SSCM integrates environmental, social and economic objectives into the coordination of supply-chain processes. Its distinctiveness lies not simply in pursuing several outcomes but in recognising their interdependence across organisational boundaries. Systematic reviews show that the field has expanded substantially, yet research remains uneven: environmental themes receive disproportionate attention, social indicators are often underdeveloped, and empirical models frequently combine conceptually different practices into global indices ([20]; [24]; [25]; [29]).

The present study treats EMP, SP and EP as correlated but distinct first-order constructs. EMP captures environmental management systems, responsiveness to customer environmental concerns and product or process design intended to reduce material and energy use. SP captures employee safety and remuneration alongside community education, employment and health-related practices. EP captures transportation economies, scientific inventory control and lean production. This disaggregation is theoretically important because the resources, stakeholders and coordination demands underlying the three dimensions differ. Empirically, the five-factor model fits significantly better than models that collapse the SSCM dimensions, confirming that treating them as interchangeable would sacrifice explanatory precision.

Environmental initiatives are inherently information-intensive. Environmental management systems require firms to translate objectives into process controls, supplier requirements and monitoring routines. Eco-design and reductions in material or energy intensity likewise depend on information moving between customers, designers, production units and suppliers. Green integration research demonstrates that internal, supplier and customer integration enable firms to coordinate environmental knowledge and align environmental requirements with operating decisions ([16]; [19]; [18]).

From a dynamic capabilities perspective, EMP can strengthen SCI by creating a need and a platform for sensing external environmental requirements, seizing opportunities for resource efficiency and reconfiguring supply relationships. Conversely, environmental practices implemented without integration may remain localised and fail to influence planning or logistics. Therefore:

H1a: Environmental management practices are positively associated with supply chain integration.

Social sustainability extends supply-chain responsibility to employees, communities and other stakeholders. Practices concerning occupational safety, remuneration, education, employment and healthcare create obligations that cannot be fulfilled through a single function. Procurement choices affect working conditions, production scheduling affects safety, and community commitments may influence location, employment and distribution decisions. Supply chain social responsibility studies increasingly show that social practices are associated with sustainable performance when they are embedded in governance and coordination arrangements rather than treated as philanthropic appendages ([9]; [35]).

SCI provides a mechanism for translating social commitments into coordinated routines. Information sharing and joint planning can align supplier behaviour, production targets and stakeholder expectations; integration can also reduce the risk that social requirements are lost as activities move across organisational boundaries. Although the social construct in this study spans employee- and community-oriented practices, confirmatory analysis supports a coherent reflective factor. Accordingly:

H1b: Social practices are positively associated with supply chain integration.

Economic sustainability in the present model concerns operational routines that protect viability through transport efficiency, inventory control, lean production and waste minimisation. These practices appear operational by definition, yet their effective implementation often requires coordination beyond the focal plant. Lean production depends on reliable supplier deliveries; just-in-time inventory depends on timely information and synchronised replenishment; and transportation economies require consolidation across inbound and outbound flows. Operations and supply-chain strategies therefore influence performance partly through their fit with integration arrangements ([27]).

The dynamic capabilities logic predicts that EP strengthens SCI because firms seeking to eliminate waste and improve flow must connect internal schedules with external supply and demand information. Integration is not merely an outcome of economic practices but a means of continuously recombining them as operating conditions change. Thus:

H1c: Economic practices are positively associated with supply chain integration.

SCI denotes the extent to which firms coordinate information, plans and activities with supply-chain partners. In the present instrument, it is reflected in communicating customers’ future needs to suppliers, updating and sharing production plans, and responding rapidly to customer needs. Integration can reduce information delays, duplicated effort, excess inventory, scheduling conflict and logistics inefficiency. Meta-analytic evidence confirms a positive overall integration-performance relationship, while recent work identifies integration as a central capability through which analytics, compatibility and other resources influence operations ([4]; [28]; [32]).

Integration should improve the specific OP dimensions measured here by enabling production plans to match demand, improving inbound material availability, reducing costly expediting, and supporting more reliable outbound fulfilment. It therefore represents the immediate capability through which sustainability-related routines are coordinated. Hence:

H2: Supply chain integration is positively associated with operational performance.

