You are on page 1of 112

Research and Analysis of Overseas Influencer

Viral Video Marketing

An Interactive Qualifying Project


submitted to the Faculty of
WORCESTER POLYTECHNIC INSTITUTE
In partial fulfillment of the requirements for the
Degree of Bachelor of Science

By:
Kristen Heller, Vanessa Sam, Michael Sensat, Jessica Vo

Date:
January 2024
Reported Submitted to:

Alice Yang
WotoKOL Technology Co. Ltd.

Professors Joseph Sarkis and Hansong Pu


Worcester Polytechnic Institute

This report represents the work of one or more WPI undergraduate students submitted to the
faculty as evidence of a degree requirement. WPI routinely publishes these reports on the web
without editorial or peer review. For more information about the project programs at WPI, see
https://www.wpi.edu/project-based-le
Abstract
With social media becoming a prominent part of everyday life, companies are opting to
use social media platforms as a means to advertise their products. In sponsorship with Hangzhou
WotoKOL Network Technology Co. Ltd, our team’s goal was to determine which key video
features relate to the virality of videos created for the advertisement of products, posted across
YouTube, Instagram, and TikTok. The results from our research provided evidence that the use
of humor and emotional appeal during the paid promotion sections of videos had a greater effect
on the overall virality of a video in comparison to other features. However, due to various
limitations of our research discussed in the paper, our findings are best served as a pilot study,
with the suggestion of further research to be conducted.

i
Acknowledgements
The team would like to recognize our sponsor contact Alice Yang and her associates at
WotoKOL for their tremendous support throughout the entirety of this IQP experience. The
completion of this IQP would not have been possible without the resources and guidance from
Alice Yang and WotoKOL. The team greatly thanks both parties for supporting and advising our
team whenever complications arose throughout our project completion.

We would also like to thank Professors Hansong Pu and Joseph Sarkis for their guidance.
As our project advisors, their feedback, support, and insight have proven vital to the project and
helped our team remain on track as we faced obstacles.

Finally, we would like to also thank the Hangzhou Dianzi University (HDU) team and
Professor Cao for their cooperation in this IQP project’s completion and for conducting the data
analysis for the project’s social media user survey.

ii
Executive Summary
Background
As of 2023, over 4.9 billion users can be found on social media across the world,
showcasing the vast extent of these platforms. With an extensive and varying user base found
across different social media platforms, many companies are now opting to utilize social media
for marketing purposes. Through collaboration between influencers on different social media
platforms, companies now have a new means to reach out to larger audience bases and gain more
views to promote their products. To reach the largest audience base and gain the most exposure,
many of these brands and influencers hope to go viral.
For this Interactive Qualifying Project (IQP), our team investigated how different
characteristics of promotional videos of two different brands, posted on YouTube, TikTok, or
Instagram affect video viewership and virality, to determine which of these characteristics
impact video virality most positively. Through an in-depth case study analysis of promotion
videos for two major clients of our sponsor and a social media user behavior survey, our team
will isolate which characteristics are most influential to virality. Following both qualitative and
quantitative analyses of our results, our work will ultimately provide our sponsor with a map of
how to best utilize these characteristics to promote video virality, ultimately creating the
framework for a standard operating procedure (SOP).
To accomplish our project goal, our team answered the following questions to help
complete our final deliverables: What content features relate to social media virality the most?
What content messages most relate to social media virality? What creator features most relate to
social media virality? What creator history characteristics affect social media virality the most?

Approach
This IQP project was completed in a two-part methodology sequence to accomplish the
main goal of understanding which video characteristics affect the performance and virality of a
video featuring a promotional advertisement. To meet this goal, the project’s primary objective
and methodology was to conduct a case study analysis on videos featuring a promotional product
advertisement for the products of two companies, Company A and Company B, both currently
partnered with our sponsor, Hangzhou WotoKOL Network Technology Co. Ltd. Our secondary
methodology focused on social media user behavior, where a survey consisting of both multiple

iii
choice and Likert scale questions was conducted with a survey pool located outside of China to
further determine which video characteristics are most impactful to video virality, but also to
further confirm the findings of the case study analysis.
In the case study analysis, our team first gathered over 70 total videos containing paid
promotional advertisements for one product of Company A and one product of Company B, with
at least 35 videos for each product. Due to a confidentiality agreement with our sponsor, we are
unable to disclose the names of Company A and Company B; however, we can disclose general
details about both companies. Company A is an electronic device brand focusing on Augmented
Reality technology, and Company B is a household appliance and cleaning device brand.

After video collection, videos and their web links were placed into an Excel spreadsheet.
In this Excel spreadsheet, the field of factors used for video analysis was created using a
framework consisting of major video characteristic categories: Content Features, Content
Message, Creator Features, and Creator History, and the characteristics that fall under each
category. These analysis fields were then broken up into three different response types: Yes or
No responses, Likert-type scale responses, and direct content feature responses. These fields in
the spreadsheet were manually determined by each member of the team.

The social media user behavior survey was created using Microsoft Forms and has three
main sections: 1. personal demographic questions, 2. social media demographic questions, and 3.
social media behavior questions. In the first section of the survey, respondents were asked
personal demographic questions regarding age, race, ethnicity, gender identity, sexual
orientation, family income, and highest degree or level of education. Section 2: social media
demographic questions, the respondents were presented with two questions. The first question
asked the user to rank how often they used the social media platforms of YouTube, Instagram,
and TikTok using a five-point Likert-type scale. This Likert-type scale was defined as the
following: 1- Never, 2- Rare, 3- Sometimes, 4- Often, and 5-Always. The second question in
section 2 asked the user how much time they spend on social media daily, with the options
spanning from less than one hour a day to over ten hours a day. The final section of our survey,
social media user behavior, was broken up into three categories of questions, all of which used a
five-point Likert scale rating. For this category of questions, the five-point Likert-type scale was
defined the same as in Section 2: social media demographic questions. The survey was

iv
distributed to several WPI student organizations, including the WPI chapter of the Society of
Asian Scientists and Engineers, Chi Omega Sorority, and the ESports club - Valorant Team.

Analysis
Following the case study video analysis, the IQP team performed both a qualitative and
quantitative analysis of the observed fields. Quantitative analysis for the case study was
conducted using Excel and Statistical Package for the Social Sciences (SPSS) software. In Excel,
we drew summary statistics and then made one-to-one comparisons between the different fields
used in the video analysis. Excel was also used to prepare and clean the data for SPSS analysis.
Within SPSS, the cleaned data was then run through linear regression models with varying target
variables.
For our team’s social media user behavior survey, data analysis consisted of analyzing
the distributions of the answers to the questions to determine which survey results were
statistically significant. This accompanied some more basic distribution and frequency analysis
using SPSS.

Final Recommendations
Based on the results the secondary research, case study analysis and a social media user
behavior survey, several recommendations are presented to aid businesses and influencers in
seeking to improve performance of their promotional advertisement videos and video virality.
However, due to limited sample size for both the case study analysis and social media user
survey, this study is best used as a pilot and that future research is needed to further support these
conclusions.

1. We recommend that influencers use humor during the paid promotional


advertisement parts of their videos. As observed in Figure 9 and Figure 10 in Section
4.2, the relationship between the average engaged viewers and whether the video
contained video characteristics, shows that videos with a higher engaged viewer count
user humor within the paid promotional advertisement sections of their videos.

v
2. We recommend that influencers use emotional appeal during the paid promotional
advertisement parts of their videos. As seen in Figure 15 in Section 4.2.4, using
emotional appeal during promotional aspects of a video increases the overall performance
of the video. This is also supported in Table 4 from our linear regression model
predicting engaged viewers on a particular video, the model has a 92.8% adjusted R2
score, and emotional appeal is the number one predictor for engaged viewers. This model
also found that the number of engaged viewers is higher when the emotional appeal is 4,
5, or 6, and lower when the emotional appeal is rated 1, 2, or 3.

3. We recommend that further research be conducted on the following content


message characteristics: production quality, music appropriateness, video creativity,
influencer excitement, emotional appeal, trustworthiness, and authenticity. Data
analysis on these content message characteristics discussed on Section 4 suggest that
these video features are significant (p<0.1), however due to the lacking sample size of
videos in the case study analyses, further testing should be conducted.

4. We recommend that further research be conducted on hashtags. Table 5 in Section


4.1.3 suggests that hashtags related to general product categories have the most effect on
the viewership of a social media post. However, due to the skew in data with the lacking
sample size of videos for the case study analyses, the team suggests additional research
be conducted to support these findings.

The team also provided the sponsor with two practical deliverables to be used by the
sponsors. One is a mind map that specifies the various categories of features and their linkage.
This mind map helps the sponsor to replicate or identify features for various industries beyond
the case study product and industry categories. The second deliverable is an aesthetically
pleasing, professional, user-friendly and high-level summary of the results that the sponsor can
have to include in customer marketing efforts.

vi
Authorship
Section Author(s)

