Cement is a hydraulic binding material that is central to concrete and mortar production (Taylor 1997). African cement markets have expanded alongside infrastructure development, urbanisation and regional trade, but their growth is accompanied by environmental, regulatory, quality-control and competition challenges (Roberts, Simbanegavi & Vilakazi 2023; Schmidt et al. 2018). In this setting, technical product performance remains essential, yet cement suppliers also compete through the reliability of their processes, explanations and relationship practices (Meyer & Pretorius 2023).

Cement is often treated as a highly commoditised industrial product. Nevertheless, repeated ordering, delivery coordination, allocation decisions and complaint handling create an ongoing service relationship in which organisational buyers evaluate both what they receive and how suppliers deal with them. Service fairness is therefore relevant because it captures buyers’ judgements about outcomes, procedures, interpersonal treatment and information provision across supplier encounters (Carr 2007).

Relationship marketing research explains durable business-to-business (B2B) relationships through credible exchange behaviour, relational value and expectations of continuity (Morgan & Hunt 1994; Palmatier et al. 2006; Rauyruen & Miller 2007). Fairness adds a more specific evaluative mechanism: buyers judge whether supplier outcomes are equitable, procedures are consistently applied, representatives act respectfully, and explanations are adequate. These judgements may become especially salient where buyers depend on reliable supply and cannot easily absorb delivery disruption.

The South African cement sector provides a pertinent setting for this analysis. Research has documented regional concentration, cartel conduct and competition concerns in African cement markets (Roberts et al. 2023; Vilakazi & Roberts 2019), while South African construction procurement is expected to operate transparently and competitively despite persistent implementation constraints (Tau, Ogunbayo & Aigbavboa 2024). Supplier relationship management is also important to participation and continuity in the local cement value chain (Rasdien, Pooe & Munyanyi 2024). These conditions make fairness judgements relevant to the credibility of supplier-buyer relationships.

Existing B2B studies have linked fairness to contract renewal, trust and loyalty, but much of the evidence concerns discrete services, service recovery or sectors outside industrial cement procurement (Cassia, Haugland & Magno 2021; Jambulingam, Kathuria & Nevin 2011; Oflaç, Sullivan & Kaya Aslan 2021). The simultaneous unique associations of the four service fairness model (FAIRSERV) dimensions with loyalty in South African cement supply relationships have received limited empirical attention.

To address this gap, the study integrated FAIRSERV with relationship marketing theory (RMT) and social exchange theory (SET). The model treated distributive, procedural, interpersonal and informational fairness as conceptually distinct but related evaluations associated with B2B customer loyalty. The study pursued the following research objectives (ROs):

The study contributes in three respects. Firstly, it applies the four-dimensional FAIRSERV framework to an industrial B2B procurement context. Secondly, it estimates the unique association of each fairness dimension with loyalty while accounting for shared variance among the dimensions. Thirdly, it explains the findings through RMT and SET by treating fair procedures and credible explanations as relational governance signals that may support continued exchange.

South African cement buyers operate in a market shaped by supplier concentration, procurement regulation and recurring coordination requirements (Tau et al. 2024; Vilakazi & Roberts 2019). Although fairness is relevant to such exchanges, existing evidence does not establish whether outcome fairness, procedural consistency, interpersonal treatment and information provision make equivalent unique contributions to loyalty. This omission limits theoretical understanding of multidimensional fairness in industrial relationships and provides managers with little guidance on which fairness practices warrant priority. The study therefore tested a focused model of four fairness dimensions and B2B customer loyalty.

Business-to-business marketing has progressively shifted from isolated transactions towards the management of enduring exchange relationships. Relationship continuity is supported when suppliers deliver value reliably and behave in ways that reinforce trust and commitment (Morgan & Hunt 1994; Palmatier et al. 2006). In industrial markets, where comparable products may be available from several suppliers, relational processes can differentiate suppliers beyond technical quality and price.

