What drives virality of online complaints? The critical roles of content and non-content factors

Abstract

The advent of digital environment has provided ample avenues for consumers to voice their complaints. Meanwhile, it has also removed the restrictions on others’ participation in these conversations. Some phenomenal examples including “United Breaks Guitars” on YouTube in 2009 and “United overbook flight 3411” on Twitter after 8 years, United Airlines obviously failed to progress in dealing with online complaints properly. Any online complaint may be substantially discussed, supported, and shared, however, not all of them are. To predict and manage complaints before they become viral is a critical but challenging task for both researchers and managers for several reasons. First, the volume, velocity and variety of user-generated content online are massive, thus, requires tremendous efforts and resources to capture, distinguish, monitor and analyse the complaints and exclude irrelevant information. Second, understanding complaints virality is a challenging task, however, there is no definite strategy or pattern for researcher’s and manager’s reference. Taking other situational factors into consideration (e.g., the traits of the industry, the equity of the involved brand, and the resource of the organization), investigating online complaints for a specific industry or company tend to be case by case analysis rather than rely on other’s experience or existing works. Finally, after analysing the complaints, what response strategy should be adopted is still unclear, which is trickier on public platforms that information is access to broad audience and online firestorm can happen without any warning. To have a comprehensive understanding of complaint virality and aim to propose a more practical method for conducting similar research, this thesis investigated various potential factors for complaint virality from diverse aspects.

A text-mining study was conducted in support of this research. Web scraping was applied to obtain complaints and relevant information from Twitter, followed by natural language processing techniques for data pre-processing, and various big data analysis techniques were adopted and compared to explore all potential factors of complaint virality. Results confirm the importance of the complainer’s and the organisation’s characteristics as well as the linguistic and psychological attributes of the negative Tweet in predicting complaint virality. The pattern of organisational response and its impact on the virality were also investigated. Finally, the interactive effects of the content attributes and topics were confirmed.

The findings of this study prove that both central and peripheral routes will come into effect when readers react to complaints on social media. The number of follower a complainer has is a predominant factor of complaint virality which is in line with the social network theory. Meanwhile, physical cues of complaints, such as word count and use of attachments, work as obvious signals for readers to assess the complaints. The density of anger is found to trigger reader’s support, confirming the action-stimulating effect of high arousal emotions. Readers are also found more likely to be influenced by expressions with higher social confidence, but they are less likely to support subjective complaints. Furthermore, different complaint topics are found to cause the variance of virality, and the attributes of complaints moderate this relationship. Finally, organisational response is proven to decrease the possibility of complaint virality. More importantly, the tipping point of response effectiveness is found to be three days in this case. These observations provide guidance on how to decide which complaints to respond and when to respond.

More information

Identifiers

ORCID for Paurav Shukla: orcid.org/0000-0003-1957-8622

ORCID for Martin Kunc: orcid.org/0000-0002-3411-4052

ORCID for Youngseok Choi: orcid.org/0000-0001-9842-5231

Catalogue record

Export record