation themes related to economic feasibility, operational effectiveness, regulatory compliance, human-centric considerations, and sustainability, which are particularly relevant in Industry 4.0 environments characterized by digitalization and human–technology interaction (55). Second, the preliminary set of criteria was refined through expert consultation to ensure contextual relevance and practical applicability (53). The expert panel comprised professionals with experience in occupational safety management, industrial engineering, and digital safety systems. Experts assessed the clarity, completeness, and non-redundancy of the proposed criteria and confirmed their suitability for evaluating labor protection measures in digitally enabled workplaces. Based on this process, six evaluation criteria were finalized for subsequent analysis. This combined literature-informed and expert-validated approach ensures that the selected criteria are theoretically grounded, practically meaningful, and aligned with contemporary occupational safety decision-making needs.

All selected criteria are modeled as benefit-type, including Life Cycle Cost (C1), ensuring that higher values consistently represent better performance across effectiveness, regulatory compliance, usability, sustainability, and ease of implementation. This design aligns the evaluation framework with safety-first principles, in which improvements in these dimensions directly enhance labor protection outcomes rather than entail trade-offs against safety. In international OSH guidance (e.g., ILO-OSH, ISO 45001, OSHA), core safety measures are framed as mandatory obligations rather than discretionary investments, which means they should not be systematically deprioritized solely on the basis of cost. In this study, cost minimization is therefore not treated as the dominant decision objective in order to avoid biasing the ranking toward purely economic considerations at the expense of safety performance. At the same time, long-term economic feasibility remains an essential component of occupational safety planning, particularly in Industry 4.0 environments where technologies differ in acquisition, operation, and maintenance demands. To reflect this, Life Cycle Cost (C1) is incorporated as a benefit-type criterion that captures affordability and long-term budgetary impact, with higher scores indicating more economically sustainable options. In this way, economic considerations are integrated into a balanced, multi-objective decision-support framework without undermining the primacy of safety-related benefits. In this way, economic considerations are integrated into a balanced, multi-objective decision-support framework while preserving the primacy of safety-related benefits in the final prioritization.

The performance scores used to construct the decision matrix (Table 1) were obtained through a structured expert-based evaluation supported by secondary sources. First, a preliminary assessment of labor protection measures was informed by relevant occupational safety standards, regulatory guidelines, and prior empirical studies to ensure baseline consistency across criteria. Subsequently, expert judgment was employed to evaluate the relative performance of each alternative with respect to each criterion, reflecting practical implementation conditions in Industry 4.0 environments. An expert panel comprising 7 professionals with backgrounds in occupational safety management, industrial engineering, and digital safety systems participated in the evaluation process. All experts had at least 8 years of professional experience in workplace safety–related roles. Experts independently scored each alternative against each criterion using a nine-point Likert scale (1 = very low performance, 9 = very high performance). To reduce individual bias, the final performance score for each alternative–criterion pair was obtained by averaging individual expert scores. The resulting aggregated scores were subsequently normalized within the MCDM framework prior to weighting and ranking. This approach is consistent with established expert-based MCDM practice and enables transparent, reproducible construction of the decision matrix while accommodating qualitative and quantitative safety considerations. Further details on the expert-based performance scoring process, including the scoring scale and an example evaluation template, are provided in Appendix A (Supplementary material).

The number of evaluation criteria was intentionally limited to six to balance analytical completeness with model parsimony. In MCDM, an excessive number of criteria may introduce redundancy, increase subjectivity in expert scoring, and reduce the interpretability of results. The selected criteria collectively capture the principal dimensions that govern occupational safety decision-making in Industry 4.0 environments, namely economic feasibility, operational effectiveness, regulatory compliance, human-centric adoption, sustainability, and implementation feasibility. Additional technology-specific attributes, such as technological maturity, scalability, and interoperability, were not introduced as independent criteria because their effects are implicitly reflected within implementation feasibility and sustainability considerations. Low technological maturity or poor interoperability directly constrains implementation feasibility, while limited scalability influences long-term sustainability and organizational resilience. Treating these attributes as separate criteria would therefore risk conceptual overlap and double-counting.

