The effects of visual congruence on increasing consumers’ brand engagement: An empirical investigation of influencer marketing on instagram using deep-learning algorithms for automatic image classification

https://doi.org/10.1016/j.chb.2020.106443Get rights and content

Highlights

  • Social media Influencers post visual content congruent with Followers' interests.

  • Visual congruence increases followers' engagement with Influencers' posts.

  • Such an increase in turn augments followers' engagement with the endorsed brand.

  • Affiliation between Influencers and Followers mediates the above relationships.

  • Deep-learning algorithms can automatically classify visual posts on social media.

Abstract

Influencers are non-celebrity individuals who gain popularity on social media by posting visually attractive content (e.g., photos and videos) and by interacting with other users (i.e., Followers) to create a sense of authenticity and friendship. Brands partner with Influencers to garner engagement from their target consumers in a new marketing strategy known as “Influencer marketing.” Nonetheless, the theoretical underpinnings of such remains unknown. We suggest a new conceptual framework of “Visual-Congruence-induced Social Influence (VCSI),” which contextualizes the Similarity-Attraction Model in the Social Influence literature. Using VCSI, we delineate how Influencers use visual congruence as representations of shared interests in a specific area to build strong bonds with Followers. This intimate affiliation catalyzes (i.e., mediates) the positive effects of visual congruence on Followers’ brand engagement. To test these hypotheses, we conducted in vivo observations of Influencer marketing on Instagram. We collected >45,000 images and social media usage behaviors over 26 months. We then applied deep-learning algorithms to automatically classify each image and used social media analytics to disclose hidden associations between visual elements and brand engagement. Our hypothesis testing results provide empirical support for VCSI, advancing theories into the rapidly growing fields of multimodal content and Influencer marketing.

Introduction

“Influencers” are ordinary individuals, not celebrities, who have amassed large numbers of Followers on social media sites by posting visually attractive content that showcases their lifestyle and merchandise preferences (Cotter, 2019). “Followers” are those individuals who subscribe to Influencers' content, and some Influencers boast tens of thousands of Followers, creating “fandom” (Abidin, 2018). Unlike celebrities, Influencers cultivate a sense of intimacy among their Followers through sharing authentic and lived experiences in the areas in which they claim expertise (Cotter, 2019). The growth of mobile applications for image-sharing, such as Instagram, has fueled the rise of Influencers (Marwick, 2013). These Influencers have surpassed celebrities as the favorite social media personalities among millennials, who have become the largest purchasing age group in the U.S. since 2019 (Fry, 2018). Recognizing millennials' purchasing power, brand managers collaborate with Influencers who have built fame in brand-pertinent areas, hoping to connect with large crowds of consumers in their niches (Abidin, 2018). An example is the beauty brand, Glossier, which enlists make-up experts to showcase Glossier products in the brand's social media posts (Ravi, 2018). Glossier is valued at 1.2 billion USD as of November 2019 (Roof & Chernova, 2019).

Nonetheless, academic research on Influencers is lagging in three intertwined aspects. First, most prior studies have focused on textual comments, omitting visual elements in posts due to the challenges in analyzing a large number of images posted daily. Second, this omission of visual elements in the research literature hinders systematic, in vivo observations involving the effectiveness of Influencers because Influencers employ images and photos as their primary means of gaining visibility among their Followers (Cotter, 2019). The lack of in vivo observations in turn limits the supply and availability of empirical support to facilitate theory advancement. As a result, a theoretical framework to explain Influencers’ roles in increasing consumer brand engagement has yet to be fully explicated. Closing the gaps in these three areas is therefore pivotal in expanding our understanding of the rapidly growing Influencer marketing. This growing field of research requires rigorous testing and firm grounding in its theoretical foundations.

To close the first and second gaps ([i] the absence of image analysis and [ii] the in vivo observations of Influencers), we employed both deep-learning algorithms and social media analytics. Deep-learning algorithms, the best-known example being Convolutional Neural Networks (CNN), have recently emerged as a robust method for classifying large numbers of images (LeCun, Bengio, & Hinton, 2015). We collected visual elements in Influencers' and their Followers' posts (>45,000 images). We then classified the themes of each collected image by fine-tuning three pre-trained CNN models. Simultaneously, we collected Influencers, their Followers, and the endorsed brand's social media data from Instagram over a data collection period of 26 months. Next, we identified associations underlying how Influencers affect their Followers' brand engagement via visually attractive content, employing social media analytics. These social media analytics allow us to identify hidden patterns in the “Big Data” collected from real-world observations (Aral & Walker, 2014).

