Visual electronic Word of Mouth: a multimodal brand approach and case study

An emerging activity on internet is to create and share visual content, our understanding of this activity and its impact however is limited. In this paper we aim to define and operationalize the concept of visual eWom and embed it in the current eWom literature. Different from existing eWom research which relies on textual information only, we conceptualize visual eWom as a multimodal construct consisting of textual and visual concepts. We test our operationalization on a dataset of 6435 consumer posts crawled from Instagram. We apply stateoftheart machine learning techniques, convolutional neural networks, for detecting the visual concepts in images posted on Instagram. We use OLS regression to test the impact of textual and visual concepts on image popularity. The results in our case study show that textual and visual concepts provide complementary information and have a different impact on image popularity.