Four complementary objective weighting methods, namely Entropy, CRITIC, MEREC, and CILOS, as formulated in Equations 5–14, are used to determine the relative importance of the six evaluation criteria. Each method captures a distinct aspect of criterion significance: Entropy reflects information dispersion, CRITIC incorporates variability and inter-criterion conflict, MEREC quantifies performance loss from criterion removal, and CILOS evaluates the contribution to closing the performance gap relative to ideal conditions. To mitigate bias associated with any single method and enhance robustness, the individual weight vectors were aggregated using the Bonferroni fusion operator, as defined in Equation 15.

Table 2 depicts the individual objective weights obtained from each method along with the final Bonferroni-fused weights. The results indicate that C1 (Life Cycle Cost/Affordability) receives the highest composite weight (0.3004). It highlights its dominant influence on decision outcomes even within a safety-first framework. It also replicates the substantial dispersion and discriminative power of cost-related performance scores across the alternatives under Entropy and CRITIC formulations. The second tier of importance comprises C6 (Ease of Implementation) and C5 (Sustainability and Long-Term Impact), with fused weights of 0.1583 and 0.1579, respectively. These criteria capture operational feasibility and long-term system resilience, both of which are critical in Industry 4.0 situations characterized by complex cyber–physical integration and sustainability-driven policy pressures. The near-equal weighting of these criteria suggests a balanced emphasis on practical deployability and enduring safety benefits. The criteria C2 (Risk-Reduction Effectiveness), C4 (Worker Acceptance and Usability), and C3 (Regulatory Compliance Level) exhibit moderately lower but comparable fused weights (0.1335, 0.1252, and 0.1247, respectively). This distribution indicates that, while these dimensions remain essential to occupational safety decision-making, their relative discriminative impact across the considered alternatives is less pronounced than that of cost, sustainability, and implementation feasibility. The close clustering of weights among C2, C3, and C4 suggests that the framework does not over-prioritize any single behavioral or regulatory dimension, thereby preserving multi-dimensional balance. The Bonferroni-fused weight vector demonstrates stability and methodological coherence by integrating diverse objective perspectives and preventing dominance by extreme values from any individual method. These fused weights are then used in the MARCOS ranking procedure (Equations 16–19) to ensure that alternative prioritizations reflect a robust, data-driven assessment of labor protection performance.

Figure 2 utilized to highlight the relative dominance of evaluation criteria with the Bonferroni-fused weights using a Pareto chart. It enables a cumulative assessment of criterion importance by ranking the fused weights in descending order and illustrating their collective contribution to the overall decision structure. Life Cycle Cost (C1) alone accounts for nearly one-third of the total decision weight, confirming its dominant role in the objective weighting framework. When combined with Ease of Implementation (C6) and Sustainability and Long-Term Impact (C5), the cumulative contribution exceeds 60% of the total weight, indicating that these three criteria form the core of critical decision-making in Industry 4.0 labor protection planning. The Risk-Reduction Effectiveness (C2), Worker Acceptance and Usability (C4), and Regulatory Compliance Level (C3) contribute a smaller yet non-negligible share, reinforcing the multidimensional nature of occupational safety assessment without disproportionately amplifying behavioral or regulatory factors. This confirms that the fused weighting scheme successfully differentiates between high-impact and supporting criteria while preserving balance across economic, operational, and safety-oriented dimensions.

The high weight assigned to C1 indicates the importance of long-term economic feasibility in sustaining safety interventions, rather than a preference for low-cost solutions at the expense of effectiveness or worker wellbeing. The dominance of C1 in the weighting results reflects expert concern for the long-term economic sustainability of safety interventions rather than a preference for low-cost solutions, emphasizing the importance of balancing economic feasibility with protection effectiveness.

Labor protection measures were prioritized using the MARCOS method, employing the Bonferroni-fused objective weights derived in Section 4.1. The MARCOS approach evaluates each alternative relative to both ideal and anti-ideal reference solutions. It enables a transparent assessment of performance proximity under a multi-criteria safety framework.

Table 3 depicts the extended normalized decision matrix. It is constructed by joining the ideal and anti-ideal reference profiles to the set of alternatives. This step establishes the upper and lower performance bounds for each criterion, ensuring that all labor protection measures are evaluated within a unified comparative space. The normalized values indicate that A3 PPE and A5 (Digital Safety Monitoring Systems) achieve several criterion-wise maxima, whereas A1 (Engineering Controls) exhibits comparatively lower normalized values for cost and ease of implementation. The weighted normalized matrix is obtained using Equation 16 and reflects the combined influence of performance scores and Bonferroni-fused criteria weights. A3 achieves the highest weighted contributions for C1 (Life Cycle Cost), C4 (Worker Acceptance and Usability), and C6 (Ease of Implementation), indicating a strong balance between affordability, usability, and deployability. A5 shows dominant weighted values in C3 (Regulatory Compliance Level) and C5 (Sustainability and Long-Term Impact), highlighting its technological and long-term strategic advantages, albeit with moderate implementation feasibility. The relative closeness of each alternative to the ideal and anti-ideal solutions is quantified through the relative indicators λi+ and λi-, as defined in Equation 18. These indicators are subsequently integrated to compute the final utility value U_i using Equation 19.