Abstract Jessica Vo

Acknowledgments Jessica Vo

Executive Summary Jessica Vo

1. Introduction Jessica Vo

2. Background Michael Sensat


Revisions: Jessica Vo

2.1 What is WotoKOL Michael Sensat

2.2 Social Media Marketing Vanessa Sam

2.2.1 Influencer Marketing Vanessa Sam

2.3 Social Media Platforms Kristen Heller


Revisions: Jessica Vo

2.3.1 YouTube General Background Kristen Heller


Revisions: Jessica Vo

2.3.2 Instagram General Background Kristen Heller


Revisions: Jessica Vo

2.3.3 TikTok General Background Kristen Heller


Revisions: Jessica Vo

2.4 Viral Videos Jessica Vo

2.4.1 Content Feature Jessica Vo

vii
2.4.1.1 Numerical Video Metrics Jessica Vo

2.4.1.2 Descriptive Video Elements Jessica Vo

2.4.1.3 Video Properties Jessica Vo

2.4.2 Content Message Vanessa Sam


Revisions: Jessica Vo

2.4.2.1 Capturing Audience Attention Vanessa Sam


Revisions: Jessica Vo

2.4.2.2 Influencer’s Impact on Virality Vanessa Sam

2.4.2.3 Rational Appeal v. Emotional Appeal Vanessa Sam


Revisions: Jessica Vo

2.4.2.4 Influence of Video Background Vanessa Sam

2.4.3 Creator Feature Kristen Heller, Jessica Vo

2.4.4 Creator History Michael Sensat

2.4.4.1 Role of Creator History on Consumers Michael Sensat

2.5 Project Goal Michael Sensat


Revisions: Jessica Vo

3. Methods Jessica Vo

3.1 Data Collection Michael Sensat

3.1.1 Case Study Methods Jessica Vo

3.1.1.1 Video Collection Jessica Vo

3.1.1.2 Factor Determination and Framework Development Michael Sensat

viii
Revisions: Jessica Vo

3.1.2 Social Media User Surveys Jessica Vo

3.1.2.1 Identifying Survey Subjects Jessica Vo

3.1.2.2 Survey Structure Jessica Vo

3.1.2.3 Survey Distribution Jessica Vo

3.2 Data Analysis Plan Vanessa Sam

3.2.1 Viewer Survey Quantitative Analyses Vanessa Sam

3.2.2 Case Study Analyses Kristen Heller

3.2.2.1 Case Study Qualitative Analyses Jessica Vo

3.2.2.2 Case Study Quantitative Analyses Vanessa Sam, Kristen Heller,


Michael Sensat

4. Results and Findings Jessica Vo, Kristen Heller

4.1 Video Content Features and Virality Kristen Heller, Michael


Sensat, Vanessa Sam
Revisions: Jessica Vo

4.1.1 Company A Vanessa Sam, Kristen Heller


Revisions: Michael Sensat,
Jessica Vo

4.1.2 Company B Kristen Heller, Vanessa Sam


Revisions: Michael Sensat,
Jessica Vo

4.1.3 Combined Case Studies Kristen Heller

ix
Revisions: Michael Sensat,
Jessica Vo

4.2 Video Content Message and Virality Michael Sensat

4.2.1 Music Appropriateness Vanessa Sam


Revisions: Michael Sensat

4.2.2 Video Creativity Vanessa Sam


Revisions: Michael Sensat

4.2.3 Influencer Excitement Vanessa Sam


Revisions: Michael Sensat

4.2.4 Emotional Appeal Vanessa Sam


Revisions: Michael Sensat

4.3 Video Creator Features and Virality Vanessa Sam

4.4 Video Creator History and Virality Michael Sensat

4.4.1 Trustworthiness Vanessa Sam


Revisions: Michael Sensat

4.4.2 Authenticity Vanessa Sam


Revisions: Michael Sensat

5. Findings and Recommendations Jessica Vo, Vanessa Sam

5.1 Deliverables Jessica Vo, Vanessa Sam


Revisions: Michael Sensat

5.2 Recommendations Jessica Vo


Revisions: Michael Sensat

5.3 Limitations Vanessa Sam, Jessica Vo

x
Revisions: Kristen Heller

5.4 Concluding Remarks Jessica Vo

References

Appendices

xi
Table of Contents
Abstract ..................................................................................................................................... i
Acknowledgements................................................................................................................... ii
Executive Summary ................................................................................................................ iii
Authorship .............................................................................................................................. vii
List of Figures.......................................................................................................................... xv
List of Tables ..........................................................................................................................xvi
1. Introduction ...........................................................................................................................1
2. Background ...........................................................................................................................3
2.1 Who is WotoKOL ............................................................................................................3
2.2 Social Media Marketing ..................................................................................................4
2.2.1 Influencer Marketing ..................................................................................................5
2.3 Social Media Platforms ...................................................................................................6
2.3.1 YouTube General Background ....................................................................................6
2.3.2 Instagram General Background ...................................................................................7
2.3.3 TikTok General Background .......................................................................................8
2.4 Viral Videos .....................................................................................................................9
2.4.1 Content Feature ........................................................................................................ 10
2.4.1.1 Numerical Video Metrics ................................................................................... 10
2.4.1.2 Descriptive Video Elements ............................................................................... 10
2.4.1.3 Video Properties ................................................................................................ 12
2.4.2 Content Message....................................................................................................... 13
2.4.2.1 Capturing Audience Attention ............................................................................ 13
2.4.2.2 Influencer’s Impact on Virality .......................................................................... 14
2.4.2.3 Rational Appeal v. Emotional Appeal ................................................................ 14
2.4.2.4 Influence of Video Background.......................................................................... 16
2.4.3 Creator Feature ......................................................................................................... 16
2.4.4 Creator History ......................................................................................................... 18
2.4.4.1 Role of Creator History on Consumers ............................................................... 18
2.5 Project Goal ................................................................................................................... 20
3. Methods ............................................................................................................................... 21

xii
3.1 Data Collection .............................................................................................................. 21
3.1.1 Case Study Methods ................................................................................................. 22
3.1.1.1 Video Collection ................................................................................................ 22
3.1.1.2 Factor Determination and Framework Development........................................... 23
3.1.2 Social Media User Surveys ....................................................................................... 27
3.1.2.1 Identifying Survey Subjects ............................................................................... 27
3.1.2.2 Survey Structure ................................................................................................ 28
3.1.2.3 Survey Distribution ............................................................................................ 29
3.2 Data Analysis Plan......................................................................................................... 29
3.2.1 Viewer Survey Quantitative Analyses ....................................................................... 30
3.2.2 Case Study Analyses ................................................................................................. 30
3.2.2.1 Case Study Qualitative Analyses ........................................................................ 31
3.2.2.2 Case Study Quantitative Analyses ...................................................................... 31
4. Results and Findings ........................................................................................................... 36
4.1 Video Content Features and Virality ............................................................................ 36
4.1.1 Case Company A ...................................................................................................... 36
4.1.2 Case Company B ...................................................................................................... 39
4.1.3 Combined Case Studies ............................................................................................ 40
4.2 Video Content Message and Virality ............................................................................ 42
4.2.1 Music Appropriateness ............................................................................................. 44
4.2.2 Video Creativity ....................................................................................................... 47
4.2.3 Influencer Excitement ............................................................................................... 47
4.2.4 Emotional Appeal ..................................................................................................... 48
4.3 Video Creator Features and Virality ............................................................................ 49
4.4 Video Creator History and Virality .............................................................................. 50
4.4.1 Trustworthiness ........................................................................................................ 50
4.4.2 Authenticity .............................................................................................................. 51
5. Conclusions & Recommendations ...................................................................................... 53
5.1 Deliverables.................................................................................................................... 53
5.2 Recommendations.......................................................................................................... 56
5.3 Limitations ..................................................................................................................... 57

xiii
5.4 Concluding Remarks ..................................................................................................... 60
6. References ............................................................................................................................ 61
Appendices .............................................................................................................................. 67

xiv
List of Figures
Figure 1: WotoKOL Business Relationship Diagram
Figure 2: Video Characteristic Original Framework, Han et al.
Figure 3: Adjusted Video Characteristic Framework Created by our team
Figure 4: TikTok Trending Hashtag Page
Figure 5: Ad Characteristics Impact Analysis
Figure 6: Hypothesis model, Chekima et al.
Figure 7: Hypothesis results, Masuda et al.
Figure 8. Company A Production Quality to Viewers relationship graph.
Figure 9: Company A Engaged Viewers to Yes/No Question Relationship Graph
Figure 10. Company B Engaged Viewers to Yes/No Question Relationship Graph
Figure 11. Company A Music Appropriateness to Video Engagement Rate Relationship Graph.
Figure 12. Company B Music Appropriateness to Viewers relationship graph.
Figure 13. Company A Video Creativity to Engaged Viewers relationship graph.
Figure 14. Company B Influencer Excitement to Engaged Viewers relationship graph.
Figure 15. Company B Emotional Appeal to Engaged Viewers relationship graph.
Figure 16. Company A Trustworthiness to Video Engagement Rate relationship graph.
Figure 17. Company A Authenticity to Video Engagement Rate relationship graph.
Figure 18. Final Presentation Slide Example
Figure 19. Humor in Videos Final Presentation Slide
Figure 20. Emotional Appeal in Videos Final Presentation Slide
Figure 21. Overall Mind Map of Which Video Characteristics Affect Virality Depending on
Industry and Product Category

xv
List of Tables
Table 1. Company A Numerical Video Metrics Relationships with P-value
Table 2. Summary of Linear Regression Results for Company A
Table 3. Company B Numerical Video Metrics Relationships with P-value
Table 4. Summary of Linear Regression Results for Company B
Table 5: Summary of Linear Regression Results for the Combined Case Study Data Set
Table 6: Dependent Variable to Dependent Variable Correlation Coefficients with p-values.
Table 7: Descriptive Statistics of Company A Characteristics
Table 8: Descriptive Statistics of Company B Characteristics
Table 9. Numerical Calculations of Company A Characteristics
Table 10. Numerical Calculations of Company B Music Appropriateness
Table 11. Gender Distribution for Company A Influencers
Table 12. Gender Distribution for Company B Influencers
Table 13. Top Audience Age Range Distribution for Company A Influencers
Table 14. Gender Distribution for Company B Influencers

xvi
1. Introduction
Have you ever wondered why MrBeast gets so many views on his YouTube videos? How
Bella Poarch receives enormous numbers of shares on her TikTok trends? How Strictly
Dumpling gains tremendous numbers of likes on his food posts?
As technology continues to advance and becomes more integrated into all aspects of life,
the use of social media has become a part of daily life. As of 2023, around 4.9 billion use social
media around the world (Wong, 2023). With such a wide and varying user base, many
companies are now opting to utilize social media for marketing purposes. Through collaboration
between influencers on different social media platforms, companies now have a new means to
reach out to larger audience bases and gain more views. In order to reach the largest audience
base and gain the most exposure, many of these brands and influencers hope to go viral. Thus,
our sponsor, Hangzhou WotoKOL Network Technology Co. Ltd. (WotoKOL) is looking to gain
innovative ideas to further support influencer-brand collaborations in increasing the virality of
their ads.
WotoKOL is a leading domestic overseas influencer marketing company that specializes
in both influencer-brand collaborations and software as a service (SAAS) technology to support
both brands and influencers in reaching virality through their social media video marketing posts
(WotoKOL, 2023). Despite having success over the past 7 years with serving over 2,500 brands
and influencer connections spanning 117 countries, the fast-moving world of social media
content demands for fresh and creative ideas to continue to market these brands (WotoKOL,
2023). With these marketing video posts focusing on three social media platforms: YouTube,
Instagram, and TikTok, WotoKOL is looking to gain insight into the characteristics aiding in a
video’s virality success. This includes analyzing previously released advertisement videos and
determining which key video factors affect video virality the most.
For this Interactive Qualifying Project (IQP), we investigated how different
characteristics of promotional videos of two different brands, posted on YouTube, TikTok, or
Instagram affect video viewership and virality, with the goal of determining which of these
characteristics impact video virality most positively. Through an in-depth case study analysis of
promotion videos for two major clients of our sponsor and a social media user behavior survey,
our team will isolate which characteristics are most influential to virality. Following both
1
qualitative and quantitative analyses of our results, our work will ultimately provide our sponsor
with a map of how to best utilize these characteristics to promote video virality, ultimately
creating the framework for a standard operating procedure (SOP).
The remainder of this report will introduce background literature—secondary research
resources—that has influenced the direction of our group’s planning, execution, and final
analysis of this study. From there, we then present our methodology section explaining our data
collection and analysis methods. Following that, we discuss the findings of our primary and
secondary research, identifying major highlights and important insights. To conclude our report,
we provide our sponsor recommendations, project challenges, and closing remarks.

2
2. Background
The following section provides background on our project sponsor, Hangzhou
WotoKOL Network Technology Co. Ltd., as well as discusses the general background
understanding of the social media influencer marketing industry. We then overview social media
marketing, specifically influencer marketing on social media platforms and will briefly introduce
the various social media platform histories evaluated in the case studies. This section will
conclude with an in-depth discussion on the various video characteristics and creator
characteristic factors that play a role in video and social media posting outcomes on these social
media platforms.

2.1 Who is WotoKOL


Our project sponsor WotoKOL is an overseas social media influencer marketing
company located in Hangzhou, China. WotoKOL was established on November 27, 2017, with
the aim of creating relationships between consumer brands and international influencers.
(WotoKOL, 2023).

WotoKOL’s mission is to provide an efficient method for brands to contact global


influencers. They also provide video ideas, video topics, and influencer recommendations to
partnered brands to produce a positive result for the brand (WotoKOL, 2023).

WotoKOL is the developer of a SAAS (software as a service) technology known as


‘WotoHub’, which is an application that allows thousands of unique brands to easily connect
with over 11 million TikTok, Instagram, and YouTube influencers from around the world. The
WotoHub is a database that brands can use to find the analytics of influencers and their videos in
order to find the creator that would be most qualified to work with their brand. Currently,
according to our sponsor, the WotoHub has 11,728,645 influencers from 117 different countries,
2,800 brands each with more than two videos, over 5000 videos uploaded, and 38,000 active
companies.

As previously mentioned, our sponsor mainly focuses on social media influencers from
three main social media platforms: TikTok, Instagram, and YouTube. They recruit these

3
influencers from around the world as contractors who can be requested to upload social media
content on their accounts to advertise WotoKOL’s brand customers as shown in Figure 1.

As a company based in Hangzhou, China, it is difficult for our sponsor to gather high-
quality video ideas and topics that are relevant to overseas influencers. In order to stand out from
their competitors, WotoKOL’s current goal is to increase the number of viewers and consumers
that are acquired through their influencers’ advertisement videos. They plan to accomplish this
goal by deepening their understanding of which video characteristics increase video viewership
retention and virality the most. As the sponsor is seeking young, international creative minds,
having the IQP team from WPI (United States) to assist in the video analysis process helps the
sponsor with their goal of determining what video factors and characteristics perform well on
social media platforms overseas.

Figure 1. WotoKOL Business Relationship Diagram

2.2 Social Media Marketing


As mentioned, social media is becoming more prevalent in everyone’s lives every day.
Brands utilize social media marketing to gain more attention and sales. They can interact and
build relationships with consumers with more ease, influence their behaviors, and shape the

4
companies’ reputation. Through these interactions, they can receive first-hand feedback and
understand consumer opinions (Sohaib et al., 2022).

