This paper utilized the dual coding theory (Yang, Hlee, Lee, & Koo, 2017) and theory of latent state–trait (LST) (Steyer, Schmitt and Eid, 1999) to the application of consumer impulse purchasing behaviors and further revealed that the number of reaching can thus be identified as a trigger for impulse buying on social media. Three inputs (number of color, text length, number of photo) and three outputs (number of likes; number of comments, and shares; and number of clicks on post) are used to develop this marketing message performance assessment model in the social medial based on the literature and expert opinions through data envelopment analysis (DEA). The disaggregate efficiencies are also assessed in order to improve the individual input resource performance in a total-factor framework. Resource-saving target ratios (RSTR) for 60 marketing message in a five-star hotel chain. The empirical findings indicate that the average total-factor text length efficiency (TFTLE) is worse than the total-factor photo efficiency (TFPE) in the Facebook platform in the five-star hotel chain. This result suggests that photo message is more attractive for viewer than the text message. Managerial discussion and the future studies are discussed.