The final utility values indicate that A3 PPE is the most preferred labor protection measure, with the highest utility score of U_i = 0.7124. This result reflects PPE's strong performance across cost efficiency, worker acceptance, and implementation simplicity, which are particularly critical in Industry 4.0 environments with diverse operational constraints. A2 (Administrative Controls) ranks second (U_i = 0.6966), demonstrating stable and consistent performance across most criteria, especially in terms of affordability and regulatory alignment. A4 (Safety Training Programs) occupies the third position (U_i = 0.6810), supported by balanced scores across all dimensions but lacking dominance in any single high-weight criterion. A5 (Digital Safety Monitoring Systems) is ranked fourth (U_i = 0.5955), despite excelling in sustainability and compliance, due to comparatively lower performance in cost and ease of implementation. A1 (Engineering Controls) ranks fifth (U_i = 0.5940), indicating that although structurally effective, such controls are less competitive when evaluated against modern Industry 4.0 constraints related to flexibility, affordability, and rapid deployment. The MARCOS-based prioritization demonstrates a transparent and interpretable ranking structure driven by proximity to ideal safety performance. The results confirm that labor protection strategies emphasizing usability, implementation feasibility, and balanced cost–performance trade-offs are most suitable for contemporary Industry 4.0 industrial environments.

The robustness and consistency of the MARCOS-based prioritization are assessed by a comparative ranking analysis using five additional, well-established MCDM methods: TOPSIS, VIKOR, EDAS, WASPAS, and PROMETHEE-II. All MCDM methods were implemented using the same Bonferroni-fused objective weights to ensure methodological comparability, and the resulting rankings are presented in Figure 3. The comparative rankings reveal strong agreement among most methods regarding the overall preference structure. A3 PPE is consistently identified as the top-ranked alternative by MARCOS, TOPSIS, EDAS, WASPAS, and PROMETHEE-II. This indicates strong and balanced performance across highly weighted criteria such as affordability, usability, and ease of implementation. The only deviation occurs in the VIKOR results, where A3 ranks third, reflecting VIKOR's emphasis on compromise solutions and on minimizing individual regret rather than pure proximity to the ideal solution. A2 (Administrative Controls) exhibits remarkable stability, achieving second place across all six MCDM methods. This consistency indicates that administrative controls provide a robust and reliable safety strategy with balanced performance under diverse decision philosophies. A4 (Safety Training Programs) shows moderate variability, ranking first under VIKOR but third under the remaining methods, suggesting sensitivity to decision models that prioritize compromise and group utility.

Kendall's tau and Spearman's rank correlation coefficients were computed using Equations 20 and (19), respectively. The resulting correlation matrices are presented in Table 4. Kendall's tau results indicate strong concordance between MARCOS and PROMETHEE-II (τ = 1.0) and high agreement with TOPSIS, EDAS, and WASPAS (τ = 0.8). In contrast, lower Kendall correlation values are observed between VIKOR and the other methods, with an average τ of 0.433, confirming its distinct compromise-oriented ranking behavior.

The Spearman's ρ matrix further reinforces these findings, with high correlation values between MARCOS and EDAS (ρ = 0.9), TOPSIS (ρ = 0.9), WASPAS (ρ = 0.9), and PROMETHEE-II (ρ = 1.0). The mean Spearman's ρ values indicate strong overall consistency for MARCOS and PROMETHEE-II (0.867), whereas VIKOR again exhibits comparatively lower alignment (mean ρ = 0.633). These results demonstrate that distance-based, outranking, and aggregation-based methods converge toward a standard ranking structure, while compromise-based methods introduce controlled divergence. The high degree of inter-method agreement validates MARCOS as a suitable primary decision-making tool for evaluating labor protection measures in Industry 4.0 environments.

