Supply Chain Digitalization and Supply Chain Resilience in Manufacturing SMES in Italy

2.5. Literature Synthesis and Gap

The literature suggests that there are information-related mechanisms through which digital capabilities can support supply chain resilience, yet there are several limitations. Previous research has studied the general problem of digitalization in the supply chain e.g., (Al Tera et al., 2024; Zhao et al., 2023), looked at data analytics specifically e.g., (Jiang et al., 2025), or examined Industry 4.0 technologies specifically e.g., (Huang et al., 2023); however, it is less clear whether IoT adoption, data analytics capability, and cloud integration function as separate antecedents of visibility. Additionally, the results regarding top management support are also mixed, as (Ali Shah et al., 2026) find that top management support strengthens the relationship, while (Xu & Zhao, 2022) report insignificant moderation. Context also remains important. While several Italian studies have investigated e-commerce and resilience of small manufacturers (Onjewu et al., 2022), proactive resilience in Italian manufacturing SMEs (Carissimi et al., 2023), and digital resources among Italian firms (Pellegrino & Gaudenzi, 2026), none focus specifically on the combination of the three digital capabilities, supply chain visibility as a common mediating mechanism, and top management support as a moderator. The present study thus investigates the relationships together in the case of Italian manufacturing SMEs.

3. METHODOLOGY

3.1. Research Design and Sampling

This study was based on a quantitative and cross-sectional approach to investigate the relationship of IoT adoption, data analytics capability, and cloud integration with supply chain visibility, and supply chain visibility association with supply chain resilience of Italian manufacturing SMEs, with a specific focus on the role of top management support as a moderator. The study also investigated the mediating role of supply chain visibility. The quantitative approach allowed the proposed relationships to be examined statistically with the use of the structured questionnaire data. Data were collected at one point in time with the individual employee as the unit of analysis. The respondents were chosen from roles that have first-hand knowledge of the organization’s supply chain and operational activities.

The study targeted manufacturing employees across the supply chain, procurement, logistics, operations, production, warehousing, planning, and managerial roles in SMEs in Italy. Respondents had to meet certain organizational and occupational requirements; thus, purposive sampling was used. Participants were screened before they started the main questionnaire to ensure that they were currently working for a manufacturing organization in Italy with less than 250 employees (Appendix A). Those not meeting either criterion were excluded. The demographic and professional description of the 344 respondents is shown in Table 1. There was a relatively equal distribution of gender, and the majority held a bachelor’s or master’s degree (63.1%). Production was the largest job role (17.7%), and food and beverages were the largest industry group (16.6%). The majority of the respondents worked in an organization that had 50-249 employees (67.4%) and had 1-10 years of experience (51.7%).

Recruitment was carried out on LinkedIn professional networks, Italian manufacturing and supply chain communities, procurement and logistics communities, and professional networks of operations and manufacturing practitioners. Invitations were also sent through contacts in manufacturing organizations. The invitation included a short description of the research, the eligibility criteria, voluntary participation information, and the link to the questionnaire. The survey was kept open throughout the collection period and yielded 344 valid responses.

To further check the suitability of the sample, G*Power 3.1 was used. Using the linear multiple regression: fixed model, R² increase procedure, six predictors, a statistical power of 0.80, an α level of 0.05, and a maximum number of predictors in the structural model was used for the calculation. A minimum sample requirement of 85 respondents was achieved. The 344 usable observations therefore exceeded the minimum requirement. The invitation was openly disseminated via multiple electronic means, and the number of recipients or those who saw it could not be determined. Therefore, a traditional response rate calculation was not possible. Due to the purposive approach, the results also cannot be used as statistically representative of the entirety of Italian manufacturing SMEs.

3.2. Instrument Development and Pilot Testing

The questionnaire was created based on existing construct conceptualizations and existing empirical measures. The items were not presented as direct adaptations from one existing instrument but were researcher-worded and inspired by previous literature. (Vass et al., 2018; and Shahzad et al., 2023) guided the adoption of IoT, while (Gupta & George, 2016; and Wong & Ngai, 2023) guided the data analytics capability, and (Shee et al., 2018; and Yao et al., 2007) guided the cloud integration. The sources for supply chain visibility were (Williams et al., 2013; and Barratt & Oke, 2007), and for supply chain resilience, (Ambulkar et al., 2015; and Chowdhury & Quaddus, 2017; Yao et al., 2007) inspired the top management support, while some related work was based on (Chen et al., 2015; and Lai et al., 2018).