EMP may improve OP directly by reducing resource consumption, defects and waste, while environmental management systems may increase process discipline. SP may contribute through safer workplaces, employee stability and stronger stakeholder relationships that protect continuity. EP is directly aligned with cost, inventory and logistics outcomes. Prior studies report positive performance consequences for green supply-chain practices, social responsibility and lean or circular practices, although effects depend on configurations and complementary capabilities ([2]; [7]; [31]; [35]). Accordingly:

H3a: Environmental management practices are positively associated with operational performance.

H3b: Social practices are positively associated with operational performance.

H3c: Economic practices are positively associated with operational performance.

The central theoretical proposition is that sustainability practices provide routines and commitments, whereas SCI provides a capability-based pathway through which those routines are translated into coordinated operating action. The mechanism involves converting environmental, social and economic information into shared plans, aligned supplier behaviour and responsive logistics. Similar conversion arguments appear in studies where supply-chain dynamic capabilities, integration, collaboration or innovation transmit the performance value of sustainability and digital resources ([6]; [13]; [17]; [19]).

The indirect-association argument does not imply that all direct associations disappear. EMP and SP may be associated with OP through internal process discipline and workforce or stakeholder mechanisms, while EP may be especially dependent on integration because inventory, lean and transportation routines require synchronisation across organisational boundaries. The relevant propositions are therefore:

H4a: Environmental management practices have a positive indirect association with operational performance through supply chain integration.

H4b: Social practices have a positive indirect association with operational performance through supply chain integration.

H4c: Economic practices have a positive indirect association with operational performance through supply chain integration.

The hypothesised relationships are summarised in Figure 1.

Figure 1. Conceptual model.

The study employed a cross-sectional explanatory survey design at the organisational level, with the manufacturing firm as the unit of analysis. The target population comprised manufacturing firms operating in Ghana. The national sampling frame was derived from the 2018 Integrated Business Establishment Survey Phase II (IBES), which identified 4375 registered manufacturing establishments across 23 subsectors. For substantive alignment with the sustainability focus of the study, the operational sampling frame was restricted to three manufacturing subsectors with a strong national presence and material environmental and social sustainability implications: food products and beverages (943 firms), wearing apparel, dressing and dyeing of fur (996 firms), and furniture manufacturing (613 firms). The resulting eligible sampling frame comprised 2552 manufacturing firms.

A multistage probability-based sampling procedure was used. First, eligible firms were organised into geographical clusters representing the middle and southern belts of Ghana, where the selected manufacturing activities were principally concentrated. Second, the firms were stratified by the three focal manufacturing subsectors to ensure representation across the industry categories. Within each stratum, firms were arranged alphabetically and selected systematically using the relevant sampling interval and a random starting point. The final target sample comprised 475 manufacturing firms, selected from the 2552-firm operational sampling frame before questionnaire administration. The achieved analytical sample also comprised 475 firms: food products and beverages (n = 167, 35.2%), wearing apparel (n = 158, 33.3%), and furniture manufacturing (n = 150, 31.6%).

One knowledgeable senior-level professional was designated as the key informant for each selected firm so that the observational unit corresponded with the organisational unit of analysis. Eligible informants included chief executives and operating officers, directors, general, supply-chain and purchasing managers, and personnel with substantive responsibility for manufacturing engineering, quality, research and development, purchasing, production, sales and marketing, human resources, finance and accounting, logistics, distribution or warehousing. Informants were selected because their positions provided direct knowledge of the firm’s supply-chain operations, sustainability practices and day-to-day organisational policies. The questionnaire also recorded respondents’ years of experience in their professional field and within the focal organisation as an additional indicator of informant competence. Each participating firm contributed one usable questionnaire; accordingly, the final analytical sample comprised 475 distinct manufacturing firms represented by 475 key informants.

The data-quality audit confirmed that every retained record contained complete focal responses. No non-numeric or out-of-range focal values, duplicate serial numbers or missing focal observations were detected. Six exact straight-liners, 16 long-string flags, 33 low-variance response patterns and two multivariate outliers at p < 0.001 were identified. These diagnostic flags were retained in the primary analysis because they did not by themselves establish invalid responding; their influence was examined through sensitivity analyses.