A sensitivity analysis was conducted to evaluate the robustness of the MARCOS-based prioritization to variations in criterion importance. A one-at-a-time perturbation strategy was adopted, whereby the weight of each criterion was independently increased and decreased while preserving normalization, as formulated in Equation 22. The sensitivity analysis design and feasibility characteristics are summarized in Table 5. Six criteria were examined across three perturbation magnitudes of ± 0.05, ± 0.10, and ± 0.15, yielding a total of 36 scenarios. Among these, 33 scenarios were found to be feasible. In comparison, three scenarios corresponding to extreme negative perturbations (δ = −0.15) for C2, C3, and C4 resulted in negative weights and were therefore treated as boundary-violating cases.

Table 6 presents the frequency distribution of MARCOS ranks across all feasible sensitivity scenarios. The results show that A3 PPE dominates, achieving Rank-1 in 22 scenarios and remaining in the top three in the vast majority of cases. This pattern indicates that a narrow weighting configuration does not drive A33′′s superior performance; rather, it is resilient across a wide range of assumptions about criterion importance. A4 (Safety Training Programs) shows a mixed pattern, frequently ranking 3rd and occasionally achieving Rank-1, suggesting moderately sensitive competitive performance to weight fluctuations. A2 (Administrative Controls) exhibits exceptional consistency, ranking 2nd in 34 scenarios, highlighting its stability as a secondary option across perturbation intensities. In contrast, A1 (Engineering Controls) and A5 (Digital Safety Monitoring Systems) predominantly occupy lower ranks. A1 consistently appears in Rank-4 and Rank-5, while A5 alternates between Rank-4 and Rank-5, indicating limited robustness under changing decision priorities. These findings suggest that alternatives requiring higher capital investment or greater implementation complexity are more sensitive to variations in criteria weighting within Industry 4.0 contexts.

To synthesize sensitivity analysis observations into a concise robustness metric, a Stability Index (SI) was computed for each alternative (Table 7). The SI values quantify the frequency with which an alternative appears in top-ranked positions across all sensitivity scenarios. A3 attains the highest SI for both Top-1 (0.61) and Top-2 (0.69) positions, confirming it as the most robust and preferred alternative. A2, while rarely achieving Rank-1, records an SI (Top-2) of 1.0, reinforcing its role as a highly stable and reliable secondary choice. A4 demonstrates moderate robustness but is sensitive to extreme perturbations, whereas A1 and A5 exhibit negligible SI values, confirming their weak resilience. The sensitivity analysis validates the structural stability and decision reliability of the proposed MARCOS-based framework.

Individual MCDM methods provide valuable perspectives on alternative prioritization, and final decision-making in complex safety contexts benefits from a unified consensus that integrates multiple ranking viewpoints. Borda count and Copeland aggregation rules were employed to synthesize the rankings obtained from all MCDM methods and sensitivity scenarios, as formulated in Equations 23–26. In parallel, the Stability Index (SI), defined in Equation 27, was incorporated to account for ranking robustness under weight perturbations explicitly (Table 8).

The Borda aggregation results reveal a clear hierarchy among the labor protection alternatives. A3 PPE achieves the highest Borda score (20) and is consequently assigned the first Borda rank, indicating that it consistently occupies top positions across the evaluated MCDM methods. This outcome confirms 3′s strong overall performance across diverse decision logics, including distance-based, outranking, and deviation-based approaches. A2 (Administrative Controls) follows closely with a Borda score of 18, reflecting stable mid-to-high ranking behavior across all methods. The Copeland results reinforce this preference structure through pairwise dominance analysis. A3 records the highest Copeland score (+4), demonstrating that it outperforms all other alternatives in the majority of pairwise comparisons. A2 again ranks second with a Copeland score of +2, confirming its consistent competitive advantage over the remaining alternatives. A4 (Safety Training Programs) occupies the third position with a neutral Copeland score, indicating balanced wins and losses in pairwise contests. In contrast, A1 (Engineering Controls) and A5 (Digital Safety Monitoring Systems) exhibit negative Copeland scores, signifying frequent pairwise losses and comparatively weaker overall preference.

The Stability Index adds a robustness-oriented dimension to the consensus analysis. A3 achieves the highest SI (Top-1) at 0.833, confirming that it not only ranks highly on average but also maintains its leading position across varying weighting conditions. A4 shows limited robustness, with a modest SI value, whereas A2, despite its strong aggregate ranking, does not achieve Rank-1 in sensitivity scenarios, highlighting its role as a stable but secondary option. A1 and A5 record zero SI values, indicating poor robustness and reinforcing their lower priority status.