2.2.1 Influencer Marketing


As of 2022, the influencer marketing industry has risen to be a $16.4 billion industry,
with more brands having been taking advantage of virtual influencers to expand the outreach of
the company’s networking (Celestino, 2023). Influencer marketing is one of the fastest-growing
marketing strategies in which influencers create content on social media to promote a brand’s
products to their audience and potentially reach the brand’s target consumers. In return, brands
compensate the influencers typically with pay and free products (Tanwar et al., 2022).
To have a better understanding as to what influencer marketing is, the term, “influencer,”
should be defined. Compared to a celebrity influencer who gained attraction from their work
through traditional media, an “influencer” is an individual in an industry that is able to sway
target audiences (Omar, 2023). Influencers on social media platforms create and post content for
their followers. They often have general expertise in a specific field and have a sizable following
as they are often seen as authentic (Haenlein et al., 2020).
Due to the influencers’ significant following and expertise in the field, many brands are
looking to them to promote their products. They oftentimes can appeal to the audience and gain
more positive attention than the marketing team of a brand as they are familiar with the platform
and audience. There are three main purposes that influencers have: reach to their audience,
endorsement, and management of social media.
To bring more attention to posts on social media, brands want to maximize and maintain
engagement. Influencers can do so by connecting with consumers in a more personal manner,
which is difficult for social media teams of brands. They can cultivate posts that audiences are
more attracted to, displaying understanding and appreciation for the followers. As the influencer
gradually introduces and keeps their followers informed, they are able to achieve serial targeting,
meaning the brands’ narratives are broken up into numerous and simpler posts. The audience is
able to digest the purpose and continuously be exposed to the same brand. Brands may also
target influencers with followers who are not directly in their target audience as they can get
exposure to a broader range.

5
2.3 Social Media Platforms
Social media platforms have only been around for about a quarter century, but they are
already an integral part of most internet users’ lives. These types of websites began with profiles
and connections, but social networking sites gradually grew into blogging, microblogging,
pictures, and video sharing. In this IQP project, the sole focus of research and analysis will be
focused on the video-sharing aspect of social media platforms per WotoKOL’s request.

2.3.1 YouTube General Background


As of 2023, YouTube estimates that 500 hours of new content is uploaded every minute
and that almost 5 billion video views occur each day and 70 billion video views on YouTube
Shorts (Reid, 2023).
As an international platform, YouTube has the highest monthly traffic in the US alone,
with 11.7 billion visits. Other countries have significant usage of the platform. South Korea in
second gets 8.25 billion per month, India gets 4.20 billion, and Brazil gets 3.60 billion to name a
few (Reid, 2023).
YouTube’s content consists of videos. Their main platform supports videos of varying
lengths which can then be organized into playlists by viewers or viewed alone. Posted videos can
vary in length, falling anywhere under 15 minutes to upwards of a maximum of 12 hours.
Creators, or influencers, post their videos on Channels, of which they can have multiple. Viewers
are then able to subscribe to specific channels and like or comment on specific videos when they
view them. Influencers on YouTube can make money by running external ads on their videos
through YouTube or by making deals with brands for advertising and exposure within the video
itself. They also often make money through affiliate marketing in which the creator gets a
commission from every sale they generate. Since viewers subscribe to specific creators on this
platform and their long-form videos are still popular, WotoKOL hypothesizes that influencers
are better able to market on YouTube because they tend to have a more loyal fan base. It is also
easier to market on YouTube because links can be embedded directly in the video descriptions
(Sherman, 2021).
YouTube is not transparent about its algorithm, but external organizations have found
some key features of it through independent research. The algorithm weighs users watch history
most heavily while for the most part disregarding user inputs such as likes and dislikes (Kiros,

6
2022). Other main factors affecting what YouTube recommends to the user are the relevance of
the video, its proportional past performance, the viewer’s interpolated personal preference, and
how YouTube interprets the creator’s authority or trustworthiness (Glover, 2023).

2.3.2 Instagram General Background


As of 2023, Instagram has about 2 billion monthly active users, falling behind YouTube
with 2.5 billion monthly active users. Instagram is also considered the world’s second favorite
social media platform with 14.3% of those surveyed calling it their favorite. Instagram’s user
demographics are mainly made up of people below the age of 35, as well as more males than
females on the platform (Moss, 2023).
Despite other social media platforms being ahead in active users, Instagram has several
advantages for brands and advertising. Instagram is the number one social media platform used
for brand research. 60.9% of Instagram users use Instagram to research brands. Also, 90% of
Instagram users follow at least one brand. One survey also showed that 25% of social media
marketers said Instagram drives the most sales of any platform; this statistic is second only to
Facebook which was selected by 34% of the social media marketers surveyed (Moss, 2023).
Instagram has 5 main types of content it supports on the platform: feed posts, reels, stories,
Instagram videos, and live broadcasts.
Feed posts are posts that are shown directly on a user’s profile page that can include
images and videos that are up to a minute long. Reels refers to Instagram’s short form vertical
video service that they also launched to compete with TikTok. These reels are much more likely
to be discovered by users not already following the account, even if posted by a smaller creator.
Stories are temporary content that can be photos or short videos only visible for 24 hours after
posting. They can be pinned to a profile in what is referred to as “Story Highlights”. Instagram
videos serve a similar function to YouTube. Instagram videos are up to 60 minutes long and are
posted on the profile separately from the feed. The last feature is live broadcasts. This is unedited
video content streamed in real time. It can however be saved and posted as a video later (Park,
2023).
Instagram has different algorithms for different sections of the app, but across the main
pages where a user can expect to find video content (main feed, reels, and explore sections) there
are four main features considered, just in different orders. The four features are your past

7
activity, information about the post itself, information about the creator who posted, and the
user’s history of interacting with a given creator (Mosseri, 2023).

2.3.3 TikTok General Background


TikTok was launched in China in September 2016 under the name Douyin. Within a year
it had reached 100 million users. TikTok, the international version of the app, was launched a
year later in September 2017 after Douyin’s parent Company Bytedance acquired Musical.ly.
TikTok exploded after that with their global monthly users increasing by 1157.76% between
January 2018 and July 2020. TikTok is 66th in global user engagement, but considering it only
launched internationally 6 years ago while larger competitors have had over twice as long to
accumulate their user bases, TikTok has had impressive growth (Dean, 2023).
As of 2023, 37.3% of American mobile internet users access TikTok at least once a
month. Globally, TikTok has 1 billion monthly active users, so of the 4.48 billion active social
media users, 22.32% use TikTok regularly. The user base of TikTok has a 2 to 1 female-to-male
gender ratio, and 47.4% of their active users are aged between 10 and 29 years old. It also has the
second highest average minutes per day in the US with 33 minutes per day. This is second only
to Facebook which gets an average of 35 minutes per day (Dean, 2023).
Similar to the Instagram Discover page, TikTok has the “For You” page which contains
content that is curated for the user; however, this page consists of one post at a time. This is the
initial landing page of the app. The content that TikTok has are mostly videos with varying
length: 15 seconds, 60 seconds, and 10 minutes. Other features available on the app are live
broadcasting, and photos and text which can be joined into a “carousel”. All posts are
accompanied by sounds, either one made by the user or one by others. Users can explore content
posted by others with the same sound, allowing users to find similar content. A unique highlight
of TikTok is that users can duet videos created by other users, promoting engagement within the
community (Hill, 2022). This splits the screen of the video into two parts: the original video and
the reaction. When the post is shown on the “For You” page, only the first two lines of the
caption are visible until the “more” button is clicked.
TikTok’s algorithm is based on matching users together based on their interests and
cross-recommending videos depending on what they like. They also rely on machine learning
algorithms to determine user preferences rather than polling users directly. So along with general

8
video metrics like likes, unique views, and shares, they use other information like how fast
someone is swiping or whether they pause a video. The algorithm then uses pattern recognition
on this data to determine what a given user will like or dislike depending on the data of other
users. The algorithm also specifically gives unknown users an opportunity to be discovered by
putting their videos in front of other users even when they are unknown (Chew, 2023).

2.4 Viral Videos


With many brands hoping to go “viral” in today’s fast-paced social media landscape, we
must first understand the measure of success and virality of a video, as well as define what it
means for a social media post to go viral. For the sake of our study, the term “virality” is defined
as “a social media post reaching a usually high number of likes, views, and shares” compared to
an account’s normal engagement metrics (Han et al., 2020). With this definition in mind, we
looked at the characteristics of a video and how they relate to video virality. Due to the multitude
of video characteristics that can be used to affect a video’s virality, involving both the video
content and message itself, as well as influencers' attributes, our team organized these different
characteristics into four categories based on Han et al. framework (see Figure 2): 1. Content
Features, 2. Content Message, 3. Creator Features, 4. Creator History (Han et al., 2020). These
four categories are divided between the content and creator, as the collaboration between the two
results in what we see in a final social media post (see Figure 3).

Figure 2. Video Characteristic Original Framework, (Source: Han, et al., 2020)

9
Figure 3. Adjusted Video Characteristic Framework (Source: IQP Team)

2.4.1 Content Feature


Content features describe the objective characteristics of a post that are readily
recognized by a viewer instantly (Han et al., 2020). This includes the numerical video metrics
commonly measured across all platforms, descriptive video elements, and descriptive video
elements. The following section breaks down the type of video features found in each category
and what the literature suggests about each feature.

2.4.1.1 Numerical Video Metrics


On YouTube, Instagram, and TikTok, all three social media platforms total the number of
likes, comments, and shares found directly on the video posting; however, only TikTok displays
the number of shares of a video. As key indicators of video performance measurement, these
three content features are commonly used to determine whether or not a video has reached
virality (West, 2021). While numerical video characteristics do not directly impact video virality
like the other video characteristics, these content features can be used as a comparison of the
performance of other video characteristic features.

2.4.1.2 Descriptive Video Elements


Similar to how a video’s numerical video metrics can be immediately discerned by

10
viewers once clicking on the video, the descriptive elements of a video are one of the first textual
features a viewer sees when looking at a video. On platforms like YouTube, the title of the video
can be found directly under the video thumbnail (YouTube, 2023). Upon clicking on the video,
viewers are then able to see the caption of the video and the hashtags tagged on it (YouTube,
2023). On TikTok and Instagram, with no distinct title feature on either platform, hashtags, or
captions play a much larger role as a descriptive video element.
As the first descriptive textual content feature the viewer observes, titles can play a vital
role in initial viewer engagement. By using eye-catching and creative titles, viewers are more
likely to initially click on a video (Isabella et al. 2023). Previous research on video engagement
has shown that when content that aligns with a viewer's interest is more likely to receive higher
engagement than content that does not align with their interest (Chowdhry, 2018). With this in
mind, influencers can curate the titles of their advertisement videos to best fit and draw in their
audience base.
Captions, similar to titles, curated to the audience base’s interest are more likely to
receive higher engagement (Isabella et al. 2023). However, in comparison, captions are much
more flexible and allow for longer descriptions and text. On YouTube, TikTok, and Instagram,
the caption section of the post is where influencers include additional details about the product.
However, on YouTube specifically, influencers are capable of directly inserting hyperlinks of the
products they use in the video (YouTube, 2023). This allows for an easy way for viewers to find
products showcased in the video.
Out of the three video descriptive video elements, hashtags can be more arbitrary and
versatile for influencers to use to increase the exposure of their videos. Made up of numbers,
letters, and symbols with a “#” at the beginning, hashtags are used by viewers as a means of self-
expression and conveyance of user preference (Wei et al., 2019). Hashtags can be used by
multiple users, allowing for social media posts to be grouped and found together when a hashtag
is searched on the social media platform. This allows the users to find social media posts of
similar content but also allows content creators to reach a wider audience of viewers (Lead
Generation, 2022).
Depending on the social media platform, different hashtags are currently trending. For
example, TikTok has a dedicated page on its platform highlighting current trending hashtags and
the top videos under each category. For brands and influencers, this social media post grouping

11
that occurs when a hashtag is used can help expand a brand’s audience and social media
presence. An example of TikTok’s trending hashtag page can be seen in Figure 4 below.

Figure 4. TikTok Trending Hashtag Page

2.4.1.3 Video Properties


While the video properties such as video quality and language are not immediately
displayed on a video, their details can be easily found under the video’s settings. For YouTube
specifically, the video quality and language can be found in the bottom left corner of the video
player (YouTube, 2023).