The focal item pool was adapted principally from the SSCM scale-development work of [11]. The archived questionnaire itself labels the SCI and OP blocks as “Das, 2018”, but does not provide a complete bibliographic entry; the traceable validated scale-development source is therefore reported as [11]. EMP was represented by three items concerning environmental-management systems, responsiveness to customers’ environmental concerns and reductions in material or energy requirements. The third EMP item is labelled EMP4 in the archived questionnaire but was coded EMP3 in the analysis dataset; its wording was unchanged. SP used five retained items concerning workplace safety, adequate remuneration, community education, employment/business opportunities and healthcare. These items combine two domains that [11] originally distinguished—socially inclusive practices for employees and socially inclusive practices for community. The present study treated them as manifestations of a broader social-practices orientation; this aggregation is supported empirically by the one-factor CFA and reliability results but remains a construct-level limitation because the employee and community domains are conceptually heterogeneous. EP comprised three items—transport economies of scale, scientific/JIT inventory control and lean production—adapted from Das’s Operations Practices domain and reframed here as economic-efficiency practices. SCI comprised three indicators concerning communication of future customer needs, shared production-plan adjustment and customer responsiveness. OP comprised three indicators capturing change in production cost and inbound and outbound logistics efficiency.

EMP, SP and EP used a five-point implementation-extent scale from 1 = Not at all to 5 = To a Great Extent. SCI used a five-point agreement scale from 1 = Strongly Disagree to 5 = Strongly Agree. OP asked respondents to assess change over the previous three years on a five-point scale from 1 = Strongly Deteriorated to 5 = Strongly Improved. The archived study materials available for this revision do not document a separate pilot-test, expert-panel refinement, or translation/back-translation procedure; none is therefore claimed. The questionnaire is in English and no translated version was identified. Full item wording, source-domain mapping and response anchors are reproduced in Appendix A (see Table A1). Item means ranged from 3.808 to 4.059 and all items showed concentration at the upper response category; the ordinal scale and ceiling concentration motivated robust maximum likelihood as the primary estimator and WLSMV as an ordinal sensitivity estimator (Table 1).

Table 1.Construct provenance, adaptation and response anchors.

Note. SP1 was excluded during the pre-model measurement audit, before CFA and structural estimation, because its wording overlapped with SCI3. No further indicator deletion was undertaken after model estimation.

The analysis was conducted in R using latent-variable structural equation modelling. Data screening preceded measurement and structural modelling. Reliability was assessed using Cronbach’s alpha, McDonald’s omega and composite reliability. Convergent validity was evaluated through standardised loadings and average variance extracted (AVE), while discriminant validity was assessed using heterotrait-monotrait ratios (HTMT), bootstrapped HTMT intervals, Fornell-Larcker comparisons and latent correlations.

The hypothesised five-factor measurement model was estimated using robust maximum likelihood (MLR) and compared with WLSMV. Alternative models combined the three SSCM dimensions, imposed a single common factor, or specified a higher-order SSCM factor. Global measurement-model fit was evaluated using chi-square, CFI, TLI, RMSEA and SRMR, with information criteria reported where appropriate. SP1 had been excluded before any CFA or structural-model estimation on the basis of questionnaire-content overlap with SCI3; this decision was not driven by model fit. Following that pre-model exclusion, no further indicators were deleted post hoc and no correlated residuals were introduced.

Common-method risk was evaluated using Harman’s first-component result, a one-factor CFA and full-collinearity VIF. A common latent method-factor model was attempted but was inadmissible because the latent covariance matrix was not positive definite; no defensible marker variable was identified. These diagnostics were treated as evidence about plausibility rather than proof that common-method variance was absent ([14]).