By jointly considering Borda scores, Copeland dominance, and stability behavior, the final consensus ranking unambiguously identifies A3 as the most preferred and robust labor protection measure for Industry 4.0 environments, followed by A2 and A4. This multi-layered aggregation approach ensures that the final decision reflects not only average performance across methods but also resilience under uncertainty. The proposed framework delivers a transparent, reliable, and defensible prioritization of labor protection strategies suitable for complex, technology-driven industrial settings.

It should be noted that the fact that PPE is the most preferred labor protection measure does not mean that traditional safety measures are more effective than digital or advanced technologies. This result reflects the status quo in Industry 4.0 safety applications. PPE is the least cutting-edge, most regulated, and therefore the quickest to adopt. Well-defined standards, a high level of employee familiarity, and relatively predictable life-cycle costs add value to the PPE, contributing to its competitiveness across economic, regulatory, and usability aspects.

On the other hand, digital systems for safety, such as real-time monitoring solutions, smart wearables, and data analytics-based hazard-detection platforms, typically struggle with tech maturity and integration complexity, as well as questions about data governance and capex levels. Although these solutions are promising in the long term for proactively and predictively safeguarding safety, their effectiveness strongly relies on contextual factors such as organizational preparedness, infrastructure setup, and workforce digital literacy.

From a safety management perspective, the findings also suggest that digital transformation should not be considered a substitute for other safety-enhancing measures, but rather a complement. Digital technologies should be considered supplements to traditional protection measures, not substitutes. The prevalence of PPE atop the list again underscores the need for basic protective tools to exist and be used, even as data-driven safety systems become further integrated into organizations.

For adopters, this means that investment in digital safety technologies must be coupled with favorable policies, training, and gradual adoption strategies to help overcome barriers to adoption. Without enabling conditions, however, an overemphasis on advanced technologies can potentially introduce new risks or compound existing ones. Hence, successful safety management in Industry 4.0 requires an appropriate mix of traditional and digital measures tailored to the organization's capabilities and maturity.

From a managerial perspective, the results offer a clear, evidence-based prioritization of labor protection measures under complex, multi-criteria conditions. The consistent top ranking of PPE across MARCOS, comparative MCDM methods, sensitivity scenarios, and consensus aggregation indicates that PPE remains highly robust and immediately deployable in Industry 4.0 settings. Managers operating in smart factories, automated production lines, and cyber–physical systems can therefore justify prioritizing advanced PPE solutions, including smart and sensor-enabled equipment, particularly when rapid implementation and workforce acceptance are critical.

The strong, stable performance of Administrative Controls underscores their role as a reliable secondary layer of protection, especially in environments where engineering redesign or significant capital investments are constrained. Organizations can leverage this insight to strengthen safety policies, work procedures, and scheduling strategies in parallel with technological interventions. In contrast, the relatively lower and less robust ranking of Engineering Controls and Digital Safety Monitoring Systems suggests that, although technically effective, these measures may face barriers to cost, complexity, or integration. Managers should therefore approach these solutions through phased implementation or pilot-scale deployment rather than immediate large-scale adoption.

The use of broad intervention categories represents a deliberate abstraction that may mask performance variability among specific Industry 4.0 technologies. For example, the category of digital safety monitoring systems may encompass smart wearable sensors, vision-based hazard-detection systems, and real-time location-tracking platforms, each with distinct cost structures, maturity levels, and implementation challenges. Similarly, engineering controls may include both traditional physical safeguards and advanced collaborative robotic systems with integrated safety functions. While this abstraction supports strategic-level prioritization, it may oversimplify technology-specific trade-offs.

Accordingly, the results of this study should be interpreted as guidance for identifying priority intervention domains rather than as prescriptions for specific technological solutions. Future research may extend the proposed framework by disaggregating these categories and evaluating individual Industry 4.0 technologies, such as virtual reality–based safety training, smart personal protective equipment, or collaborative robots, as distinct alternatives once sufficient empirical performance data are available.

