Mining delivery customer claims in social media in Colombia. An exploratory analysis applying machine learning algorithms.

Abstract

The amount of available data, computational developments, and growing social demands from customers pose enormous challenges for delivery companies. Recognizing this need, we sought to investigate how delivery customer claims through Twitter. For this purpose, we used a sequential mixed methodology, starting with a literature review by applying a bibliometric analysis and its subsequent interpretation through content analysis. Subsequently, we scraped the mentions made by different users to deliver brands on Twitter to consolidate a significant corpus of data that would allow us its subsequent review through an exploratory analysis. Finally, applying natural language processing and machine learning algorithms, we discovered how users of these delivery companies usually compare brands with their competitors when complaining. We specifically highlighted the psychological bias of users, who polarize between love and hate for brands, respond to other users' posts, and achieve meaningful interactions with their likes as the main predictors of this type of claim.