The use of this expansive methodology renders empirical support that warrants adequate grounding in proposing a theoretical framework for Influencers’ effectiveness. In particular, we contextualize the Similarity Attraction Model (SAM) in the Social Influence (SI) literature. SAM suggests that shared attitudes, interests, and opinions are predictive of frequent interactions and affiliations between two parties in a dyadic relationship (Byrne, Griffitt, & Stefaniak, 1967). Similarity in SAM has traditionally been operationalized from textual comments, which impedes SAM from being applied to multimodal elements that are increasingly prevalent in social media posts (Highfield & Leaver, 2016). Empowered by the deep-learning algorithms and social media analytics mentioned above, we propose a new concept, visual congruence, which denotes the extent to which the themes of images posted by two parties overlap. We then argue that Influencers carefully curate visual congruence in their posts in order to accentuate shared interests with their Followers in their attempts to attract these Followers to their content, based on SAM.

Next, we posit that these increased affiliations between Influencers and Followers induce Followers to engage with the brand's posts that Influencers endorse, based on the SI literature. SI refers to the influences that individuals exert on the ways in which others expect a product's utilities (Godes et al., 2005). SI augments the propagation of ideas and economic behaviors throughout social networks (Aral & Walker, 2014). One's SI on the other party in a dyadic relationship increases when the two parties are engaged in frequent and intimate interactions (Aral & Walker, 2014). Thus, we argue that frequent interactions, which Influencers garner from their Followers through creating visual congruence, facilitate Followers' brand engagement. Accordingly, this framework suggests the mediating role of strong affiliations between Influencers and their Followers in catalyzing the positive impact of visual congruence on Followers' brand engagement. To denote the new mechanisms of conducting influence through visual elements, mediated by increased interactions, we suggest “Visual Congruence-induced Social Influence.”

In conclusion, we extend the application of SAM from textual to visual realms and contextualize SAM into SI to delineate the process by which visual congruence increases Followers' engagement with the brand. In so doing, we contribute to knowledge advancement in not only the Influencer marketing field but also SAM and SI. We also propose an expansive methodology that will aid in future researchers who are challenged in their pursuit of this rapidly growing, yet understudied area due to technical difficulties.

Section snippets

Growth of influencers

In order to amass and maintain a large number of Followers, Influencers create visually attractive content in a niche area in which they claim expertise (Lueck, 2015). Well-known examples of these niches include health (specialty foods and cooking), beauty (fashion and makeup), fitness (workout and body images), and video games (Abidin, 2016; Park, Ciampaglia, & Ferrara, 2016). There is a growing need among social media users to learn from other ordinary people's authentic experiences in these

Contextualization of the Similarity-Attraction Model in the social influence literature

As the overarching theory of our investigation, we chose SAM, which goes back to Newcomb in his 1956 work on the prediction of interpersonal attraction. In that work, Newcomb claimed that attitudinal similarity is the strongest predictor of interpersonal relationships. Inspired by Newcomb's work, Byrne and his associates have conducted extensive studies over three decades and have confirmed the importance of similarity in attraction: “the expression of similar attitudes by a stranger serves as

Hypothesis development

Fig. 1 delineates our research model. We argue that visual congruence, manifested in Influencers' posts with their Followers in the areas that are pertinent to the brand, will increase Followers' engagement with the Influencers' posts (H1), and in turn with the brand's posts (H2).2

Data collection and selection of methods

We collected Instagram data during two time periods (T1 and T2), each of which covers 13 months (a total of 26 months in Table 1). Collecting observations for more than two years allows us to examine how Influencers earn engagement from their Followers, and how they gradually pass positive brand attitudes to their Followers. That is, an observational period of 26 months enables us to capture the longstanding effects that Influencers have on their Followers’ brand engagement, whereas a

Deep learning algorithms for image classification

We developed deep-learning algorithms to classify three visual themes—(i) physical activities (to measure visual congruence for hypothesis testing), (ii) Lululemon-style clothing (which was not used for hypothesis testing, but was necessary to confirm that the Lululemon ambassadors featured the brand in their posts in Section 5.2, Table 2), and (iii) pets (which were not used for hypothesis testing, but were used for one of the robustness checks in Section 7.4, wherein detailed justifications

Follower's engagement

We counted the number of times the Follower liked or commented on the Influencer's posts in T1 as a measure for F–I engagement. Recall that Instagram does not provide sharing or reposting features. In the same vein, we counted the number of times the Follower liked or commented on the brand's posts in T2 as a measure for F–B engagement. The reason we observed F–I engagement in T1 and F–B engagement in T2 is that we wanted to see how F–I induces F–B engagement, while preventing reverse causality.