The structural specification estimated EMP, SP and EP as predictors of SCI and EMP, SP, EP and SCI as predictors of OP. The primary coefficients were estimated with MLR using std.lv = TRUE, a mean structure and FIML for missing data. Because non-parametric bootstrap confidence intervals were obtained in a separate estimation step, the identical measurement and structural specification was re-estimated with conventional maximum likelihood (ML), std.lv = TRUE and a mean structure. After a 2000-resample convergence/admissibility check, the final defined indirect and total associations used 5000 non-parametric resamples and percentile 95% bootstrap confidence intervals. This procedure follows the modern emphasis on directly estimating indirect effects rather than inferring them from the significance of component paths. A parsimonious control model included firm-size category, manufacturing sector and standardised respondent tenure. Robustness checks compared control specifications, outlier exclusions, a two-indicator SCI specification, WLSMV, complete-case and FIML estimation, observed-score path analysis with HC3 standard errors, and the SP1-retained sensitivity model.

As shown in Table 2, the five construct means were above the scale midpoint: EMP = 3.864, SP = 3.892, EP = 3.940, SCI = 4.029 and OP = 3.954. Standard deviations ranged from 0.906 to 0.993. The distributions were negatively skewed, consistent with the ceiling concentration observed at the item level, but no focal data correction or imputation was required. Complete-case and FIML estimation therefore used the same 475 observations.

Table 2.Descriptive statistics for focal constructs.

Note.Construct scores are shown descriptively; hypothesis tests are based on the latent-variable model.

As shown in Table 3, reliability and convergent validity were satisfactory for all constructs. Cronbach’s alpha ranged from 0.814 to 0.862, omega and composite reliability ranged from 0.815 to 0.862, and AVE ranged from 0.557 to 0.635. All MLR standardised loadings were positive and statistically significant, ranging from 0.671 to 0.848. SP2 had the smallest loading but remained substantively meaningful and was retained to preserve content coverage.

Discriminant validity was also supported. The largest HTMT ratio was 0.702 for SCI and OP, with a 95% bootstrap interval of [0.620, 0.774]. All HTMT values were below 0.85, no interval included 1, and each square root of AVE exceeded the relevant interconstruct correlations. The largest latent correlation was 0.695 between SCI and OP, which was high enough to be theoretically meaningful but not indicative of construct redundancy.

Table 3.Reliability, convergent validity and factor-loading ranges.

Note.All loadings were statistically significant at p < 0.001.

Table 4 summarises the measurement-model fit and comparison results. The hypothesised five-factor model demonstrated excellent fit under both estimators. MLR produced χ² (109) = 117.700, p = 0.268, CFI = 0.997, TLI = 0.997, RMSEA = 0.014, 90% CI [0.000, 0.029], and SRMR = 0.024. WLSMV yielded χ² (109) = 115.542, p = 0.316, CFI = 0.995, TLI = 0.994, RMSEA = 0.021, 90% CI [0.000, 0.039], and SRMR = 0.029. The largest modification index was modest, and the model was retained without post hoc residual correlations.

Alternative models performed materially worse. Combining EMP, SP and EP into one first-order SSCM factor produced CFI = 0.727 and RMSEA = 0.134; the one-factor model produced CFI = 0.588 and RMSEA = 0.163. A higher-order SSCM model fitted well in absolute terms, but robust comparison favoured the unconstrained five-factor model, Δχ² (4) = 13.587, p = 0.009. The results therefore support analysing the three sustainability dimensions separately.

Table 4.Confirmatory factor analysis and measurement-model comparison.

Note.Robust difference tests favoured the hypothesised five-factor model over all constrained alternatives.

Harman’s first unrotated component explained 35.81% of item variance. More importantly, the one-factor CFA fitted very poorly, and full-collinearity VIF values ranged from 1.270 to 1.819. The attempted common latent method-factor model was inadmissible and therefore not interpreted. Taken together, the diagnostics suggest that a single method source is unlikely to account fully for the covariance structure; however, the cross-sectional single-informant design means common-method influence cannot be ruled out.

As shown in Table 5, EMP (β = 0.377, p < 0.001), SP (β = 0.284, p < 0.001) and EP (β = 0.320, p < 0.001) were positively associated with SCI, supporting H1a-H1c. SCI was positively associated with OP (β = 0.440, p < 0.001), supporting H2. Because all exogenous predictors were freely covaried and direct paths from the SSCM dimensions to both endogenous constructs were estimated, the structural portion was just-identified conditional on the validated five-factor measurement model. Consequently, its global fit indices are identical to those of the measurement model and do not constitute an independent test of the directional or indirect structural specification. Structural inference is therefore based on path estimates, uncertainty intervals, explained variance and robustness analyses rather than on repeated global-fit statistics.