For policymakers and regulatory bodies, the study provides quantitative evidence supporting a multi-layered safety strategy aligned with international OSH frameworks such as ISO 45001 and ILO-OSH. The dominance of criteria related to affordability, ease of implementation, and sustainability implies that regulatory guidelines should not only mandate compliance but also encourage practical feasibility and long-term resilience in safety interventions. The results further suggest that incentive schemes, subsidies, or technical assistance programs could be particularly effective in facilitating the adoption of advanced safety technologies that otherwise score lower due to implementation barriers. The demonstrated robustness of the consensus ranking reinforces the value of data-driven, multi-criteria evaluation tools for regulatory impact assessments. Authorities responsible for workplace safety audits and compliance planning can adopt similar MCDM-based frameworks to transparently justify policy decisions and prioritize interventions across sectors with varying technological maturity.

The study provides a comprehensive, replicable decision-support framework that advances existing MCDM applications in occupational safety. The fusion of multiple objective weighting techniques using the Bonferroni operator reduces subjectivity. It enhances stability, while integrating MARCOS with rank correlation, sensitivity analysis, and consensus methods strengthens the credibility of the results. This methodological architecture can be readily extended to other Industry 4.0 decision problems, such as selecting safety technology, evaluating risk mitigation strategies, or sustainability-driven industrial planning. The explicit incorporation of a Stability Index alongside traditional aggregation methods demonstrates the importance of robustness-aware decision-making under uncertainty. Future studies can build upon this approach by integrating fuzzy data, probabilistic weights, or dynamic criteria to capture evolving industrial risk landscapes.

The study provides a practically grounded, policy-relevant, and methodologically rigorous foundation for improving labor protection decision-making in Industry 4.0 environments, bridging the gap between advanced analytical techniques and real-world safety management needs.

The study has several limitations that should be acknowledged. First, the analysis is based on a fixed and relatively limited set of alternatives and evaluation criteria. Although these were selected through literature review and expert validation to represent key economic, operational, regulatory, human-centric, and sustainability dimensions, they may not fully capture the diversity and complexity of occupational safety practices across all industrial sectors and geographical contexts. In particular, sector-specific requirements and risk profiles, such as those found in construction, healthcare, or heavy manufacturing, may influence the relative effectiveness of labor protection measures.

Second, the performance scores used to construct the decision matrix are based on expert judgment, supported by secondary standards and guidelines. While expert-based evaluation is appropriate in data-scarce safety contexts, and objective weighting methods were employed to mitigate subjectivity, contextual bias cannot be entirely eliminated. Differences in professional background, experience, or organizational context may influence expert perceptions of alternative performance.

Third, the proposed framework assumes static criteria weights within each evaluation scenario. Although sensitivity analysis was conducted to assess the robustness of rankings under local weight perturbations, the model does not fully capture dynamic or time-varying changes in safety priorities that may emerge as Industry 4.0 technologies mature, regulatory frameworks evolve, or organizational safety cultures develop. The analysis is therefore limited to a single-period decision context and does not incorporate long-term learning effects, adaptation processes, or feedback loops.

Another limitation arises from the use of aggregated intervention categories, such as engineering controls or digital safety monitoring systems. While this abstraction supports strategic-level decision-making and generalizability, it may obscure performance heterogeneity among specific Industry 4.0 technologies, including smart wearables, vision-based monitoring systems, or collaborative robotic solutions.

Future research may address these limitations by extending the framework to incorporate dynamic weighting schemes, real-time performance data from digital monitoring systems, and longitudinal evaluations that reflect evolving safety priorities. Sector-specific studies and technology-level disaggregation of alternatives would further enhance contextual relevance. Additionally, integrating empirical accident data, sensor-based indicators, or adaptive learning mechanisms could improve objectivity and strengthen the practical applicability of the proposed decision-support framework.

The proposed framework may be enhanced by incorporating fuzzy, interval-valued, or probabilistic data representations to better handle the uncertainty and vagueness inherent in occupational safety assessments. Dynamic or time-dependent weighting schemes could be explored to reflect evolving safety priorities as industrial systems transition toward higher levels of automation and digitalisation. The model can be expanded by integrating real-world accident statistics, sensor-generated safety data, or digital twin simulations to strengthen empirical grounding and reduce reliance on expert-based scoring. Sector-specific adaptations in construction, mining, chemical processing, or logistics would further improve contextual relevance and generalizability. Hybridizing MARCOS with artificial intelligence or machine learning techniques, such as reinforcement learning for adaptive weighting or clustering for alternative grouping, represents a promising avenue for intelligent safety management systems. Finally, future studies may incorporate cost–benefit or life-cycle sustainability assessments alongside safety performance to support holistic industrial decision-making.