Voices are important in many areas of marketing communications such as in tv and radio advertisements. Also in service interactions, for example, with frontline employees and voice-based technologies (e.g., smart assistants; Dawar & Bendle, 2018) voices are ubiquitous. Yet, it remains unclear which voice types companies and brands may use for different purposes as marketing research did only investigate single voice characteristics (e.g., Chattopadhyay et al. 2003). This is in contradiction with the notion of Gestalt psychology, which suggests that all characteristics of voice need to be examined together to grasp the holistic impression that sensory elements create (e.g., Koffka 1922). Therefore, the goal of our research is to systematically develop a voice typology and to examine the impact of different voice types on important marketing outcomes (e.g., brand personality).
To derive a voice typology, we adapt a procedure of Orth & Malkewitz (2008) and employ cluster and factor analysis. We systematically collect a sample of 462 voice recordings reflecting how voice is used in marketing communications (e.g., voices employed in ads) and a set of 850 adjectives used in research and practice to describe voices. In a large-scale study, 2,600 customers evaluate the voice recordings based on a systematically reduced set of 68 adjectives. Based on a preliminary analysis of the data of 1,700 customers, we find a five-cluster solution. Exploratory factor analysis reveals eleven factors: warm, bright, credible, vital, careful, surprised, heroic, compelling, excessive, strange, and uncertain. Based on the means of factors in the different clusters, we describe the voice types as warm, professional, rapid, dramatic, and normal. In addition to the adjectives, we asked participants to which degree brand personality dimensions are associated with different voice types. The analysis shows that, for example, warm voices score high on competence, excitement, sincerity, and sophistication. Dramatic voices score high on the dimension ruggedness. These results show that the voice typology is relevant for important marketing outcomes. The final results of this study will be presented at the conference. In a future study, we will experimentally manipulate voice types in a scenario experiment and show the robustness of the effects on brand personality in a controlled experiment.
We contribute to research as we develop a voice typology. The voice typology creates a better understanding of voice in marketing research and enables researchers to systematically examine voice. Additionally, we show the effects of voice types on marketing outcomes enabling more detailed examinations in the future. Practitioners can benefit from our typology as it helps to understand the voice options available for marketing managers. Additionally, we guide the choice of voices for different purposes and brands as we show which voice types may be used to evoke desired outcomes.

Purpose: Customers are increasingly confronted with services robots. Unlike machines such as self-scans and ATMs, service robots are typically anthropomorphized (i.e., being experienced by customers as being more “humanlike” than “machinelike” ) because of their design. Indeed, existing research showed that customers are more likely to imbue human characteristics to robots when their embodiment is more humanoid compared to mechanics. Despite its importance, service robot research has yet not focused on the anthropomorphic design caused by linguistic human cues. This is surprising since machines such as service robots also need to communicate with customers to provide the service, such that verbal communication features (such as tone, speed, pitch, excitement) become prevalent. Like the service robot appearance (i.e., hardware), the speech of the service robot (i.e., software) can be designed to be more humanlike versus machinelike. Bridging the service and communication literature, the current research aims to unveil the impact of linguistic human cues on customers’ reactions toward the service robot.

Design: This study adopts a multiple-method approach: (1) structural equation modeling (SEM) to quantify the main effects of the variables under investigation (2) lexically based text mining to illustrate the meaning of the variables being tested, including expressed sentiments, and (3) fuzzy set qualitative comparative analysis (fsQCA) technique to detect contrarian cases, and various routes that can lead to favorable customer outcomes, thereby allowing alternative routes that deviate from the findings of the main effects from SEM. In total 360 subjects participated in our pretest and main study, which focuses on the hospitality industry (here, check-in and restaurant at the airport).

Findings: Our pretest confirms that it is both viable (here, by using a voice simulator) and realistic (i.e., scenario realism) to manipulate the linguistic cues of a service robot, in which a more anthropomorphized linguistic design is associated with a more extraverted service robot personality – in line with the communications literature. SEM further unveils that a more extroverted service robot is associated with more warmth, competence, overall experience, trust, and behavioral intent, but also more discomfort. These relationships are mediated via the perceived anthropomorphism of the service robot. In addition, our results show that contextual effects matter since, a heterophily effect (i.e., when the service robot and the customer have opposite personality traits with regard to extraversion/introversion) and hedonic (vs. utilitarian) service encounters result in better customer outcomes. Our text mining results additionally reveal the specific sentiments that are experienced when being confronted with these service robots. Finally, our fsQCA’s results provide evidence that alternative routes of service robot design x customer personality x service encounter type x anthropomorphism level can lead to favorable customer outcomes.

Implications: The personality of a customer or a human service provider is difficult to manipulate; in the contract, the personality trait of service robots can be easily manipulated by design as it relates to the anthropomorphic linguistic cues. Our findings help managers to better understand which linguistic cues lead to the best outcomes, thereby taking individual and situational contextual heterogeneity into consideration.