The direct EMP-OP association was positive (β = 0.247, p < 0.001), as was the direct SP-OP association (β = 0.176, p = 0.002), supporting H3a and H3b. The direct EP-OP path was positive but not statistically supported (β = 0.088, p = 0.120), so H3c was not supported. The three SSCM dimensions explained 49.3% of the variance in SCI, and the full structural specification explained 55.7% of the variance in OP. Figure 2 summarises the standardised structural relationships.

Table 5. Structural path estimates and hypothesis decisions.

Note.B denotes unstandardised coefficients; β denotes standardised coefficients. R²SCI = 0.493; R²OP = 0.557.

Figure 2. Standardised structural model.

As shown in Table 6, all three hypothesised indirect associations through SCI were supported by the percentile-bootstrap intervals. Using the separate ML bootstrap model with 5000 resamples, the EMP indirect association was 0.249 unstandardised and 0.166 standardised, 95% bootstrap CI [0.149, 0.384]. The SP indirect association was 0.188 unstandardised and 0.125 standardised, CI [0.099, 0.309]. The EP indirect association was 0.212 unstandardised and 0.141 standardised, CI [0.123, 0.330]. H4a - H4c were therefore supported as indirect-association hypotheses.

The EP pattern is especially informative. Although EP was positively associated with SCI and had a positive indirect association with OP, its direct latent path to OP was not statistically distinguishable from zero. The cross-sectional design does not establish temporal mediation; rather, the fitted model indicates that the operational relevance of economic practices in this sample was expressed principally through the SCI pathway.

Table 6. Bootstrapped indirect associations through supply chain integration.

Note. Percentile-bootstrap confidence intervals are based on a separate ML model with the same measurement and structural specification as the primary MLR model and 5000 non-parametric resamples.

The focal conclusions were stable across the principal sensitivity checks. Adding firm-size category, sector and respondent tenure changed no standardised focal coefficient by more than 0.01; SCI R² rose from 0.493 to 0.498 and OP R² from 0.557 to 0.562. Excluding the two multivariate outliers, estimating SCI with two indicators, using WLSMV, comparing complete-case with FIML, and estimating observed-score paths with HC3 standard errors did not change the conclusion that all three indirect associations through SCI were positive and statistically supported.

The observed-score model produced a small statistically significant direct EP-OP coefficient, whereas the latent SEM did not. Because the latent model accounts explicitly for measurement error and was specified after an acceptable measurement model, the hypothesis decision is based on the latent estimate. Sector multigroup analysis supported configural, metric and scalar invariance, and equality of structural paths was not rejected, robust Δχ² (14) = 13.302, p = 0.503. These subgroup results are exploratory. An alternative-ordering model fitted equivalently, underscoring that the cross-sectional data cannot establish temporal direction.

As an additional reviewer-requested sensitivity analysis, the full model was re-estimated with SP1 restored to the SP factor while all other indicators and structural paths were held constant. Although SP1 and SCI3 were not response-level duplicates (r = 0.058; 33.05% identical responses), SP1 had a near-zero, nonsignificant loading on SP (λ = −0.007, p = 0.872). Restoring SP1 reduced SP reliability (α = 0.862 to 0.809; CR = 0.862 to 0.812) and AVE (0.557 to 0.464), while HTMT for SP-SCI remained below conventional thresholds (0.488 versus 0.517). The absolute magnitudes and statistical support of SP → SCI and SP → OP were unchanged, R² remained 0.493 for SCI and 0.557 for OP, and the SP indirect association remained statistically supported with essentially the same magnitude. The apparent sign reversal in the SP coefficients under the six-item specification reflected arbitrary latent-factor orientation rather than a substantive reversal. These results support retaining the five-item SP specification on measurement grounds without changing the study’s structural conclusions.