With over 91% of businesses using videos as a marketing tool and 96% of those
businesses valuing video content as an ‘important part’ of their marketing strategies, quality
videos become necessary to stand out from competitors (Wyzowl, 2023). When the quality of the
video is poor, studies have shown that video viewership falls by 77% (Tribbey, 2016). We must
define and understand what aspects of a video play a role in what constitutes ‘good video
quality.’

When recording a video, one of the most important technical aspects to examine is the
video resolution. Video resolution is the measure of the number of pixels in the standard aspect
ratio 16:9 (Leonard & Kurniawan, 2023). A video with a high resolution means that there is a

12
higher number of pixels, typically greater than 480p, present during any single frame. Videos
with lower pixel counts, less than 480p, indicate that the video has a lower resolution.
Commonly used resolutions across all video platforms today are 720p and 1080p, meaning that
in any given frame of a video, many pixels are present providing clearer images for the video
(Leonard & Kurniawan, 2023). With high-quality video production, brands and influencers then
have a technical means to increase their viewer retention rate.

Aside from video resolution, the use of camera work can help elevate the quality of the
video (Doman, 2015). Proper framing of the scene allows for the influencers to highlight the
subject they would like to focus on during the video. Providing different angles for such scenes
provides viewers with a more well-rounded perception of the product, especially in the case of
product advertisement (Doman, 2015).

2.4.2 Content Message


The content message characteristics of a video describe the features that viewers have to
read or watch to comprehend; this includes affective cues, facts, humor, and music. In addition to
the requirement of interpreting the video, these characteristics typically are subjective in terms of
enjoyment and interest; as a result, they are more difficult to define, analyze, and advise upon.
However, they do make an impact on a video’s virality. In this section, we will be going over the
effects of different content message features.

2.4.2.1 Capturing Audience Attention


One thing to consider when posting a video on social media is including a hook or an
interesting aspect to the video to catch the audience’s attention as the audience tends to increase
the performance of the video (Ling et al. 2022). Once the audience’s attention is captured with an
eye-catching hook, many viewers continue to watch the video due to the content. The creativity
of a video tends to keep the viewers watching as it provokes positive attitudes. Viewers feel
more encouraged to process the ad and purchase the featured item when viewing creative
advertisements (Lee, 2016). Not only does it improve the users’ attention, but it also causes the
video to be more memorable as well as creates a following for the creativity and advertisement
of the product (Alegro & Turnšekn, 2021).

13
2.4.2.2 Influencer’s Impact on Virality
The words that the influencer speaks and the way they interact with the viewers affect the
reactions and attitudes of the viewers on the message and product. Burgoon and Miller stated that
the “influencer’s psychological closeness, interactivity with viewers,” and other linguistic styles
can increase the video performance (Alegro & Turnšek, 2021).

Initially, the audience would recognize if the influencer’s face were present. According to
Thelwall, forming parasocial relationships between the viewer and the influencer makes the
influencer more relatable as it is similar to friendship in everyday life, so if the influencer does
not show their face, it creates inauthenticity (Thelwall et al., 2022).

To develop more friendly attitudes, some lifestyle influencers mimic conversations


between friends (Thelwall et al., 2022). However, another approach is breaking the third wall
and interacting with the audience. “By asking questions and inviting feedback” the influencer
can appear more trustworthy and personal, improving chances of better video performance (Lee
& Theokary, 2021).

The influencer can also give the impression of being relatable by giving personal reviews,
including being honest about their opinion on the product. Speaking about the flaws of the
product allows the consumers to view the influencer as authentic, making them more likely to
trust to purchase the product. Then, discussing the influencers’ experiences and personal
thoughts on the advertised item would establish a more positive relationship (Ge & Guo, 2022).

2.4.2.3 Rational Appeal v. Emotional Appeal


The topic that influencers speak about and the strategies they use will have an impact on
the way it is perceived. According to the study done by the Journal of Contemporary
Administration in Brazil, addressing real-life problems can bring a sense of purpose to a product,
increasing the chances of the product and video being shared with others. Using rational appeal
is best for products that are more technical, niche, or newly presented to the audience, as the
audience does not have a confident understanding of the product. Also, viewers who already
want to buy a specific product and want more information about the given item prefer videos

14
with higher rational appeal. Although it is important to discuss the functionality of a product as it
brings insight to the consumer, the more technical facts there are, the less chance there is for the
video to be viewed and shared (Isabella et al., 2023).

Informational videos are beneficial for new products in the market, but videos that evoke
emotions tend to impact virality more positively, especially those with positive feelings (Isabella
et al., 2023). Figure 5 summarizes results showing ads that induce positive feelings cause more
shares (Tellis et al., 2019). While the study shows that coefficients of the negative emotions are
also high and show significance in causing more shares, the focus of our team’s study will be on
positive emotions, per the sponsor’s request. The emotion that is most likely to be shared is
inspiration, followed by warmth (Tellis et al., 2019).

Influencers who present as excited and passionate correlate with higher video
performance (Lee and Theokary, 2021). Emotional contagion, meaning the audience experiences
the emotions that they subconsciously mimic, is a reason for this outcome. Contagion causes
more views and subscribers as viewer attitudes and emotions are influenced by video mood.

Figure 5. Ad Characteristics Impact Analysis (Lee & Theokary, 2021)

A specific emotion that helps increase engagement is humor. According to Tucker,


humor is a great contributor to receiving viewers' engagement especially since it would not

15
damage the advertisement message like other strong emotions (Tucker, 2015). However, the
influencer promoting the product needs to consider the campaign they are in as it may not fit
with brands with more serious issues. On the other hand, fun brands, those with a younger
audience, and those that hope to revive their image should prioritize joyful and humorous themes
(Dobele et al., 2007).

2.4.2.4 Influence of Video Background

Video background, including the setting and music, relate to performance of a video. The
audience would not be as satisfied if the video had an irrelevant background to the video
content’s topic (Ge & Guo, 2022).

Music is another content message characteristic that requires cognitive effort to


understand. There are many diverse types of music in the world, meaning many emotions and
messages can be expressed using music. However, not all will be significant in virality. Many
influencers try to use popular sounds to get their videos viral. Yet, using viral sounds does not
guarantee an improved video performance. When influencers use music, it should match the
video’s plot as well as the visuals. The likelihood of the video being shared is increased when the
soundtrack is popular and fits the video (Cunnane & Corcoran, 2018).

One music genre that does exceptionally well in advertising is coming-of-age songs,
those that are listened to during the adolescence phase. When viewers are presented with songs
that relate to their experiences or pasts, they tend to be more engaged, building more effective
persuasion (Lourenço et al., 2023).

2.4.3 Creator Feature


Creator features are objective features that are immediately recognizable to the viewer.
Some of these features include the influencer’s demographic details such as their gender, race,
and age, as well as the influencer account’s average numerical video metrics. In addition to the
creator features as outlined in Han et al., 2020, information found through the sponsor’s
WotoHub platform such as their WotoHub score, average engagement, commercial content

16
engagement percentage, top audience region, audience gender, and audience age range also fall
within this category.

The average numerical video metric of a creator (average number of likes, views,
comments, and shares for an influencer’s account) is the predominant measure of a video’s
success, thus, when new viewers see high numbers in these areas, they tend to have a greater
initial trust in the creator (Seo et al., 2019). However, this is not always the case. In advertising,
the likes correlation can vary. In a study done by researchers in Auckland and South Korea, they
found that when an advertisement does not explicitly acknowledge its self-serving goal of
promoting sales or awareness, then more likes are interpreted negatively. When the self-serving
goals are explicitly acknowledged in the advertisement, more likes on the post are viewed
positively. This may be due to growing consumer skepticism in social media. Consumers can
become skeptical of posts and believe the likes were bought rather than rightfully earned. (Seo et
al., 2019). The content of a video has been found to be significantly more important than the
opinions of other viewers expressed via likes (Foster, 2020).

There is a high correlation between the number of likes and the number of followers
(Moon & Yoo, 2022). There is a suggestion that the relationship between the number of
followers and the number of likes have an inverted U-relationship, proposing that there is a point
where having a larger number of followers will not increase the like count but instead a decrease
could be seen.

Personal demographics can impact a viewer’s perception of certain creator history


factors, which will be discussed in the next section. For example, a creator’s age and gender
would affect how attractive they seem to the user. A person’s gender or race may undermine or
overinflate creator authority on a given subject. A recent example is a female with a PhD in
Physics who kept having the educational creator, Hank Green tagged in her comments, despite
him not having a PhD in physics and both creators telling people to stop doing this (Leighton,
2022).

17
2.4.4 Creator History
Creator history, like content message characteristics, are the traits that are known by an
influencer’s viewers but are not known about the influencer at first glance. Creator history can
include an influencer’s credibility, authenticity, reputation, attractiveness, expertise on the
subject, trustworthiness, originality, expectations for their videos, and the creator’s authenticity.
All the creator's historical traits are subjective and are developed over time while the creator
continues to post online content. These characteristics can only be gauged by users over a longer
period, and thus we classify them as creator history (Han et al., 2020). Creator history will be
very key to our study; therefore, it is important to have an idea of how the creator's history can
affect the performance of content posted on social media.

2.4.4.1 Role of Creator History on Consumers


Creator history can have a big role in the consumer trends of social media users when
viewing a social media influencer’s ad. During a study on the effects of creator attractiveness,
expertise, and trustworthiness on social media viewers' consumer trends, researchers using the
results of their experiment were able to support their hypothesis that attractiveness, expertise,
and trustworthiness individually and combined will all have a positive influence on a social
media viewer’s advertisement attitude, product attitude, and purchase intention (Chekima et al.,
2020) (see Figure 6).

Another study on the effects of different creator history attributes on viewer purchase
intention was able to confirm a few hypotheses on the subject. This study confirmed their
hypothesis that trustworthiness, expertise, and parasocial relationships all positively influenced a
social media user's intention to purchase the advertised product (Masuda et al., 2022). However,
that doesn’t properly state the importance of a good creator history. Masuda et al., were able to
verify that influencers’ physical and social attractiveness both positively influenced the creator's
trustworthiness, perceived expertise, and parasocial relationship.

Another important creator history characteristic is authenticity. Authenticity is defined as


a post or photo that demonstrates some aspect of the celebrity’s true self. Authenticity will result
in positively influencing viewer purchase likelihood (Kowalczyk & Pounders, 2016).

18
Although it is uncertain whether creator history has a bigger effect on a video’s
performance compared to content feature, content message, and creator features, it is certain that
if the creator making the social media post is attractive and perceived as trustworthy, the post
will most likely perform better than if the creator was unattractive and perceived as not
trustworthy (see Figure 7).

Figure 6. Hypothesis model, (Chekima et al., 2020)

Figure 7. Hypothesis results (Masuda et al., 2022)

19
2.5 Project Goal
Throughout our IQP project, the project goal was to provide social media video
recommendations, using the previously defined video characteristics framework, to our sponsor
to improve social media video virality. The format we provided the video recommendations in is
a mind map used to identify key features in social media videos and a PowerPoint presentation
that explained in detail the results of our project.

To accomplish our project goal, we decided answering the following questions would be
an effective method to help complete our final deliverables: Which content features affect social
media virality the most? Which content messages affect social media virality the most? Which
creator features affect social media virality the most? Which creator history characteristics affect
social media virality the most? The next section of this paper will discuss how our team worked
to answer these questions.

20
3. Methods
In this section, we will discuss the methodologies conducted for our team’s research.
Before our team arrived in Hangzhou, China, our team had agreed with our sponsor to conduct a
large Excel sheet analysis, social media user behavior surveys, and influencer and brand
interviews. However, once arriving in China, we discovered that the large Excel sheet analysis
lacked many rows and informative fields, preventing the conduction of an accurate analysis of
the information. In addition, WotoKOL shifted the project methodology and requested for a case
study analysis to be conducted in place of the data analysis. For more details on this specific
limitation the project faced, please see Section 6.

With this shift of the project methodology in mind, the new primary focus in
methodology was to conduct a case study analysis on promotional videos for companies:
Company A, a wearable electronic device brand, and Company B, a household appliance brand
focusing on vacuums. From there, the secondary focus of our methodology became the social
media user surveys. The main objective of the user survey was to gain a better understanding of
the video creation process and the role of the four video characteristics categories in the virality
of a video. Due to the shift in methodology and project priorities, the originally proposed plan to
conduct influencer interviews was removed from our project. Both the case study analysis and
social media user survey were created with the original framework, as shown in Figure 3.