There were 32 measurement items: 5 IoT adoption, 5 data analytics capability, 5 cloud integration, 6 supply chain visibility, 6 supply chain resilience, and 5 top management support (Appendix A). All the responses were obtained on a 5-point Likert scale ranging from 1 = strongly disagree to 5 = strongly agree. This format allowed for adequate differentiation of responses and has been applied to similar research on supply chains and analytics (Dennehy et al., 2021; Zhang et al., 2023). A pilot test was carried out prior to the main survey with 35 participants whose characteristics matched those of the target population. Feedback emphasized clarity, wording, comprehensibility, and interpretation of the items. Before administering the final questionnaire, necessary wording refinements were made, and pilot responses were not included in the final data set.

3.3. Data Analysis and Data Quality Assessment

SmartPLS software was used to analyze the data by means of Partial Least Squares Structural Equation Modelling (PLS-SEM). It is a variance-based structural equation modelling technique that allows for the estimation of relationships between latent constructs, the assessment of measurement and structural models and mediation and moderation effects for predictive research (Hair et al., 2011). Thus, this approach is suitable as the proposed model has several latent constructs and multiple relationships in the model, including direct, mediation, and moderation relationships. Indicator loadings, internal consistency reliability, convergent validity, and discriminant validity were used to evaluate the measurement model first. Path coefficients, significance level, effect size, and explained variance were then used to assess the structural model. Within the structural model, mediation effects of supply chain visibility and the moderating effect of top management support were explored. Significance and confidence intervals of the hypothesized relationships were obtained using bootstrapping with 5,000 resamples. Also, the direct paths from the three digital capabilities to resilience were also estimated to assess whether resilience was associated with these capabilities beyond the indirect visibility pathway.

The non-response bias in data quality assessment is examined by comparing 50 early responses and 50 late responses. No significant differences were observed between the two groups on the study variables, thus indicating that systematic non-response bias was unlikely to have an impact on the results. Furthermore, procedural and statistical methods were used to assess potential Common Method Bias (CMB). The respondents were told that they would be anonymous, that their answers would remain confidential, and that the information that they provided would be gathered for academic purposes only. Statistically, the first factor was found to contain 34% of the total variance, which is below the 50% criterion generally used in the application of Harman’s first factor test. The full-collinearity VIF values were also checked, and all of them were less than 3.3, which means that there is no severe CMB problem (Kock, 2017). The highest VIF was 2.192 for the SCV → SCR relationship (Table 2).

4. RESULTS

4.1. Measurement Model

Table 3 shows the measurement model results for the studied constructs. All factor loadings exceed 0.70, ranging from 0.772 to 0.887. Cronbach’s alpha and composite reliability values are >0.70, and AVE values are between 0.669 and 0.752. Hence, the constructs have sufficient indicator reliability, internal consistency, and convergent validity.

4.2. Discriminant Validity

The HTMT ratios used for checking the discriminant validity are reported in Table 4. The values are all within the range of 0.118 to 0.837, which is below the recommended limit of 0.85. Although SCV-SCR was close to the threshold of 0.85 (0.837), it still falls under the suggested threshold. It suggests that the constructs are empirically different from each other, which means that the measurement model has satisfactory discriminant validity.

4.3. Model Fit

Table 5 presents the model fit results for the saturated and estimated models. SRMR values are 0.041 and 0.054, which are below the acceptable value of 0.08. NFI values of 0.905 and 0.902 were considered as acceptable. The d_ULS and d_G are reported for completeness and are not indicative of notable fit concerns.

4.4. Path Coefficient

Table 6 shows the direct effects between the variables of the study. CI, DAC, and IOTA each have significant positive associations with SCV (β = 0.301, 0.299, and 0.349; p < 0.001), supporting H1a–H1c. SCV is also positively related to SCR (β = 0.389, p < 0.001), confirming H2. The direct effects of CI, DAC, and IOTA on SCR are significant (β = 0.272, 0.093, and 0.221), although DAC has a small effect size (f² = 0.021). The relationship between TMS and SCR is positive (β = 0.146, p < 0.001), while the interaction between TMS and SCV is not significant (β = 0.031, p = 0.235), which does not support H4.