Evidence from B2B markets indicates that customer experience, satisfaction and relationship quality are associated with loyalty, although the strength of these relationships varies across contexts and customer groups (Human et al. 2020; Masitenyane & Dhurup 2023). Fairness provides a complementary explanation because it focuses on how buyers evaluate supplier conduct during exchanges rather than on overall service quality alone.

Service fairness refers to a customer’s evaluation of whether service outcomes, procedures, interpersonal conduct and explanations are justifiable and impartial (Carr 2007). In B2B settings, these evaluations arise across repeated transactions involving multiple organisational representatives. Fairness can therefore operate as a signal of whether a supplier is a credible exchange partner. Prior studies show that procedural and distributive fairness can shape contract renewal intentions in discrete B2B transactions, while interfirm justice is also associated with information sharing behaviour (Cassia et al. 2021; Huo, Liu & Li 2023).

In cement procurement, the four dimensions have concrete operational referents: (1) allocation and delivery outcomes reflect distributive fairness; (2) ordering and complaint handling rules reflect procedural fairness; (3) courtesy and respect reflect interpersonal fairness; and (4) explanations for shortages, price changes or delays reflect informational fairness. Examining the dimensions simultaneously is necessary because buyers may distinguish them conceptually while still evaluating some of them in closely related ways.

Business-to-business customer loyalty denotes an organisational buyer’s intention and commitment to maintain a supplier relationship over time. It includes continuity intentions and attitudinal attachment rather than a single purchase decision (Human et al. 2020; Rauyruen & Miller 2007). In South Africa’s concrete-products market, relationship quality and commitment have been linked to loyalty and repurchase intentions (Masitenyane & Dhurup 2023). The present study focused on loyalty as the substantive relational outcome and excluded a separate repurchase-intention construct because the questionnaire’s repurchase items duplicated the wording of the original loyalty items.

Relationship marketing theory explains how firms establish, maintain and strengthen exchange relationships over time (Berry 1983; Morgan & Hunt 1994). In B2B markets, continuity depends on credible conduct, mutual value and reliable interaction between exchange partners. Applied to cement procurement, RMT suggests that fairness evaluations inform whether a supplier is regarded as a dependable long-term partner. Consistent procedures and credible explanations are particularly relevant because they help buyers coordinate orders and interpret supply variations.

Social exchange theory complements RMT by explaining how relational conduct is evaluated through reciprocity and balanced exchange (Blau 1964; Cropanzano & Mitchell 2005). When suppliers apply procedures consistently, allocate outcomes fairly, communicate adequately and treat representatives respectfully, buyers may interpret these actions as evidence of goodwill and fulfilment of relational obligations. Loyalty can then be understood as a reciprocal orientation towards continued exchange. Conversely, unfair conduct may weaken the perceived balance of the relationship.

The study used a quantitative cross-sectional survey design. The target population comprised organisational buyers and users of cementitious products in Gauteng, including civil and building engineering contractors, concrete product manufacturers, and ready-mix concrete suppliers. These organisations were relevant because their operations depend on repeated cement procurement and supplier coordination. The sampling frame was drawn from entities registered with the Construction Industry Development Board (CIDB). The achieved complete-case sample of 208 observations exceeded the commonly applied minimum of 200 observations for a moderately specified covariance-based structural equation model and was evaluated alongside model complexity and fit (Hair et al. 2019).

After ethical clearance had been obtained, publicly available organisational contact details were used to approach eligible CIDB-registered entities. A link to the online questionnaire was distributed to 300 contacts between August 2025 and February 2026. A total of 220 completed questionnaires were received, giving a gross response rate of 73.3%. Seven multivariate outliers were removed using Mahalanobis distance across the substantive Likert-scale items (p < 0.001), and five further cases with missing observations on the AMOS indicators were excluded. The final complete-case dataset comprised 208 responses, representing a usable response rate of 69.3% of the distributed invitations. Because participation was voluntary and the sample was confined to formal, CIDB-registered organisations in Gauteng, inference is bounded to comparable organisational buyers rather than all cement users in South Africa.