Following the completion of the case study analysis and user survey execution, a
quantitative and qualitative analysis using Microsoft Excel, SPSS, and NVivo were used to
conclude to support our literature research. In the remainder of this section, the methodologies
will be described in detail, as well as outlining the limitations of each method described.

3.1 Data Collection


To carry out the project goal, data was gathered using two methods: case studies looking
at characteristics of social media videos and surveys sent out to social media users asking about
social media preferences. When collecting data from the case studies and surveys, the main focus
was on four categories of video characteristics and their effect on the virality of a social media
video. Data was collected on those same characteristics in the survey to compare the results of

21
the case studies with that of the survey results, and the findings of the background research that
was conducted.

3.1.1 Case Study Methods


Following the sponsor’s request and the shift in methodology, the primary focus of this
project became a case study analysis of influencer videos with paid advertisements for the
products of two different companies. Due to a confidentiality agreement with WotoKOL, our
team is unable to disclose the names or products of these brands. It will refer to them as the
following: Company A, an electronic device brand focusing on Augmented Reality technology,
and Company B, a household appliance and cleaning device brand.

The case study analysis of these two companies was broken up into four stages: 1. video
collection, 2. factor determination and framework development 3. video review and analysis, and
4. quantitative and qualitative analysis of the collected data. As this section of the paper focuses
on the process of the methods, please refer to Section 4.2 for more information on case study
analysis.

3.1.1.1 Video Collection


For the Company A case study: promotional videos were collected using one of the
companies' product names and using the search bar feature on each social media platform to find
related videos. This search was completed using YouTube, Instagram, and TikTok. Using the
videos found on each platform, the video links were compiled into a Microsoft Excel spreadsheet
in one column. A total of 33 videos were found and compiled into the Excel spreadsheet for
analysis.

WotoKOL provided 37 videos for analysis and review for the Company B case study.
This is different from the Company A case study video collection, as Company B is a large client
of WotoKOL, thus WotoKOL was able to provide historical video data. These videos are a
combination of those that performed successfully based on Company B’s standards, some that
performed moderately, and some that performed poorly. The performance of these videos was
determined via our sponsor and through Company B. These 37 videos were listed on a separate
page of the Excel spreadsheet in one column.

22
3.1.1.2 Factor Determination and Framework Development
Following the collection of the videos for both Company A and Company B, the field of
factors used for video analysis was determined. Using the framework discussed in Section 2.4,
several analysis fields were defined and focused on content features, content messages, creator
features, and creator history factors. These factors include many of the features discussed
previously such as captions, titles, hashtags, and the number of likes a video has received (Han et
al., 2020). Furthermore, emotional appeal and creativity were incorporated as additional features
in the video characteristics framework, as Segev et al.’s definitions on these features aligned with
that previously discussed in Section 2.4.1 (Segev et al., 2023). These analyses fields were then
broken up into three different response types: Yes or No responses, Likert scale responses, and
direct content feature responses.

Y/N Questions

When conducting the case study, many different content and creator characteristics were
observed, likewise, some of these characteristics are best captured in whether they appear in the
video. An example of this type of question would be “Does the video contain humor?”. Yes and
no questions can be asked about all types of content and creator features.

In the case studies, eight different yes and no questions were asked. These questions
were: whether the video contains music at any point, whether the influencer of a video shows
their face, whether the video has a hook in the video, whether humor is contained at least once
throughout the video, whether the video contains a form of rational appeal, whether the video
states con about the product, whether the video gives an opinion about the product, whether the
influencer prompted viewers questions, and whether the influencer asked viewers for feedback
on the video.

To identify whether these features were contained in the videos while conducting the case
studies, two team members watched through the videos looking for these features. If both
members agreed that the feature was present in the video, it would be identified as yes. If only
one member identified the feature being present in the video, a third member of the team would

23
make the final decision of whether the feature was in the video. If both members agreed the
feature was not found in the video, the feature would be identified as no for that video.

Each question in this section was based on the secondary background research conducted
in Section 2.4. “Does the video contain music” was included as a yes or no question because
secondary research, found that including music in your video can affect the performance of the
video, along with the type of music also influencing the performance (Lourenço et al., 2023). To
further support this secondary research finding, music appropriateness was included in the Likert
questions as well.

Going back to the secondary research, whether a video contains humor was included in
the yes or no question section, as research found that humor is one of the strongest emotions that
can also increase viewer engagement (Tucker, 2015). It is also very difficult to measure humor,
thus, including it as a yes or no question would be the most accurate way to measure humor.

Rational appeal was included in the yes and no questions as research found that rational
appeal did not have much positive influence on virality, therefore, our team wanted to confirm
these findings (Isabella et al, 2023).

Whether the influencer states cons about the product fell into the yes and no section
because our research found that being honest about a product can cause the viewer to trust the
video and be more likely to trust and purchase the advertised product (Ge & Guo, 2022).

Whether the influencer includes an opinion of the product was included in this section
because research found speaking on the cons of a product, as well as the influencer saying their
opinions on a product can make the viewer more likely to purchase the advertised product (Ge &
Guo, 2022).

Whether the influencer asked questions to viewers, and whether the influencer asked
viewers to comment on the video was included in this section, secondary research showed that an
influencer interacting with the influencer’s viewers can increase viewer engagement (Ge & Guo,
2022).

24
Any feature that was decided not to be included as a yes and no question was better fit as
a Likert question, or was readily available from viewing the video, the creator’s page, or the
creator’s WotoHub page.

Likert Questions
Similar to the yes and no questions from Section 3.1.1.2, there was a need to assess
many video content and creator characteristics. These features include content message or
creator history characteristics, which makes them difficult to rate with yes and no. For example,
the field “The video is creative,” could be ranked with yes or no, but to get a more accurate
rating of how creative this video is, the team decided to rank these questions using a scale from
one to seven, one being strongly disagree, and seven being strongly agree.

Eleven features were ranked using the Likert-type questions. These questions include
creativity, music appropriateness, video setting appropriateness, video production quality,
influencer excitement, influencer passion, emotional appeal, influencer attractiveness, influencer
trustworthiness, influencer authenticity, influencer subject expertise.

Likert scale is used to evaluate characteristics to get more nuanced scores of the features.
The Likert scale was devised to measure ‘attitude’ in a scientifically accepted and validated
manner in 1932 (Joshi et al., 2015). As the Likert scale is scientifically accepted, our team found
that this would be the best way to go about measuring these features.

Each question in this section was based on the background research discussed in Section
2.4. Video creativity was included in our Likert scale question sections as our team felt creativity
would be best measured in a Likert scale form. The research shows that creative video
advertisements are more memorable to viewers, which could affect virality. Another feature
included in the Likert question section is the appropriateness of the music in a video. As stated in
the yes and no question section, based on secondary research, the appropriateness of music can
influence video virality, and our group found that analyzing music appropriateness on the Likert
scale would be more useful to the results. Video setting appropriateness was another feature
measured on the Likert scale, as in secondary background research it was found that setting
appropriateness can influence video virality.

25
Past research on setting appropriateness and music appropriateness also used a Likert
scale for these characteristics, leading our team to follow suit. Production quality was another
feature that our team chose to incorporate in the Likert question section, as rating it using this
scale influenced viewership. Our team chose to include influencer excitement, influencer
passion, and influencer attractiveness in the Likert question section as our research showed that
these three characteristics all lead toward positive ad attitude, product attitude, and purchase
intention. Emotional appeal, influencer excitement, and influencer passion are the three final
factors that we included in the Likert question section, as in our secondary research, we found
that all three of these factors resulted in increased video performance.

Content and Creator Features


During the case study, three types of data were gathered, yes and no data, Likert-type
scale data, and content features. Content features and creator features include video and creator
characteristics easily recognizable just by viewing the video, or the analytics of the video using
the WotoHub. Some examples of content and creator features are the view count of a video, the
like count of a video, the title, the hashtags of a video, the number of followers that the creator
has, and the number of videos that the creator has posted.

The team evaluated the numerical video metrics, descriptive video elements, and the
videos’ properties, as discussed in Section 2.4.1 and Section 2.4.3. Some other features that were
recorded were only available using the WotoHub. These features are the ‘WotoHub Score’, the
creator’s average like count, the average views in the last 60 days (about 2 months), the region
with the highest count of viewers that watch the creator, the gender with the highest view count
of the creator, the age range with the highest count of viewers that watch the creator, the average
engagement rate of the creator, and the average commercial engagement rate of the creator.

During the process of performing the case studies, the content features and creator
features were measured by having a team member look at the video page, the influencers page,
and the influencer’s page on the WotoHub. This made content features and creator features the
easiest of the features to measure. We did not have to use a second team member to check these
metrics as none of them are subjective features.

26
During the case study, content and creator features were used to determine a video's
virality. All the features the team recorded were easily compared with the Likert-rated
characteristics, and yes and no questions could come to conclusions on video virality. We needed
to have something to measure virality with, thus these content and creator features served that
purpose.

3.1.2 Social Media User Surveys


To support the conclusions drawn from our case study analysis for Companies A and B,
our team conducted a survey of social media users to gather information on user behavior. We
had several objectives for our survey. First, we wanted to understand viewers’ opinions about the
four main video characteristics defined in Section 2.4. These characteristics are intertwined in
the observational data we gathered through the case studies; thus, the main goal of the survey
was to determine which features are the most important to international social media users. The
second project goal was to collect demographic data of different social media users to determine
how different groups use social media. This process helped to identify biases in the survey data
and what limits existed on any conclusions we may draw. The third objective joined the first two
objectives to analyze which video characteristics are most important to different types of
viewers.

3.1.2.1 Identifying Survey Subjects


The intended audience for the survey is anyone outside of China who uses social media.
WotoKOL only assists brands with expanding their global reach, thus the focus on international
social media platforms rather than domestic Chinese platforms. Due to the restrictions on using
WPI’s fully undergraduate email alias to broadly deliver surveys to users, a convenience sample
consisting of friends, family, and other participants was used. In addition, several student
organizations were approached to complete the survey. These groups included the WPI chapter
of the Society of Asian Scientists and Engineers, Chi Omega Sorority, and ESports club -
Valorant Team.
In an ideal study a control group of those who do not use social media would have been
used. Time constraints did not allow for a control group.

27
3.1.2.2 Survey Structure
Using Microsoft Forms, the social media user behavior survey was created and broken
into three main sections: 1. personal demographic questions, 2. social media demographic
questions, and 3. social media behavior questions. Before the questions were presented to the
survey takers, they were given a page informing them of our team’s research goal, that their data
would not be presented individually or tied back to them, and that they had the option to choose
not to fill out any of the questions on the survey if they so choose. This was done to ensure we
had survey taker informed consent to use their data in the study. For more details, please see
Appendix A for the full survey Preamble and Confidentiality Notice.
Survey takers were asked personal demographics questions regarding age, race, ethnicity,
gender identity, sexual orientation, family income, and highest degree or level of education, in
the first section of the survey. The purpose of these questions was to help determine correlations
between certain groups and opinions. It was also used later in our process to help our team
identify bias within our survey sample.
In section 2: social media demographic questions, the survey takers were presented with
two questions. The first question asked the user to rank how often they used the social media
platforms of YouTube, Instagram, and TikTok using a five-point Likert scale. This Likert scale
was defined as the following: 1- Never, 2- Rare, 3- Sometimes, 4- Often, and 5-Always. The
second question in section 2 asked the user how much time they spend on social media daily,
with the options spanning from less than one hour a day to over ten hours a day.
The final section of our survey, social media user behavior, was broken up into three
categories of questions, all of which used a five-point Likert scale rating. For this category of
questions, the five-point Likert scale was defined as the following: 1- Strongly Disagree, 2-
Disagree, 3- Neutral, 4- Agree, and 5- Strongly Agree.
In this section, the first category of questions centered around the different types of social
media genres. Using the framework developed in Section 2.4, as a reference, and WotoKol’s
feedback, our team selected 12 different video genres that best fit the purpose of our survey. For
each statement given to the survey taker, they were asked to rank how much they agreed with
each of the statements.