Indirect effects (through SCV) are shown in Table 7. All three indirect effects are positive and significant (p < .001), consistent with H3a, H3b, and H3c. The VAF values are 38.1% for IOTA, 30.1% for CI, and 55.7% for DAC. Since the corresponding direct effects remain significant, SCV provides partial mediation in all three relationships.

4.5. Simple Slope Analysis

Table 8 shows the conditional effects of SCV on SCR as well as the indirect effects at varying levels of TMS. The SCV–SCR association increases from 0.358 at low TMS to 0.420 at high TMS. There are also indirect effects that rise in proportion with TMS. But since the interaction effect in Table 6 is not significant, these conditional differences do not show a significant moderation effect.

4.6. Explanatory power and Predictive Relevance

The explanatory and predictive relevance of the endogenous constructs are shown in Table 9. The model explains 69.8% of the variance in SCR (adjusted R² = 0.693) and 49.5% in SCV (adjusted R² = 0.490). The Q²predict values are >0 and positive (0.610 and 0.483), showing that these models are predictive. The RMSE and MAE also indicate the model’s prediction errors on SCR and SCV.

5. DISCUSSION

The results offer a nuanced perspective on the connection between the digital capabilities of SMEs in the Italian manufacturing sector. The findings do not necessarily mean that digitalization is a single capability, but rather that there is heterogeneity in its relevance depending on IoT adoption, data analytics capability, and cloud integration, and that supply chain visibility is a critical linking capability. This is generally aligned with Dynamic Capability Theory (DCT) and Information Processing Theory (IPT): a digital capability can be a means to create and process information within the supply chain, and visibility is an information-processing condition that can be linked to resilience.

All three digital capabilities were positively associated with supply chain visibility, with IoT adoption having the highest association (β = 0.349, f² = 0.206), followed by cloud integration (β = 0.301, f² = 0.148) and data analytics capability (β = 0.299, f² = 0.149). The ordering may have a theoretical significance. IoT is closely related to the physical observation of products, materials, equipment, and processes and therefore has a more direct informational relationship with visibility. Cloud integration and analytics, however, are about how information is linked and processed once it is available. The results, thus, complement the studies of (Vass et al., 2018; and Shahzad et al., 2023), and extend the overall picture of digitalization evidenced in (Al Tera et al., 2024). Notably, the equivalent variables for analytics and cloud integration show that visibility may not rely on one technological path. For Italian SMEs, the various digital capabilities can thus offer partial substitutes but more or less distinct ways of enhancing the availability of information in the supply chain.

The relationship between visibility and resilience was relatively stronger (β = 0.389, f² = 0.229), which is the main argument of IPT, where the information processing capacity is more relevant under uncertainty conditions. This finding is in line with the result reported by (Williams et al., 2013), yet adds to their qualification that visibility cannot act alone in the organizational situation. The findings are consistent with (Zhao et al., 2023; and Jiang et al., 2025), who both suggested that visibility is a key mechanism between digital capabilities and resilience. The current results, however, provide a more holistic interpretation, as visibility is explored as a shared phenomenon across three different digital capabilities instead of a wider digitalization index.

These relationships are also set against the backdrop of the additional direct-effect analysis. The direct pathways from each of the three digital capabilities to supply chain resilience were not hypothesized, but examined to determine if other digital capabilities beyond the visibility pathway were related to resilience. The findings show that these direct relationships were not consistent, and that the relationship between digital skills and resilience does not depend only on technology use. Together, this evidence suggests that the link between individual capabilities and resilience should be analyzed at the pathway level rather than assuming that all digital capabilities have similar direct impacts on resilience.

