Consumer sustainability judgments in omnichannel luxury retail: social media engagement and perceived last-mile delivery performance in Saudi Arabia

Consumers may form sustainability judgments from general brand-engagement and service-performance cues even when those cues do not directly measure sustainability communication or environmental performance. This study examines the associations of social media engagement (SME) and perceived last-mile delivery performance (LMD), operationalized as respondent-reported delivery-performance perceptions, with customer satisfaction, consumer sustainability perception, and attitudinal brand loyalty in Saudi omnichannel luxury retail, while testing digital literacy and brand perception as moderators of the SME–sustainability association. A cross-sectional English-language survey conducted in Jeddah and Riyadh yielded 900 completed submissions; sequential eligibility screening followed by a post-eligibility response-quality review produced a primary analytical sample of 384. The survey did not separately verify personal receipt of a focal-brand delivery or the informational source underlying each LMD rating; accordingly, LMD may combine experience-based evaluations with inferred or halo-based brand beliefs. The scoring strategy uses compatible response formats: CS1–CS4 for satisfaction, S1–S3 for sustainability perception, BL1–BL3 for loyalty, SME1–SME2 for social media engagement, and LE1–LE3 for delivery-performance perception; binary CS5 and SME3 are analyzed separately rather than combined with the Likert composites. All hypotheses were estimated using HC3 heteroskedasticity-robust inference. In the primary N = 384 analysis, SME was positively associated with satisfaction ( β =.277, p < .001) and loyalty ( β = .351, p < .001), but not with sustainability perception ( β = .136, p = .084). LMD was positively associated with satisfaction ( β = .199, p = .023), sustainability perception ( β = .288, p < .001), and loyalty ( β = .205, p = .020), and the LMD–sustainability association remained significant after adjustment for brand perception, satisfaction, and respondent-profile controls (B = .189, p = .006). A sample-selection sensitivity analysis using all 453 eligibility-qualified submissions preserved H1, H2, H3, H5, H6 and the fully adjusted LMD–sustainability association, while H4 and H8 were sensitive to inclusion of the 69 quality-screened submissions; H7 remained unsupported. The findings therefore support a cautious cross-touchpoint inference account while showing that H4 and H8 are sample-selection-sensitive and that H7 remains unsupported with an unstable direction.