Purpose – The importance of customers and their perceived value is widely accepted. Companies are increasingly attentive to customers (Strandvik et al., 2012) in order to design offerings expected to have value in use and context (Chandler and Vargo, 2011; Vargo and Lush, 2016). In B2B relationships the value concept is actor-specific (e.g. Hedaa and Ritter, 2005; Tuli et al., 2007). However, while value as a phenomenon is extensively investigated (Ravald and Grönroos, 1996; Zeithaml, 2020), actors have different perspectives and understandings of value, and this is largely missing from the literature (Rintamäki and Saarijärvi, 2021; Strandvik et al., 2012).
The diffusion of AI-driven technologies represents a great challenge for companies to address customer value (Nylén and Holmström, 2015; Wirtz et al., 2018; De Keyser et al., 2019). Recently, firms are introducing new intelligent tools to ensure continuous and immediate interaction with customers (Van Pinxteren et al., 2020; Selamat and Windasari, 2021). Mostly known as “chatbots”, but used with several synonyms, they allow companies to communicate in a way that resembles a human-human conversation (Epstein and Klinkenberg, 2001). While research about the benefits of chatbots is emerging (Johannsen et al., 2021), what needs to be considered more explicitly is the different perspectives of the value inherent in them. The development of this technology involves two different players; on the one hand the provider who develops the technology, and on the other the customer who uses the technology to improve interactions with their end-customers. Understanding potential differences in how the provider and customer perceive the value in use of the chatbots is an important task to successfully design and adopt them in the service encounter.
This paper, therefore, aims to explore the dyadic perceptions of conversational agents from the dyadic perspective of the provider and the customer.

Methodology – The study adopts a qualitative approach, often used in social and business studies (Cassell et al. 2006). The data collection conducted in a B2B setting is based on eight semi-structured interviews with four providers and four customer companies. A ‘constant comparative analysis’ (Dubois and Gadde, 2002) was used to abstract the dyadic perspective. Secondary data, such as business news, company websites and reports, were also used.

Findings – The provider’s value proposition does not always match the customer’s value in use. Providers indicate that chatbots allow firms to always be present for the customers, automate some processes and reduce costs. A six dimension model depicts the mismatch between how the provider and customer view conversational agents: providers focused on technical features such as usability, responsiveness, natural language processing, while customers highlighted more complex benefits such as chatbots’ ability to convey engagement, enable cross-selling and facilitate problem-solving.

Originality – The study contributes to the service literature by revealing the difference between customer’s needs and provider’s offering, thus facilitating the analysis and characterization of the value in use and offer concept for chatbots.

Chatbots have a track record of being a valuable and cost-effective means to handle service requests, bridging firm and customer communications. Coupled with artificial intelligence methods (e.g., sentiment analysis, natural language processing), chatbots are able to simulate human-like conversations by assisting and collecting data from customers in a similar fashion to their frontline employee counterparts. Due to their versatility, their usage as a text-based service communication channel increased by 92% since 2019 (Drift, 2020) and they are predicted to handle 75-90% of healthcare & banking queries by 2022 (Juniper Research, 2020) making their design choices critical in order to avoid service and customer experience failures.

Chatbot interactions occur via text message exchanges between the conversational agent and customers, generating a real-time dialogue. Such text-based communications are claimed to have reduced capacity for emotional exchange, an essential component in service encounters. One of the few nonverbal clues that carry emotions in written communications are typefaces (i.e., font styles). Research suggest that typefaces can elicit certain emotions and convey meaning beyond the semantic written content. For example, Serif fonts can project a sense of tradition and respectability, while Sans Serif fonts convey neutrality and directness. Script fonts on the other hand provide the feeling esthetics and beauty of calligraphy and handwriting.

Chatbot design features aid in how it is perceived as a conversational partner and are critical in delivering positive customer experiences. Yet, research into how typefaces, a core element of text-based communications, influence chatbot perceptions by customers remains sparse. Furthermore, typefaces possess the ability to trigger customer thoughts and feelings in specific contexts, and thus could be perceived as appropriate or not depending on the expected message they are attempting to convey. This is also an aspect of typefaces that could aid in optimizing chatbot design depending on context, leading to improved customer experiences.

To address these gaps, this study seeks to understand how different font types employed in chatbot interactions influence customer experiences. Specifically, this study aims to tackle the following main issues. (1) The effect of how font types may imbue the chatbot with specific personality traits that in turn influence how the customer perceives and experiences the service encounter, (2) the effect of specific chatbots typefaces based on specific contexts (e.g., a comic font may be perceived as inappropriate in a serious banking encounter), and (3) aims to provide a research agenda to aid academics and practitioners in further exploring and optimizing the effects of chatbot fonts to provide better customer experiences.

Theoretically, this study aims to fill a gap in the technology and customer experience service literature, aiming at developing a deeper understanding of how typefaces influence customer experiences in chatbots encounters. Managerially, this study aims to inform and advise marketing executives and chatbot service designers about the importance of the often-undermined effects of font styles in the chatbots design decision making process.
References are available upon request.