Mining Twitter lists to extract brand-related associative information for celebrity endorsement

Abstract

Twitter lists (i.e., curated collections of Twitter accounts) are user-generated and serve primarily as a tool to group other users. Grouping judgments are grounded in the implicit assumption that co-listed members share common associations. As such, Twitter lists are ideal for directly exploring associative links between brands and/or other entities. This research capitalizes on Twitter list membership data to provide a new metric indicating the similarity of users’ list membership profiles. This metric is used as a proxy for perceptions of brand–celebrity (mis)fit (i.e., the degree of congruency or similarity between the celebrity and the brand) in celebrity endorsement situations, where a celebrity's fame or social status is used to promote a brand. To validate the accuracy of the method, we compare the list similarity metric with directly elicited survey data for a test set of 62 celebrities and 64 brands, ranging across eight industry sectors. This research contributes to the extant literature of studies extracting brand-related associative information (i.e., information held in consumers’ memory that contains the meaning of a brand) from large volumes of consumer online data. This research also introduces new ways of data mining to operational research literature and provides managers with a new methodology to directly infer perceptions of brand–celebrity (mis)fit.