The questionnaire contained demographic questions and five-point Likert-type measures ranging from 1 (strongly disagree) to 5 (strongly agree). The final CFA retained distributive fairness (DIST1–DIST3), procedural fairness (PROC1–PROC4), interpersonal fairness (INTER1, INTER3 and INTER4), informational fairness (INFO1–INFO4) and B2B customer loyalty (LOY3–LOY5). Interpersonal fairness 2 (INTER2) (‘My preferred cement supplier staff members treat me with an unbiased attitude’) was removed because its standardised loading in the initial model was 0.571 and its reference to an ‘unbiased attitude’ overlapped conceptually with bias-free procedures and outcomes. B2B customer loyalty 1 and 2 (LOY1 and LOY2) had initial loadings of 0.291 and 0.253, respectively. Their residuals displayed a large modification index (MI = 105.042). Removing these indicators reduced construct contamination and retained the three indicators that directly expressed loyalty and long-term commitment. The refined loyalty construct remained reliable (composite reliability [CR] = 0.921) and convergently valid (average variance extracted [AVE] = 0.799). Item removal was therefore guided jointly by construct meaning, standardised loadings and localised model diagnostics rather than by fit improvement alone.

Statistical Package for the Social Sciences (SPSS) was used to prepare the data, compute descriptive statistics and conduct supplementary one-way ANOVA. Analysis of Moment Structure (AMOS) was used to estimate the CFA and structural equation model following the two-step approach of Anderson and Gerbing (1988). Model fit was assessed using Minimum Discrepancy divided by Degrees of Freedom (CMIN/DF), Goodness-of-Fit Index (GFI), Normed Fit Index (NFI), Comparative Fit Index (CFI), Tucker-Lewis Index (TLI) and Root Mean Square Error of Approximation (RMSEA). Reliability and convergent validity were evaluated using Cronbach’s alpha, CR and AVE. Discriminant validity was assessed using both the Fornell-Larcker criterion and the heterotrait-monotrait ratio (HTMT); an HTMT value below 0.90 was used for conceptually related constructs (Fornell & Larcker 1981; Henseler, Ringle & Sarstedt 2015). A residual covariance between DIST1 and DIST2 was retained because both items referred narrowly to equitable outcome allocation; no cross-construct residual covariances were specified. For the supplementary ANOVA, effect size was reported as eta squared (η²). Tukey HSD comparisons were used when homogeneity of variance was tenable, whereas Games–Howell comparisons were used when it was violated (Hair et al. 2019). Preferred brand and duration of support were treated as exploratory grouping variables rather than structural predictors.

Ethical clearance was obtained from the School of Consumer Intelligence and Information Systems Research Ethics Committee of the University of Johannesburg (reference 2025SCiiS085; valid for 3 years). Written informed consent was obtained before participation. Responses were coded and analysed in aggregate form to protect respondent anonymity and confidentiality.

The sample comprised respondents with direct exposure to cement supplier relationships. Buyers (41.3%) and procurement officers (26.0%) formed the largest occupational groups. Most respondents (71.2%) had supported their supplier for at least 5 years, including 30.3% with relationships of 10 years or longer. Table 1 presents the profile of the 208 complete cases.

One-way ANOVA was used to examine whether construct scores based on the indicators retained in the final CFA differed across preferred brand and duration of support groups. Table 2 reports the omnibus tests and eta squared effect sizes.

Preferred cement brand was not associated with statistically significant differences in any retained construct. Duration of support was associated with distributive fairness, F(3, 202) = 5.565, p = 0.001, η² = 0.076; procedural fairness, F(3, 202) = 4.046, p = 0.008, η² = 0.057; informational fairness, F(3, 202) = 2.925, p = 0.035, η² = 0.042; and B2B customer loyalty, F(3, 202) = 6.015, p < 0.001, η² = 0.082. Table 3 reports the adjusted post-hoc comparisons for these significant duration effects.