EMP exhibited the strongest association with SCI among the three SSCM dimensions (β = 0.377, p < 0.001), supporting H1a. Theoretically, this result does more than confirm that environmental practices and integration coexist. It supports the dynamic-capabilities proposition that sustainability-oriented routines create value when firms possess mechanisms for sensing external requirements, seizing opportunities for resource efficiency and reconfiguring inter-organisational activities accordingly. Environmental management systems, eco-design and reductions in material or energy intensity generate information and coordination demands that extend beyond a single function. Their positive association with SCI therefore indicates that EMP is linked to the organisational capacity to translate environmental requirements into supplier specifications, production plans and coordinated operating responses. This interpretation is consistent with studies showing that green integration connects environmental practices to broader sustainability and performance outcomes ([16]; [17]; [19]), but the present evidence adds a more explicit capability implication: environmental routines appear to activate, or at least accompany, the integrative processes through which firms coordinate sustainability across organisational boundaries.

H3a and H4a provide an important qualification to a simple single-path account. EMP retained a positive direct association with OP (β = 0.247, p < 0.001) while also exhibiting a positive indirect association through SCI (standardised indirect = 0.166, 95% bootstrap CI [0.099, 0.256]). Dynamic Capabilities Theory would be overstated if these findings were interpreted to mean that integration is always necessary for environmental practices to be associated with operational value. Instead, the evidence suggests two pathways. Some environmental routines may yield relatively immediate internal benefits through waste reduction, process discipline and lower resource intensity, whereas SCI supplies an additional inter-organisational conversion pathway by connecting customer concerns, supplier communication and production planning. Thus, the theory is supported but also bounded: integrative capability enhances the operational deployment of EMP, yet it does not exhaust the mechanisms through which environmental practices are associated with performance.

SP was positively associated with SCI (β = 0.284, p < 0.001), supporting H1b, and was also directly associated with OP (β = 0.176, p = 0.002), supporting H3b. From a dynamic-capabilities perspective, the H1b result indicates that social sustainability is not merely a reputational or legitimacy-oriented domain; it also creates coordination requirements that firms must integrate into operational routines. Safety, remuneration and community commitments involve multiple stakeholders and organisational functions, so their implementation requires information to be interpreted, priorities to be reconciled and resources to be redeployed. In this sense, SCI provides a mechanism through which dispersed social obligations can be converted into coordinated operating responses. The result therefore extends the capability argument beyond the environmentally dominated applications common in SSCM research and is consistent with evidence that social responsibility contributes to sustainable performance when embedded in governance and coordination arrangements ([9]; [35]).

The supported indirect association for SP through SCI (standardised indirect = 0.125, 95% bootstrap CI [0.066, 0.206]) provides evidence for H4b, but the simultaneous direct SP-OP association again qualifies a universal conversion claim. If SCI were the only capability through which social practices became operationally consequential, the residual direct pathway would be difficult to explain. Instead, the findings imply that social practices may generate internal operational benefits—such as workforce stability, reduced disruption, stronger organisational commitment and stakeholder cooperation—while integration creates a complementary cross-boundary route through supplier communication and production planning. This distinction refines Dynamic Capabilities Theory in the present context: not every performance consequence of a sustainability routine requires reconfiguration across the supply chain. Some value may arise from relatively local organisational mechanisms, whereas dynamic or integrative capabilities become particularly important when the practice must be coordinated across functions and partners. The empirical coherence of the five-item SP factor supports this interpretation at the construct level, although the mixture of employee- and community-oriented practices remains a boundary condition for future research.

The EP findings provide the most stringent test of the capability-conversion argument. EP was positively associated with SCI (β = 0.320, p < 0.001), supporting H1c, but its direct latent association with OP was not statistically supported (β = 0.088, p = 0.120), so H3c was rejected. At the same time, the EP → SCI → OP indirect association was positive and statistically supported (standardised indirect = 0.141, 95% bootstrap CI [0.082, 0.220]), supporting H4c. This pattern is theoretically important because transport economies, inventory control and lean production are ostensibly designed to improve operations directly. Yet, once measurement error and SCI were accounted for in the latent model, their operational relevance was expressed principally through integration. The evidence therefore accords closely with the dynamic-capabilities claim that possession of efficiency-oriented routines is insufficient when those routines depend on coordination with external and internal actors.