28
The second category of questions focused on how each user finds new social media
content. Similar to the previous genre of questions, these questions asked the survey takers to
rank how much they agreed with each of the statements within that section. Using the platform
features previously mentioned in Section 2.3.3 and our team’s personal social media experience,
our team chose seven potential discovery methods that the survey takers may find new social
media through.
For the final category questions, using our team’s previously discussed framework as
adjusted from Han et al., we created a series of 23 questions regarding the user’s social media
use. Like the previous two categories of questions, this series of questions used the same five-
point Likert scale to allow the survey taker to rank how much agreed with an individual
statement. The full survey format and the question format appear in Appendix B.

3.1.2.3 Survey Distribution


The team used various communication platforms to distribute the survey and recruit the
convenience survey pool. This includes using personal text messages, Discord, Slack, and
Microsoft Outlook email. While this introduced a bias into the data set, it was chosen to acquire
as many responses as possible. This bias may also lead to a narrow or skewed demographic data
set, which are further discussed in limitations.

3.2 Data Analysis Plan


We analyzed two different data sources to identify features with the strongest relationship
to video performance. The first source was the survey sent out to social media users outside of
China to learn user video-watching tendencies. After receiving 63 responses, our team
transferred the results to the HDU students for evaluation. The second data source was the videos
that were studied by our team. This dataset consists of textual, Likert-type scale rating, time and
date-based, continuous, categorical, and binary data. Due to the variety of data types, we
conducted qualitative analyses and quantitative analyses. NVivo, a qualitative analysis software,
was utilized to design word clouds for textual data. Quantitative analyses were conducted
through Excel and the Statistical Package for Social Sciences (SPSS), a software for advanced
statistical analysis, to run statistical and regression analyses, respectively.

29
In this section, we will be detailing the process in which the survey analysis and case
study analysis were done. Beginning with the survey analysis, we describe how we prepared and
transformed the survey responses into a spreadsheet for the HDU students to analyze. In the case
study analyses subsection, we will review the steps to develop the word cloud, statistical, and
regression analysis.

3.2.1 Viewer Survey Quantitative Analyses


Our user survey was used to supplement our case study data. The data analysis consisted
of analyzing the distributions of the answers to the questions to determine which survey results
were statistically significant. This accompanied some more basic distribution and frequency
analysis. This was conducted using Excel and SPSS.

3.2.2 Case Study Analyses


The case study data was analyzed upon completion of the case study data collection.
First, we analyzed Company A’s and Company B’s videos separately. Then we combined and
analyzed the joint data sets from Company A’s and Company B’s videos. The analysis process is
discussed in this section. The analysis of the results appears in Section 4.
We performed both qualitative and quantitative analyses. Our qualitative analysis
consisted of word cloud creation using the software NVivo to show common words used in our
textual fields. The quantitative analysis uses Excel and SPSS. Summary statistics and one-to-one
comparisons between different fields are completed on Excel. Excel was also used to prepare and
clean the data for SPSS analysis. Within SPSS, a multiplicity of linear regression models—
automated and determined through SPSS. A more detailed description of the analysis methods is
discussed in Section 3.2.2.1 and 3.2.2.2.
Before further analysis some method limitations are discussed. Due to the time restraints
of this project, our team was only capable of analyzing a limited number of videos for each case
study. This issue is further compounded when the data is segmented or when some videos do not
have every feature due to differences between platforms and video styles. In addition, the
subjective video rating system used in this study introduced bias into the data set. For Likert
scale opinion questions, 2-4 team members were rating each video. Ideally, with more resources,

30
this number would be much higher, but for the scope and time frame of our study, this was
determined to be the most efficient option in balancing between bias and time. Another facet that
comes with the small sample size of the raters is the lack of diversity among the raters. The raters
in this study are not diverse in terms of factors like age, gender, socioeconomic background,
geographic background, and education level.

3.2.2.1 Case Study Qualitative Analyses


Using the titles, hashtags, and captions collected through each video in both case studies,
NVivo was used to create word clouds based on the occurrence of keywords or phrases. These
word clouds were created both with all videos used in each respective case and for the top-
performing videos of each case.
To create the word clouds using the NVivo platform, the captions, hashtags, and titles of
the videos were placed into their own respective Word documents. These Word documents were
uploaded to NVivo, where the software then created a Microsoft Excel sheet, calculating the
number of occurrences of each word or phrase. With these words and phrases accounted for,
word clouds were then generated for each category using NVivo’s visualization tools. To
identify the top-performing videos for each case study, our team found the videos with the
highest combination of likes, views, and shares. Using this combination, the top 15 videos for
each case were then chosen.

3.2.2.2 Case Study Quantitative Analyses


To determine what affects the performance of a video the most positively, we examined
the different video features that could affect the virality of a video. To begin the analysis, we
established a list of analyses that would help us achieve our goal. However, when we started on
creating the comparison, we statistically summarized all the features of the spreadsheet. Two
statistical software packages are used: Excel and SPSS. For statistical analyses, we designed
charts that represented the dataset visually in Excel. For regression analyses, we executed linear
regression models to determine which factors contributed most to our virality metrics. For
correlation analyses, we examined the correlation matrices generated for the factors. In this
section, we will describe the steps in the quantitative analysis of the case studies.

31
Excel Quantitative Analysis
To ensure that the analyses aided our team in producing the deliverables and were
conducted thoroughly, the objectives for the analyses were first identified. A checklist of the
comparisons was created to outline which comparisons could be made. See Appendix C for more
details. We described each comparison with the proper steps to design graphs that would
correctly measure that comparison and divided the task of creating those comparison graphs
amongst the team. As the dataset for both companies contained the same features, the objective
list was implemented for both case studies.
The first step in executing a complete statistical analysis was to look at the general
statistics of the spreadsheet. We generated a tab for each dataset. By categorizing the features by
the data type, we were able to organize tables and efficiently determine general descriptive
statistics. For the numerical, time-based, date-based, and Likert scale rated questions, we
calculated the average values, and percentages of N/A contained in each column. We also
identify the minimums, lower quartiles, medians, upper quartiles, and the maximums, followed
by the interquartile range (IQR), and the differences between the mean and median found in each
column.
Our team then found the portion of each category for all categorical variables, and similar
to the categorical variables, we later computed the percentage of the yes and no response for the
binary variables. The last category that we computed was categories containing textual
information. An example of textual information is the hashtags contained in a video. With this
textual information, we counted the number of videos that each hashtag was used in and its
percentage.
To perform the case study quantitative analysis, the team decided to conduct numerous
comparisons of the data. The first comparison was for content features such as views, comments,
and shares; correlations for these features was also determined. The second type of comparison
was comparing binary response questions with number of engaged viewers. The third
comparison compared the Likert scale questions with video performance including engagement
rate and views. These comparisons were completed for both Company A and Company B.
Initially we need to determine how the number of views, comments, shares, followers,
engaged viewers, and video engagement rate compared with each other. To establish the
relationship between the six features, scatter plots and trendlines using Excel were used to

32
display relationships and trend direction. Correlation coefficients and p-value were used to
determine the strength and direction of the relationship.
The yes and no questions are compared to the video view count and the video
engagement rate. The comparison of each yes and no question to the view count and video
engagement rate was separately conducted for each case study. A comparison of the yes and no
questions from using joint data from both case studies was then completed. This analysis resulted
in three graphs containing all the yes and no question comparisons—see Section 4.
To determine how the content message characteristic Likert questions were related to the
video performance, these responses were compared with video performance features. Although
correlation between the number of views, comments, shares, followers, engaged viewers, and
video engagement rate exist, graphs and tables for the Likert question relationships with views,
shares, engaged viewers, and video engagement rate to learn the feature with the strongest
association were created. Consequently, for each Likert question, we graphed four scatter plots
with a trendline. The correlation coefficient and p-value were then computed to identify which
relationships were significant.

Following the completion of the yes and no comparisons, and Likert scale comparisons,
additional comparisons and robustness checks considered relationships between other factors.
These comparisons were performed to analyze all fields outside the yes and no, and Likert-type
questions. For example, the WotoHub score was compared with different video performance
features to look for a relationship between the values. Other features comparisons, such as the
percentage of male viewership and video age to performance features were also conducted. We
computed the correlation coefficient and p-value to identify which relationships were significant
for these comparisons.

The graphs were compiled into one Excel sheet. The correlation coefficient and the p-
values were calculated to determine significance and trends of relationships.

SPSS Data Analysis Preparation and Process

Before inputting our case study data into SPSS, we first cleaned the data in Excel. This
process began after data collection. Several minor inconsistencies in the data required resolving.

33
The fields containing “yes” or “no” responses had “y” and “n” characters in some cases and “Y”
and “N” in other cases. These were fixed to all be lowercase for SPSS to recognize them as two
categories correctly, rather than four distinct categories. The remaining single-word textual
categories were transformed into sentence cases. Entries marked as “N/A” were removed and left
as empty cells to not be misinterpreted by SPSS.

After these consistency corrections, some textual data was transformed into numeric data.
To do this, we split several key variables into binary dummy variable columns. Dummy variables
split a column with multiple categorical variables stored in one variable into multiple binary
columns representing each possible variable. For example, the variable Platform can contain
“YouTube”, “Instagram”, or “TikTok.” So instead of one variable storing one of those three
values, it is transformed to three variables representing each potential value. If the original value
was “YouTube” then the “YouTube” variable would be filled with a one and the “Instagram”
and “TikTok” variables would both have zeroes. We split Platform, Language(s) in the video,
Hashtags, Highest possible resolution, and gender. For hashtags, we did not include a column for
every hashtag featured in a video, only hashtags in three or more videos had a dummy variable
column created. Then we transformed all of our yes/no questions into binary variables as well
where one indicates yes and zero indicates no.

Once the data was uploaded from the Excel file into SPSS, we then needed to adjust the
“Measure” and “Type” of each field. “Measure” refers to whether SPSS interprets the field as
“Scale”, “Ordinal”, or “Nominal”. Scale is for continuous numerical interval and ratio data.
Ordinal is for ordered categorical data. Nominal is for unordered categorical data. “Type” refers
to the data type of the field. We used date, string, numeric, and percent. See Appendix D for the
data structure of the database on each variable’s identifying “Type” and “Measure” fields.

A linear regression model using SPSS’s automatic linear modeling tool was used to
evaluate the data. This tool includes automatic data preparation and automatic subset selection.
Since we included all fields in the model, we did not utilize the automatic subset selection,
however we did use the automatic data preparation. SPSS’s automatic linear modelling tool
completes 4 main data cleaning steps: it replaces missing values; it converts time data into
duration data; it specifies categorical predictors; it identifies and handles outliers (Yang, 2013).

34
There are some drawbacks to the automatic nature of the tool, such as in the depth of
analysis researchers may do to determine outlier cases. However, that is not thoroughly analyzed
in this study, so the outlier analysis provided is enough for our purposes. We ran this six times
per data set, Company A’s data, Company B’s data, and the combined data. Across the six
models, the independent variable inputs remained the same, but the dependent variable target
was varied amongst the following fields: view count, like count, comment count, share count,
video engagement rate, and total engaged viewers. In total, we ran 18 models. See Appendix E
for the independent variables used. We aimed to include all predictors in every generated model.

A correlation matrix and a corresponding significance matrix were generated after


determining the most effective linear regression models. The correlation matrix consists of the
correlation values across pairwise cells for the features. These values may be negative to positive
where the closer the correlation’s absolute value is to one, the more significant the relationship.
The closer the value is to zero, the less correlated the two factors. We mostly analyzed the data
comparing the independent variables to dependent variables; however, high correlations between
independent variables were not ignored because they could be indicative of bias or overlap
between factors.

35
4. Results and Findings
This section evaluates each of the research questions based on primary and secondary
research findings. The specific questions and subsections in this results section are titled around
the research questions: “What content features most relate to social media virality?”, “What
content messages most relate to social media virality?”, “What creator features most relate to
social media virality?”, and “What creator history characteristics relate to social media virality
the most?” These research questions and dimensions are based on video characteristic to virality
study relationships such as (Han, et al., 2020).

To answer each question, we will provide supporting primary and secondary research
evidence to provide insight into the research question, eventually supporting recommendations to
the sponsor.

To avoid providing too much unnecessary information that may be tangential to answer
the research questions, we will not be evaluating all the data, but only presenting and evaluating
data to support insights into the research question. We prioritized the outcomes that were the
most significant or outcomes that we found interesting or matched our literature review.
Additionally, due to confidentiality restrictions, we are unable to disclose the two company
names and specific details about the companies.

