Changing Brand Attitudes through Influencer Marketing

Liang, Yuyang and Argyris, Young Anna and Muqaddam, Aziz (2018) Changing Brand Attitudes through Influencer Marketing. AMERICAS CONFERENCE ON INFORMATION SYSTEMS (AMCIS).

Changing Brand Attitudes through Influencer Marketing.pdf

Download (118kB) | Preview


Influencer marketing evolved from celebrities’ product endorsement, leveraging the popularity of “micro-celebrities” on social media. Influencers are these “micro-celebrities” who are not necessarily public figures but rather gathered sizable followers in social media and consistently endorse a set of brands or products from one category they claim expertise on. In this form of marketing campaigns, product endorsement becomes subtle. Brands would collaborate with influencers, to give them their products for free, in return for reviewing them in their social media accounts. However, the caveat for influencer marketing is the lack of brands’ control over these influencers since influencers use their personal social media account to communicate with their followers. As a result, it is important to assess the effectiveness of influencer marketing. To achieve this goal, we focus on the congruence between influencers’ posts and followers’ posts. Recently, Instagram has become the primary platform for influencer marketing for fashion brands for its affordances for exchanging photos and videos. On this platform, we choose to use the activewear fashion brand, Lululemon, which has proactively employed influencer marketing and obtained substantial success especially in terms of creating and maintaining a large group of loyal customers without spending on mass media advertising. Based on the source-congruence approach and electronic word of mouth literature (References needed here), we hypothesize that when the congruence between influencers’ and followers’ posts is higher, the followers will “like” the influencers’ posts; as a result, the followers’ attitudes towards the brand will improve. First, we collected a list of Lululemon brand ambassadors from the brand’s official website, and we randomly selected 30 profiles from the list for further analysis. For each influencer, we obtained 50 images posted on their Instagram account. Further, we selected 60 followers per influencer and a half of these 60 followers have never liked any of the influencers’ posts (called “non-likers”), and the other half have (“likers”). 50 images per follower, a total of 91,500 images, were downloaded for the analysis. Next, we manually code 300 images as training data according to two brand-related themes: physical activity (1=active; 0= non-active) and clothing style (1= active wear/athletic clothing; 0= other styles). Cohen’s Kappa showed acceptable inter-coder reliability: .88 (style) and .73 (activity). Then, we applied the Convolutional Neural Network (CNN) method to automatically classify the total of 91,500 images. The final machine learning model has an accuracy rate of 83.5%. To measure the similarity of the posts between the influencers and their followers, we used the cosine similarity score. First, we compared the similarity score between likers and non-likers via independent T-test. The test result indicates that, on average, likers (Mean = 0.88, SD = 0.25) have a significantly higher similarity score with the influencers they follow than non-likers (Mean =0.84, SD = 0.30), with t(1798) = 3.14, p = 0.0017). Thus, in general, likers’ posts are more congruent with influencers’. Moreover, within likers, we investigated the relationship between the number of times they liked the influencers’ posts and their similarity with the influencers, using a Poisson mixed model with the similarity score as the independent variable and a random intercept. The estimated coefficient for the similarity score is 0.20 (s.e. = 0.05, p < 0.001), which shows that the similarity score has a significant positive relationship with the number of likes. Thus, our hypotheses are supported.

Item Type: Article
Subjects: H Social Sciences > HB Economic Theory
Divisions: Faculty of Industrial Engineering and Informatics > Information System
Depositing User: staff repository 1
Date Deposited: 06 Sep 2018 13:30
Last Modified: 06 Sep 2018 13:30

Actions (login required)

View Item View Item