Rather than simply being “consistent with” Dynamic Capabilities Theory, the EP pattern identifies a boundary condition under which the theory has particular explanatory force. Lean and just-in-time routines require reliable supplier deliveries, synchronized production schedules and timely demand information; transportation economies likewise depend on coordinated inbound and outbound flows. Without such integration, firms may relocate inventory, delay, or risk rather than eliminate it. The supported indirect pathway and unsupported direct latent pathway therefore suggest that SCI functions as a feasibility mechanism through which economic sustainability routines are recombined and deployed across organisational boundaries. This finding strengthens earlier evidence that operations and supply-chain strategies generate performance when aligned with integration capabilities ([27]; [32]), while also qualifying the assumption that economically oriented SSCM practices possess automatic operational value. Among the three dimensions, EP consequently offers the clearest empirical support for the paper’s central conversion proposition.

SCI was the strongest immediate predictor of OP (β = 0.440, p < 0.001), supporting H2, and the full model explained 55.7% of the variance in OP. This result provides the central test of whether the proposed mediator behaves as the type of capability posited by the theoretical framework. Dynamic Capabilities Theory argues that superior outcomes arise not simply from owning resources or routines but from the capacity to sense changing conditions, seize emerging opportunities and reconfigure resource combinations. The SCI indicators in this study—communicating future customer needs to suppliers, updating and sharing production plans, and responding to changing requirements—map onto this logic because they concern the interpretation of external signals, coordinated commitment of resources and adjustment of operating arrangements. The positive SCI-OP association therefore supports the proposition that integrative coordination is a consequential mechanism through which firms convert sustainability-related routines into lower costs and more efficient inbound and outbound logistics. This interpretation is consistent with meta-analytic and capability-based evidence linking integration to performance ([4]; [28]).

The findings nevertheless impose an important theoretical boundary. They do not establish that SCI is intrinsically or universally a dynamic capability, because routine information exchange and stable coordination can constitute ordinary capabilities. Nor do the cross-sectional data demonstrate that SCI developed before the observed performance differences. The stronger inference is narrower: the form of SCI measured here has capability-like properties, and the observed structural pattern is compatible with a conversion process in which coordinated sensing, planning and response help deploy SSCM practices. The alternative-ordering model fitted equivalently, which prevents the data from adjudicating temporal direction. Consequently, the study supports the explanatory usefulness of the dynamic-capabilities lens while withholding a stronger causal claim. This distinction is theoretically important because it converts the theory from a decorative label into a falsifiable interpretive framework: the evidence supports its coordination-and-reconfiguration mechanism, but the design cannot test the temporal process assumptions that distinguish a genuinely dynamic capability from a correlated organisational routine.

The first theoretical contribution is therefore a move from an adoption logic to a conditional conversion logic. The supported H1a-H1c and H4a-H4c pathways show that all three sustainability-practice domains are associated with SCI and possess positive indirect associations with OP through it. This pattern supports a core dynamic-capabilities proposition: organisational routines are not valuable solely because they exist; their performance implications depend partly on how firms combine and deploy them. Yet the simultaneous direct effects for EMP and SP demonstrate that this proposition should not be universalised. Integration is one conversion mechanism, not the only mechanism. The evidence therefore refines Dynamic Capabilities Theory by showing that the need for integrative deployment varies with the operational content of the underlying sustainability routine.

Second, the results establish dimension-specific theoretical boundary conditions. The measurement analysis showed that EMP, SP and EP are empirically distinct, and the structural model showed that their performance pathways are also non-equivalent. EMP and SP retained both direct and indirect associations with OP, whereas EP was expressed principally through SCI in the latent model. Theoretically, this means that SSCM should not be represented as a homogeneous resource bundle whose components depend identically on dynamic capabilities. Environmental practices may generate internal process efficiencies as well as integration-enabled benefits; social practices may operate through workforce and stakeholder stability alongside coordinated supply-chain action; and economic routines appear especially dependent on synchronisation across organisational boundaries. Collapsing these dimensions would therefore obscure when and why an integrative capability matters.