Summary of the findings

Influencer marketing is growing rapidly, especially among millennials, who have become the largest purchasing age group in 2019. Nonetheless, academic research has not yet fully explicated how Influencers garner their Followers' brand engagement. As such, we have maintained that visual congruence, which is manifested in Influencers' posts to accentuate shared interests with their Followers, is positively associated with increases in Followers' engagement with Influencers, based on SAM. This

CRediT authorship contribution statement

Young Anna Argyris: Conceptualization, Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization, Supervision, Project administration. Zuhui Wang: Methodology, Investigation, Software, Validation, Data curation, Writing - original draft. Yongsuk Kim: Methodology, Formal analysis, Writing - original draft. Zhaozheng Yin: Methodology, Supervision, Resources, Funding acquisition.

Declaration of competing interest

Authors have no conflict of interest to report.

Acknowledgement

This study was funded by the National Science Foundation (NSF CAREER grant IIS-1351049).

References (73)

  • S. Aral

    Commentary--Identifying social influence: A comment on opinion leadership and social contagion in new product diffusion

    Marketing Science

    (2011)
  • S. Aral et al.

    Tie strength, embeddedness, and social influence: A large-scale networked experiment

    Management Science

    (2014)
  • R.M. Baron et al.

    The moderator-mediator variable distinction in social psychological research: Conceptual, strategic and statistical considerations

    Journal of Personality and Social Psychology

    (1986)
  • N. Bhatt et al.

    Impact of celebrity endorser's source credibility on attitude towards advertisements and brands

    South Asian Journal of Management

    (2013)
  • S. Bishop

    Managing visibility on YouTube through algorithmic gossip

    New Media & Society

    (2019)
  • R. Bond et al.

    A 61-million-person experiment in social influence and political mobilization

    Nature

    (2012)
  • R.J. Brodie et al.

    Customer engagement: Conceptual domain, fundamental propositions, and implications for research

    Journal of Service Research

    (2011)
  • D. Byrne et al.

    Attraction and similarity of personality characteristics

    Journal of Personality and Social Psychology

    (1967)
  • J. Constine

    How Instagram's algorithm works

    (2018)
  • G. Contestabile

    Influencer marketing in 2018: Becoming an efficient marketplace

    (2018)
  • K. Cotter

    Playing the visibility game: How digital influencers and algorithms negotiate influence on Instagram

    New Media & Society

    (2019)
  • K. Cotter et al.

    Explaining the news feed algorithm: An analysis of the “News feed FYI” blog

  • M. De Veirman et al.

    Marketing through Instagram influencers: The impact of number of followers and product divergence on brand attitude

    International Journal of Advertising

    (2017)
  • U.K.H. Ecker et al.

    The components of working memory updating: An experimental decomposition and individual differences

    Journal of Experimental Psychology. Learning, Memory, and Cognition

    (2010)
  • J.S.B. Evans et al.

    Dual-process theories of higher cognition: Advancing the debate

    Perspectives on Psychological Science

    (2013)
  • R. Fry

    Millennials expected to outnumber boomers in 2019

    (2018)
  • D. Godes et al.

    The firm's management of social interactions

    Marketing Letters

    (2005)
  • M. Granovetter

    The strength of weak ties

    American Journal of Sociology

    (1973)
  • M. Granovetter

    The strength of weak ties: A network theory revisited

    Sociological Theory

    (1983)
  • B. Gu et al.

    The allure of homophily in social media: Evidence from investor responses on virtual communities

    Information Systems Research

    (2014)
  • J. Han et al.

    The influence of the sigmoid function parameters on the speed of backpropagation learning

  • R. Hannerman et al.

    Introduction to social network methods: Chapter 10: Centrality and power

    (2005)
  • K. He et al.

    Deep residual learning for image recognition—semantic scholar

    (2016)
  • T. Highfield et al.

    Instagrammatics and digital methods: Studying visual social media, from selfies and GIFs to memes and emoji

    Communication Research and Practice

    (2016)
  • C. Jensen et al.

    Finding others online: Reputation systems for social online spaces

    Proceedings of the SIGCHI Conference on Human Factors in Computing Systems

    (2002)
  • S.V. Jin et al.

    Product placement 2.0: “Do brands need influencers, or do influencers need brands?

    Journal of Brand Management

    (2019)
  • Cited by (95)

    View all citing articles on Scopus
    View full text