The mediation results, considered alongside the additional non-hypothesized direct-effect tests, strengthen this interpretation. All three digital capability–resilience relationships were partially mediated by SCV, with some mediation being stronger than others. Data Analytics Capability had the highest VAF (55.7%) while IoT adoption (38.1%) and cloud integration (30.1%) had lower scores. This distinction is relevant, as the direct link between analytics and resilience was rather weak (β = 0.093, f² = 0.021), yet was still statistically significant and had a meaningful indirect link via visibility (β = 0.117). Therefore, it seems that analytics capability is more reliant on the information-visibility pathway than IoT or cloud integration for its association with resilience. This finding is in line with (Jiang et al., 2025), who also finds that visibility is a link between analytics capability and resilience. In contrast, the direct link between cloud integration and resilience (β = 0.272, f² = 0.175) indicates that cloud-enabled connectivity might be of relevance to something other than visibility. These differences taken together suggest that digitalization should not be understood as a uniform organizational resource.

The additional direct-effect results further distinguish the three capabilities; a direct association was found between IoT adoption and resilience, while the direct association between data analytics and resilience was relatively low and between cloud integration and resilience was the highest of the three. Given that these paths were not postulated as hypotheses, they are more likely to be interpreted as supplementary evidence than as evidence of additional hypothesized relationships. However, they do explain the varying levels of power in mediation across the digital capabilities.

There is a second equally important qualification; the role of top management support was not significant. TMS was also directly positively associated with resilience (β = 0.146, f² = 0.056), but the interaction between TMS and visibility was not significant (β = 0.031, p = 0.235, f² = 0.003). The conditional estimates increased from low to high TMS, but the nonsignificant interaction indicates that this difference does not constitute statistically significant moderation. This is in contrast to (Ali Shah et al., 2026), which showed that TMS strengthens the relationship between digital technology and resilience, and is consistent with the context-dependent moderation reported by (Xu & Zhao, 2022). The difference could reflect that managerial support is associated with resilience more broadly, but not necessarily with the strength of the link between visibility and resilience. In other words, management commitment might support the organizational context of digitalization and the informational value of visibility might be relatively stable, regardless of the level of managerial support.

The results combined show a picture of digitalization in Italian manufacturing SMEs as a capability that is diverse rather than uniform. The results indicate that IoT seems especially relevant to the visibility of supply-chain activities, cloud integration exhibits a relatively more direct connection with resilience, and analytics has a relatively closer connection with resilience via the visibility pathway. These patterns were derived from the hypothesized relationships and additional direct-effects tests, and therefore, produce a more nuanced understanding of the relationship between individual digital capabilities and resilience.

This indicates that the usefulness of various digital capabilities to organizations could be related to the function within which it is integrated within supply-chain activities. The strong explanatory power for resilience (R² = 0.698) also suggests that these relationships are significant within the context of the study; however, the relationships cannot be interpreted as evidence that digital capabilities are a direct cause of resilience. The negligible TMS moderation suggests that the visibility–resilience link is not systematic in this sample and is not moderated by managerial support. Altogether, the results depict an Italian manufacturing SME landscape where digital capabilities, information visibility, and resilience are deeply connected but not necessarily in the same way.

THEORETICAL CONTRIBUTION

This study provides a more differentiated application of DCT and IPT to supply chain digitalization. DCT distinguishes between IoT adoption, data analytics capability, and cloud integration as organizational capabilities, rather than a single technological resource as digitalization. IPT, in turn, positions supply chain visibility as the information-processing condition connecting these capabilities with resilience. The results also suggest that this relationship is not equal for all capabilities, as seen in the higher mediation results for analytics. The lack of significance for the moderation of top management support also implies that a positive managerial support should not be assumed to automatically change the visibility–resilience relationship. The study, combined, represents a context-specific application of the two perspectives, both direct and information-mediated.

PRACTICAL IMPLICATIONS

Digital investment for Italian manufacturing SMEs should be linked to the function of the supply chain that they should support. Where continuous tracking and operational monitoring are important, IoT investments can be prioritized, and investments in analytics must be accompanied by processes that consolidate and make supply-chain information accessible across functions. The integration of the cloud should be evaluated more through the lens of connectivity between internal operations and external supply-chain partners than on cloud adoption alone. Managers might also consider assessing digital programs based on their contribution to information continuity in disruption. Since managerial support did not strengthen the visibility–resilience link, allocating additional management attention alone may be insufficient; organizational processes surrounding information use and cross-functional coordination also warrant consideration.