Third, the study clarifies what the evidence does—and does not—test about dynamic capabilities. The strong SCI-OP path, the three supported indirect associations and the stability of those conclusions across controls, outlier exclusion, the two-indicator SCI specification, WLSMV and observed-score analyses strengthen the argument that integration is a substantive conversion mechanism rather than a modelling artefact. At the same time, the equivalent fit of an alternative temporal ordering and the cross-sectional single-source design mean that the study cannot establish the evolutionary sequence of sensing, seizing and reconfiguring proposed by Dynamic Capabilities Theory. The contribution is therefore one of mechanism specification and theoretical qualification, not causal verification. SCI is positioned neither as a generic “good practice” nor automatically as a dynamic capability; rather, the findings show that its measured coordination behaviours are compatible with the capability logic and are especially consequential where sustainability routines, such as EP, require cross-boundary synchronisation. This interpretation extends capability-based SSCM research while preserving the distinction between evidence that supports a theoretical mechanism and evidence that would be needed to demonstrate that mechanism causally ([10]; [13]; [17]; [31]; [33]).

Managers should avoid treating sustainability and operations as separate portfolios. Environmental targets, social commitments and economic-efficiency initiatives should be embedded in supplier communication and production planning. Practically, firms should communicate anticipated customer sustainability requirements to suppliers, revise production plans jointly when those requirements change, and connect environmental and social indicators to inventory and logistics decisions.

The EP result has a particularly direct managerial implication: lean production, inventory control and transport consolidation may produce limited operational value when information remains fragmented. Managers should therefore evaluate whether suppliers receive timely demand and production information, whether purchasing and manufacturing schedules are synchronised, and whether customer feedback reaches upstream partners quickly.

The results also suggest that the three SSCM dimensions deserve differentiated implementation strategies. EMP appears to provide the largest integration leverage, whereas SP contributes through both integration and direct organisational stability. Managers should preserve these distinct pathways while using SCI as a shared platform that connects sustainability priorities to operational execution.

Several limitations circumscribe the conclusions. First, the cross-sectional design cannot establish temporal precedence. The indirect coefficients are compatible with the proposed conversion mechanism but do not demonstrate causal mediation. Longitudinal studies should measure SSCM practices before integration and operational outcomes, while quasi-experimental or panel designs should address reverse causality and time-varying confounding.

Second, the focal variables were collected in the same survey period from a single knowledgeable informant in each participating firm. Measurement-model comparisons and collinearity diagnostics weaken a simple single-factor explanation, but common-method variance and informant-specific perceptual bias cannot be eliminated statistically. Future research should therefore adopt longitudinal and multi-informant designs and, where possible, combine sustainability-practice reports from operational managers, integration measures from supply-chain partners and performance indicators from archival or other objective sources.

Third, OP is perceptual and narrowly defined around production cost and logistics efficiency. This focus improves construct precision but does not capture quality, flexibility, market, financial or environmental outcomes. Replication should examine whether the capability-conversion mechanism varies across performance domains.

Fourth, all items exhibited ceiling concentration, which may reduce discrimination among high-performing firms. WLSMV sensitivity analysis mitigates some distributional concerns, but future instruments should use more behaviourally specific response anchors or objective practice indicators.

Fifth, SCI contains only three indicators, and one item concerns responsiveness through finished-goods availability. Although the construct achieved good reliability, convergent validity and fit, future work should measure internal, supplier and customer integration more comprehensively. Sixth, the retained SP factor combines employee-oriented practices (safety and remuneration) with community-oriented practices (education, employment/business opportunities and healthcare), which [11] originally distinguished as separate domains. The acceptable one-factor measurement results support the pragmatic aggregate used here, but the conceptual heterogeneity remains a construct-level limitation and should be tested explicitly with multidimensional social-sustainability models in future studies. Finally, the sectoral sample is concentrated in food products, apparel and furniture manufacturing. Scalar invariance and equal structural paths were supported across sectors, but broader replication is needed before generalising beyond these industries or the study setting.