The significant comparisons indicate that the shorter-duration groups generally reported higher distributive fairness and B2B customer loyalty than the two longest-duration groups. Procedural fairness differed only between the 1 year – 3 years and 5 years – 10 years groups. No adjusted informational fairness contrast was significant despite the significant omnibus test. These exploratory differences indicate group variation, not change caused by relationship duration.

The refined measurement model showed acceptable fit: CMIN/DF = 1.616, GFI = 0.915, NFI = 0.940, CFI = 0.976, TLI = 0.970 and RMSEA = 0.055. Retained loadings shown in Table 4 ranged from 0.592 to 0.982. Cronbach’s alpha ranged from 0.748 to 0.944, CR from 0.758 to 0.929 and AVE from 0.514 to 0.813, supporting internal consistency and convergent validity. The comparatively lower loadings for PROC4 (0.592) and INTER1 (0.606) were retained because they remained statistically meaningful, the corresponding constructs met the CR and AVE criteria, and the items represented relevant facets of the construct domains (see Table 4).

The Fornell-Larcker criterion was met for most pairs (see Table 5). The distributive-procedural correlation (r = 0.902) equalled the square root of distributive fairness AVE, indicating substantial empirical proximity and prompting the additional HTMT assessment requested by the reviewer. Table 6 reports the HTMT values.

All HTMT values were below 0.90. The highest value was 0.887 between distributive and procedural fairness. Discriminant validity was therefore supported under the 0.90 criterion, although the value confirms that buyers’ outcome and process judgements were closely related. The two constructs were retained because FAIRSERV defines them as theoretically distinct and because the structural coefficients should be interpreted as unique associations after controlling for their shared variance.

The model explained 56.5% of the variance in B2B customer loyalty. Procedural fairness (β = 0.438, p = 0.003) and informational fairness (β = 0.186, p = 0.006) had significant positive associations with loyalty, supporting H2 and H4. Distributive fairness (β = 0.252, p = 0.070) and interpersonal fairness (β = −0.088, p = 0.172) were not significant; therefore, H1 and H3 were not supported (see Table 7). Figure 2 presents the standardised paths and R².

The study makes three theoretical contributions. Firstly, it extends FAIRSERV from predominantly consumer service settings to an industrial B2B cement supply context. Secondly, it demonstrates dimensional equivalence within FAIRSERV: Procedural and informational fairness were significant, unique predictors, whereas distributive and interpersonal fairness were not. Thirdly, it integrates RMT and SET by identifying consistent procedures and credible explanations as concrete exchange signals through which buyers may judge supplier reliability and relational goodwill. The high distributive-procedural association additionally shows that theoretical distinctiveness does not necessarily imply strong empirical separation in every procurement context.

Cement suppliers should audit whether ordering, allocation, delivery scheduling and complaint-resolution procedures are applied consistently across organisational customers. Exceptions should be governed by explicit criteria and recorded so that buyers can understand how decisions were reached. Supplier representatives should also provide timely, specific and truthful explanations for shortages, delivery changes, price adjustments and service failures. Interpersonal courtesy remains an appropriate baseline standard, but the results indicate that courteous treatment cannot substitute for reliable procedures or adequate explanations. Because longer-duration groups reported lower scores on several outcomes, periodic relationship reviews may be useful, particularly for established accounts.

The cross-sectional design precludes temporal or causal inference. The sampling frame was restricted to formal CIDB-registered organisations in Gauteng, and voluntary participation may have introduced non-response or self-selection bias; the results should therefore not be generalised to informal buyers, other provinces or the entire South African cement market without replication. The small Mamba subgroup limited the precision of preferred brand comparisons. The high distributive-procedural association indicates partial construct proximity despite an HTMT value below 0.90, and the removal of INTER2, LOY1 and LOY2 may have narrowed the content represented by those scales. The ANOVA was supplementary and involved multiple exploratory comparisons. Future studies should use longitudinal or multi-source designs, test the measurement structure in an independent sample and examine whether dependence, contract type or service-failure conditions change the relative salience of the fairness dimensions.