Effective Chatbot Conversations: Experiments with Bot Identity and Tone of Voice

Firms have been around the online world long enough to learn the skill of attracting visitors to the company website, whether by organic or paid means. What’s lagging behind is their ability to convert that site traffic into new business. Average conversion rates are in most cases alarmingly low and, by physical store measures, practically suicidal. It is no surprise then that the marketing technology scene is booming with providers who obsess over driving conversion rates and online sales up. One of the most prominent solution categories in the space is chatbots and other conversational interfaces, which can be used as the corporate website’s lead generation and sales engine, customer service function, and to bring overall enhancements to the online customer experience. Chatbots are by no means a new innovation, but they have so far received little direct academic attention. More explored themes in the same domain include website interactivity, the state of flow and concepts related to online consumer behavior. This thesis contributes to the existing bodies of research by studying the effectiveness of chatbots on three live company websites with a true experimental research design. More specifically, the focus is on discovering how chatbot identity (brand – human) and tone of voice (formal – casual) influence the effectiveness of chatbot conversations in terms of conversions and engagement in dialogue. Findings from three conducted studies give no conclusive answer to whether a bot with a branded or human identity, or formal or casual tone of voice performs more effectively than its counterpart. In some of the experiments, human identity and casual tone yield the best results, but in others, tables are turned in favor of a branded identity and formal tone of voice, or one of the other combinations. Even though the variants do not in all cases produce significantly different results, there is solid indication that choices related to bot identity and tone of voice matter. Conclusions of the thesis suggest that the choice between a branded or human bot identity and formal or casual tone of voice should be made with the specific brand in mind. Furthermore, it is found that bot variants behave differently when placed in different stages of the buyer’s journey. There is also evidence that suggests a categorical difference in user preferences between the contexts of B2B and B2C. These and other observations are listed as potential avenues for future research. Overall, the set of mixed findings stresses the importance of understanding customers’ needs and motivations, as well as their relationship with and expectations toward the brand. When the chatbot succeeds in addressing users’ most pressing questions, it presents an attractive proposition for improving the website conversion rate and customer experience. But structure alone is not enough – results of the conducted studies show that matching the bot’s identity and tone of voice with the brand is a reliable source of beneficial behavioral outcomes