Data modalities, consumer attributes and recommendation performance in the fashion industry

Abstract This paper investigates determinants of recommendation systems’ performance in an online experiment a large European Internet footwear store. By combining transactional data and archival customer records, unique database was compiled from which proxy variables were extracted to represent dimensions consumer loyalty shopping involvement. These combined regression analysis with technical characteristics two types algorithms employed for generating recommendations: the EMDE algorithm, relying on LSH method, industry-standard CF-RS. Statistical reveals that recommendations are more successful when visual modality is behavioural data. Better found be associated lower levels involvement shopping, as well higher trust engagement vendor. Experience vendor showed negative correlation through both its main effect by interactions other consumer-related variables.