4.1 Video Content Features and Virality


The case studies resulted in various video content message features having a significant
correlation with different virality metrics (See Table 1). This section presents the content features
that had a p-value < 0.1, which is typically considered the largest p-value for weak statistical
significance for empirical data (Fisher, 1956). This section is broken up into three subsections:
content features results from Company A, Company B, and Combined Case Studies.

4.1.1 Case Company A


To establish the relationships between the six video performance metrics, we calculated
the correlation coefficients between them. Table 1 displays that the count of views, likes,
comments, shares, and engaged viewers all have a significant relationship, while the video

36
engagement rate stood alone with no correlations to the other metrics. This lack of uniformity
meant it would not be feasible to create one combined metric representing a relationship with all
of them.

Views Likes Comments Shares Engagement Engaged


Rate Viewers
Views Correlation 1 0.975 0.522 0.582 -0.149 0.968
coefficient
p-value <0.001 0.011 0.004 0.272 <0.001
Likes Correlation 1 0.549 0.586 0.019 0.993
coefficient
p-value 0.007 0.004 0.469 0.000
Comments Correlation 1 0.929 0.000 0.640
coefficient
p-value 0.000 0.500 0.002
Shares Correlation 1 -0.045 0.673
coefficient
p-value 0.427 0.001
Engagement Correlation 1 0.014
Rate coefficient
p-value 0.477
Engaged Correlation 1
Viewers coefficient
p-value
Table 1. Company A Numerical Video Metrics Relationships with p-value

Table 2 highlights the results of six linear regression models on the Company A data set.
The independent variables can be found in Appendix D. All significances are relatively low
showing there were few overarching trends in the data for the model to fit, with the exception of
the share count model. We believe this model overfits due to having fewer viable entries feeding
into the model, as the share count on some platforms was not available for data collection. The
main trend to notice here is that #augmentedreality’s presence attached to a video was a major
predictor of the success of the video. Correlation does not imply causation.
Survey results—going beyond the case studies evaluated for the two companies, see
Appendix F—showed that most social media users don’t pay attention to hashtags, suggesting
that any association related to performance is likely due to the content creator or platform
algorithm. Although given the unique characteristic of Company A showed one hashtag has
some significance in some of the measures based on the linear regression results of Table 2. An

37
N/A in the third column represents that no statistically significant estimated means relationships
with independent variables were found in the model.

Linear Regression
Dependent Variable Adjusted Statistically Significant Independent
(Performance) R2 (Predictor) Variables
View Count 21.0% N/A
Like Count 38.8% N/A
Comment Count 14.9% #augmentedreality
Share Count 100% Commercial content engagement
rate, average creator engagement
rate, average views (in the last 60
days), average likes, WotoHub
score, number of posts, time length,
number of followers, production
expertise, attractiveness
Engaged Viewers 40.7% Average views (in the last 60 days)
Video Engagement 60.1% N/A
Rate
Table 2. Summary of Linear Regression Results for Company A
Video Virality Relationships to Video Content Features

Figure 8 shows one significant set of direct relationships investigated with the data from
this study. Production Quality vs Views from Company A had a significant p-value < 0.05. This
comparison aligned with secondary research findings from Section 2.4.1.3, saying that videos
with worse production quality would receive less views (Tribbey, 2016).

Figure 8. Company A Production Quality to Viewers Relationship Graph.

38
4.1.2 Case Company B
Unlike Company A, many of the comparisons between the video performance metrics
were not significant, as shown in Table 3. As secondary research has shown and discussed in
Section 2.4.3 it is stated that there should be a relatively high significance relationship amongst
these elements (Moon & Yoo, 2022). A possible explanation for the contrast in results is the
small dataset. There is a strong positive relationship between views, likes, and engaged viewers,
but no further relationships with or between any of the other metrics. This made creating one
aggregate outcome (dependent variable) metric or factor even less feasible than with Company
A.

Views Likes Comments Shares Engagement Engaged


Rate Viewers
Views Correlation 1 0.933 0.031 0.289 -0.149 0.934
coefficient
p-value <0.001 0.454 0.139 0.290 <0.001
Likes Correlation 1 0.165 0.236 0.150 1.000
coefficient
p-value 0.271 0.190 0.289 0.000
Comments Correlation 1 0.465 0.606 0.172
coefficient
p-value 0.035 0.006 0.262
Shares Correlation 1 0.177 0.246
coefficient
p-value 0.256 0.179
Engagement Correlation 1 0.154
Rate coefficient
p-value 0.285
Engaged Correlation 1
Viewers coefficient
p-value
Table 3. Company B Numerical Video Metrics Relationships with p-value

Similar to Company A, 6 linear regression models were completed for Company B’s data
set on video content features. The results are summarized in Table 4. The model fits (coefficient
of determination R-squared) across these models are better than they were for Company A. The
only exception is the model with share count as a dependent variable, which has an opposite
relationship from the Company A model. It may be an underfitting issue from having too few

39
valid data points to create a model. Some top predictors across the models were emotional
appeal and video creativity.
Linear Regression

Target Adjusted R2 Statistically Significant Relationships


View Count 93% Video Creativity, Number of
Followers, WotoHub Score
Like Count 88.9% Emotional Appeal, Video Creativity
Comment 55.3% N/A
Count
Share Count 1.4% N/A
Engaged 92.8% Emotional Appeal, Video Creativity,
Viewers Number of Followers, Expertise,
Commercial Content Engagement
Rate, Appropriate Music
Video 45.2% N/A
Engagement
Rate
Table 4. Summary of Linear Regression Results for Company B of
Video Virality Relationships to Video Content Features

4.1.3 Combined Case Studies


Six multiple linear regression models were run on the combined case study data set. The
models had virality factors as dependent variables and video content features as independent
variables (see Table 5). Similar to the models run on the company A data set these models
showed low accuracies and the top predictors are mostly hashtags that relate to the product.
The only model with top predictors other than hashtags was the video engagement rate
model with the second highest adjusted R2 of the combined data set models and had top
predictors of whether the video was on Instagram and the emotional appeal score. A major
takeaway from these models is that each product and video type will have its own set of
important predictors. More research is needed for advertising different types of products.

40
Target Adjusted R2 Statistically Significant Estimated
Means Relationships
View Count 29.1% N/A
Like Count 28.4% N/A
Comment 22.6% #augmentedreality, Redacted Hashtag
Count 2
Share Count 95.6% #augmentedreality
Engaged 28.8% #augmentedreality
Viewers
Video 46.8% Instagram, Emotional Appeal
Engagement
Rate
Table 5. Summary of Linear Regression Results for the Combined Case Study Data Set

Joint Company A and Company B metrics show additional significant relationships than
there were for Company B and fewer relationships than there were for Company A, further
supporting our conjecture that further study is required to determine the relationships between
each of these and the sales metrics of the product being advertised. In the combined data set,
there is a strong positive correlation between shares and comments and there is a separate
positive correlation between engaged viewers, views, and likes (See Table 6).
Views Likes Comments Shares Engagement Engaged
Rate Viewers
Views Correlation 1 0.929 0.469 0.534 -0.163 0.936
coefficient
p-value <0.001 0.002 <0.001 0.174 <0.001
Likes Correlation 1 0.407 0.441 0.091 0.995
coefficient
p-value 0.008 0.004 0.302 0.000
Comments Correlation 1 0.926 -0.039 0.496
coefficient
p-value 0.000 0.413 0.001
Shares Correlation 1 -0.073 0.528
coefficient
p-value 0.339 0.001
Engagement Correlation 1 0.080
Rate coefficient
p-value 0.324
Engaged Correlation 1
Viewers coefficient
p-value
Table 6. Dependent Variable to Dependent Variable Correlation Coefficients with p-values.

41
4.2 Video Content Message and Virality
The case studies resulted in various video content message variables (factors) having a
significant correlation with different virality metrics. This section summarizes the content
message features that had a p-value < 0.1 relationship or better.

Contains Influencer Contains Humor Contains Discusses Gives Influencer Influencer


Music? Face Hook? Used? Rational cons of opinion? asks viewer invites
Shown? Appeal? product? questions? comment?

Y* (number) 21 29 7 7 31 10 25 3 13
Y (percent) 64% 88% 21% 21% 94% 30% 76% 9% 39%
N (number) 12 4 26 26 2 23 8 30 20
N (percent) 36% 12% 79% 79% 6% 70% 24% 91% 61%
Table 7. Descriptive Statistics for Binary Evaluations of Video Message Features for Company A Characteristics.
*Y= Yes, N = No
Our results from the yes and no comparisons between content message characteristics and
engaged viewers and views had few significant features and few with positive and negative
correlations. The comparison with the most significant positive correlated result is humor being
used. Some other yes and no results to look at are ‘influencer face shown’ with a positive
correlation and ‘do they give an opinion’ as well with a positive correlation. Some yes and no
results with significance and a negative correlation are ‘Do they give cons,’ ‘influencer asked
viewers questions, and ‘influencer invited comments/feedback.’ The other yes and no results
were not significant enough to draw strong conclusions (see Figure 9).

Figure 9. Company A Engaged Viewers to Yes/No Question Relationship Graph

42
Additionally, there was an imbalance in videos that utilize humor and videos that don’t.
Only six videos were found to contain humor as shown in Table 7. The features, ‘Do they give
cons about the product?’ and ‘influencers asking viewers questions’, are skewed as well. From
this table, it can be concluded that they do not accurately represent all videos.

Humor Discusses cons Influencer asks


Used? of product? viewer questions?

Y* (number) 6 2 1
Y (percent) 16% 5% 3%
N (number) 31 35 36
N (percent) 84% 95% 97%
Table 8. Descriptive Statistics for Binary Evaluations of Video Message Features for Company B Characteristics.
*Y= Yes, N = No

There is an imbalance in the statistical summary of influencer excitement, emotional


appeal, and setting appropriateness—and should be considered in the evaluation given that a
small number may bias the result towards the ‘majority’ classification (Beralla, et al., 2021).
Based on skewness tests, influencer excitement, emotional appeal, and setting appropriateness
were found to have skewed data.

Company B case study results are similar to Company A results. Numerous comparisons
between content message and virality outcomes, either with significant positive or negative
correlation, were found. The primary research analysis with the strongest findings occurred with
yes and no evaluations the Likert-based scoring evaluations from the case study video empirical
evaluation. We will summarize the characteristic results for each evaluation, identifying the
features with significance and without significance.
After analyzing Likert questions and video performance metrics, some results matched
secondary research and some contradicted literature (See Section 2). However, the Company B
data set showed more significance in comparisons in this dataset than that of Company A. As
mentioned in Section 2.4.1.3 significance was defined as p-values less than 0.1.

Like the Company A case study results, the results from the yes and no comparison with
engaged viewers and views for the Company B case study had some video message content
features that were significant, and a few with positive and negative correlations. The comparison
with the most significant positive correlated result is whether humor is used. Some other yes and

43
no results to look at are ‘rational appeal’ with a positive correlation and ‘do they give an opinion’
as well with a positive correlation. Some yes and no results with significance and a negative
correlation are ‘Do they give cons’, ‘Influencer asked viewers questions’, and ‘Does it have a
hook’. The other yes and no results were not significant enough to draw a conclusion (see Figure
10).

Figure 10. Company B Engaged Viewers to Yes/No Question Relationship Graph

The next four subsections summarize some of the more significant video content features
and some analysis of their results from Company A and Company B’s data sets and evaluations.
The four features discussed include music appropriateness, video creativity, influencer
excitement, and emotional appeal. These content message characteristics are further discussed
within this section as data analyses indicated high significance for these features.

4.2.1 Music Appropriateness


The statistical summary of the video music appropriateness ranking by the team for the
Company A case study (see Table 9), shows a mean score of 2.75. There are no videos that were
scored 6 and 7 in the music appropriateness category, showing a skewness and bias by the team
scoring this factor. Authenticity is also presented Table 9 but is discussed in section 4.4.2. With
this result in mind, we acknowledge that the results of these features will be skewed—based on
the biases of the scoring team and the subjectivity of evaluation by the scoring approach needs to
be carefully monitored based on how many people and who is completing the scoring. Even a

44
scoring within a single video will show a distribution of agreement or disagreement, these values
are not included, but can be evaluated with a more nuanced evaluation of inter-video agreement
team scoring.

Music Appropriateness Authenticity


Mean 2.75 3.24
N/A* Count 13 0
N/A Percentage 39.39% 0.00%
Minimum 1 1
Q1 1.75 2
Med 3 4
Q3 4 4
Maximum 5 5
IQP 2.25 2
Mean-Median -0.25 0.24
Table 9. Numerical Calculations of Company A Music Appropriateness
*N/A: not applicable – feature was not present in the social media post

Due to the skew in the data, the comparison between music appropriateness and
engagement rate is likely to be biased. Figure 11 provides insight into this relationship. Figure 11
had a p-value < 0.05 with a correlation coefficient of 0.5052. The relationship graph in Figure 11
indicates more appropriate music has a lower video engagement rate. The result disputes our
hypothesis based on our literature review as Section 2.4.2.4 which mentions that if a video has
music that is irrelevant to the message and visuals, the less likely the viewer would enjoy the
video (Cunnane & Corcoran, 2018). We believe this is due to the small sample size and the
absence of videos with a 6 and 7 ranking of music appropriateness. It may also relate to
inappropriate music causing a humorous contrast that causes viewers to be more engaged. This
issue needs further investigation but given the biases of the scoring and data distribution, it
would be difficult to draw a strong conclusion and recommendation based on this finding.

45
Figure 11. Company A Music Appropriateness to Video Engagement Rate Relationship Graph.

Unlike the chart from Company A, the Music Appropriateness vs Views comparison
from Company B did match the secondary research which states the likelihood of the video
being shared is increased when the soundtrack is popular and fits the video. (Cunnane &
Corcoran, 2018). Figure 12 shows that more suitable video music related to the greater views it
receives. Although the p-value < 0.10 showed some significance, this graph could be more
accurate if there were more data since as displayed in Table 10, 30% of this feature is N/A which
is a significant amount of the dataset.

Music Appropriateness
Mean 4.28
N/A* Count 12
N/A Percentage 32.43%
Minimum 1
Q1 4
Med 5
Q3 5
Maximum 6
IQP 1
Mean-Median -0.72
Table 10. Numerical Calculations of Company B Music Appropriateness
*N/A: not applicable – feature was not present in the social media post

46
Figure 12. Company B Music Appropriateness to Views Relationship Graph.

4.2.2 Video Creativity


Similar to Music Appropriateness, the comparison between video creativity and views is
significant with a p-value < 0.10. Figure 13 shows the relationship that the more creative a video
the less likely a viewer would engage. Secondary research in Section 2.4.2.1 stated that viewers
enjoy videos that are more creative as they are intrigued with unique content (Lee & Hong,
2016). A possible reason for this counter-intuitive finding is the narrow coverage of the video of
levels of creativity.

Figure 13. Company A Video Creativity to Engaged Viewers Relationship Graph.

4.2.3 Influencer Excitement


Despite there being a narrow range of videos with different levels of influencer
excitement, the relationship between influencer excitement and engaged viewers is significant

47
with a p-value < 0.05. Figure 14 displays that the more excited an influencer is, the more
engaged viewers a video will receive. This aligns with the secondary research literature review
which states that certain linguistic styles can increase video performance, as referred to in
Section 2.4.2.2 (e.g. Alegro & Turnšek, 2021).

Figure 14. Company B Influencer Excitement to Engaged Viewers Relationship Graph.

4.2.4 Emotional Appeal


Another significant graph from Company B was Emotional Appeal vs Video Engagement
Rate with a p-value < 0.01. Figure 15 exhibits that the more emotions the video provokes, the
higher the percentage of viewers that would engage with the video, aligning with the secondary
background research found in Section 2.4.2.3 (e.g. Isabella et al., 2023). However, as mentioned
previously in Table 9, the average is low, and we can notice it since there are no videos with a 7
ranking and there is only one with an emotional appeal ranking of 6. This conveys that the
dataset does not accurately cover the whole range and is likely skewed.

48
Figure 15. Company B Emotional Appeal to Engaged Viewers Relationship Graph.

4.3 Video Creator Features and Virality


Unlike content message characteristics, no significant conclusions about creator features
could be made following data analysis. Only general characteristics of the video creator feature
items are presented here to provide the reader with some characteristics considered.

The distribution of some creator features in the dataset are now presented. A majority of
(81.82%) are male, as shown in Table 11. There were only two female influencers who posted an
advertisement video for Company A. Due to the limited data values, no conclusions could be
drawn. The differences in performance and trends of the videos by the two genders were then
compared. This was similarly the case for Company B (see Table 12), with most influencers
being female and only one influencer identifying as male.

Gender Frequency Proportion


Male 27 81.82%
N/A* 4 12.12%
Female 2 6.06%
Table 11. Gender Distribution for Company A Influencers
*N/A: not applicable – influencer gender could not be identified

Gender Frequency Proportion


Male 1 2.78%
Female 35 97.22%
Table 12. Gender Distribution for Company B Influencers

There were only four influencers whose top audience gender is female in Company A,
which is not significant enough to be considered. Consequently, the video characteristics of the
videos with female and male top audience gender were not compared. In Table 1, the top age
range of the audience is shown. Many of the influencers were not part of the WotoHub database,
thus the top audience age ranges for these influencers were unidentifiable. Therefore, around
40% of the influencers with N/A top age range were unable to conduct analyses using the top age
range.

49
Top Age Frequency Proportion
Range
18-24 18 54.55%
N/A 13 39.39%
25-34 1 3.03%
35-44 1 3.03%
Table 13. Top Audience Age Range Distribution for Company A Influencers

Similar to Company A, the distribution of the creator features of Company B is skewed


and many of the creators were missing from the WotoHub, resulting in inconclusive results.
Table 14 shows that there is a major imbalance as there is only one male influencer for Company
B.

Overall, recommendations cannot be made from the creator features of videos from either
Company A or Company B.

Gender Frequency Proportion


Female 33 89.19%
Female (account is 3 8.11%
about cats)
Male 1 2.7%
Table 14. Gender Distribution for Company B Influencers

4.4 Video Creator History and Virality


The case studies resulted in various video creator history features having a significant
correlation with different virality metrics. This section presents the creator history features that
had a p-value < 0.1. These features consist of trustworthiness and authenticity when being
compared to video engagement rate.

4.4.1 Trustworthiness
From Company A’s advertisement videos, Trustworthiness vs Video Engagement Rate,
exhibited in Figure 16, was one of the few relationships with high significance. The trend shows
that more credible creators actually result in lower video engagement rates. This graph has a p-
value < 0.05 and a correlation coefficient of 0.3859. This graph contradicts the research we had
previously reviewed (Chekima et al., 2020). Section 2.4.4 states that social media users are more
likely to find and watch videos from creators that are more trustworthy as they know that the

50
creators would be truthful. Some potential reasons for this may be due to the scorer bias (the
team) on trustworthiness (other viewers may not have the same viewpoint which means a larger
data set of viewers is needed that can score trustworthiness), the data sample of videos is small,
or that other confounding issues are arising such as negative trustworthiness in a video may have
some correlation with humor and humor may cause greater engagement. These are nuanced
issues that require further investigation.

Figure 16. Company A Trustworthiness to Video Engagement Rate relationship graph.

4.4.2 Authenticity
As previously displayed in Section 4.2.1 in Table 9, there was a skew in the data for
lower-rated videos in authenticity in Company A; Figure 17 displays a negative relationship
between authenticity and video engagement rate, meaning the more authentic the influencer, the
lower the video engagement rate would be.

Although the relationship is significant with a p-value < 0.10 with a correlation
coefficient of 0.3179, the dataset is skewed as mentioned in Section 2.4.4; there are no videos
representing the rank 6 and 7 of authentic creator. This finding differs from the literature review
as research specifies that videos of creators who are considered authentic would receive a higher
video performance (see Section 2.4.4.1). These results are quite tentative and will benefit from
more data and additional scorers.

51
Figure 17. Company A Authenticity to Video Engagement Rate relationship graph.

52
5. Conclusions & Recommendations
This project had findings from primary research—through the case study analyses and
social media user behavior surveys—and the secondary research. These findings provided
evidence to support the final deliverables and recommendations for the sponsor. This section
summarizes the sponsor deliverables. Recommendations to the sponsor (WotoKOL) and drawn
from the research and evidence are summarized. We then provide some reflections on, among
other things, challenges and limitations faced during the project. We then have a short conclusion
section that brings the IQP report to an end.

5.1 Deliverables
The team’s final deliverables to the sponsor consisted of two visually focused items: (1) a
business-focused final PowerPoint presentation summarizing the results and conclusions drawn
from the case study analysis, with supplemental details for the social media user survey to
support the conclusions of the case study; and (2) a mind map outlining the video features
characteristics introduced in Section 2.4. The mind map also explains how these features can be
used to positively increase virality for different categories of brand products.
The team’s major focus was to create a final product for the sponsor that was
“marketable”. A final PowerPoint presentation that could be directly used by our sponsor
company for their own presentations was this first deliverable product. Canva software was used
to create the PowerPoint presentation to highlight key points of the case study, in an aesthetically
pleasing manner.
Figures 19-21 are example slides of the final presentation. The deliverable to the sponsor
contained 37 slides with many visual and multimedia elements—included were images and video
clips taken directly from videos used in the case study. In addition to using more graphics and
images, our team opted to implement thumbs-up and thumbs-down icons to represent the results
of the case study analysis in a simple and effective way. This decision was made following
feedback from our sponsor, noting that many of the graphs taken from Excel and SPSS were too
difficult to read during a presentation and provided more detail than was necessary to make an
effective argument.

53
Figure 18. Final Presentation Slide Example

Figure 19. Humor in Videos Final Presentation Slide

Figure 20. Emotional Appeal in Videos Final Presentation Slide

During the middle of the project the sponsor was presented with case study initial
findings, the sponsor informed us that members of her team who were not involved in our project

54
would be attending. A table of defined terms was created—see Appendix G—for reference of the
terms used in the presentation.
The second deliverable was a mind map (see Figure 21). The mind map summarizes the
conclusions found in the case study analysis for video characteristics that positively affect video
virality and apply those concepts to different product categories. Per the sponsor’s request,
negative features were not included in the mind map. For future developments of this research,
negative features can be an area of future exploration.
In Figure 21, there are two main categories: creative videos and advertisement videos.
There are multiple subcategories to advertisement videos that refer to the industries that the
sponsor requested, as these are the WotoKOL industries of interest. Some industries were split
into further categories since they, based on our research, have different objectives and audiences.

Figure 21. Overall Mind Map of Which Video Characteristics Affect Virality Depending on Industry and Product
Category

55
The case studies conducted were in the technology—VR/AR and lifestyle industry
categories. Thus, we are most knowledgeable in these industries. Following our sponsor’s
request, the conclusions made were based on the primary and secondary research.
To use this mind map, the sponsor first needs to identify the type of video being posted.
If it is a video without a sponsor or paid advertisements, the sponsor would refer to the
characteristics on the right of the “creative video” bubble. These characteristics are ordered from
the most important to less necessary features. If the video contains an advertisement, the sponsor
will need to determine the industry of the product and follow the lines to the appropriate bubbles
to understand what characteristics fit that industry ad.

5.2 Recommendations
We recommend that influencers use humor during the paid promotional advertisement
parts of their videos. The evidence based on the case analyses approach, survey feedback, and
the literature show that videos with a higher engaged viewer count use humor within the paid
promotional advertisement sections of their videos. Influencers can benchmark emerging and
current issues that help to show humor. These are results are supported by background research
discussed in Section 2.4.2.3. In summary, here are some strategies influencers can take to
implement humor into their social media advertisement post: use authentic humor that fits in
with the influencer’s public image, use popular culture and currently trending references, and use
visual comedy (Patzer, 2023).
We recommend that influencers use emotional appeal during the paid promotional
advertisement parts of their videos. As seen in Figure 15, using emotional appeal during
promotional aspects of a video increases the overall performance of the video. This is also
supported in Table 4 from our linear regression model predicting engaged viewers on a particular
video, the model has a 92.8% adjusted R2 and emotional appeal is the number one predictor for
engaged viewers. This model also found a statistically significant relationship where the number
of engaged viewers is higher when the emotional appeal is 4, 5, or 6, as opposed to when
emotional appeal was rated 1, 2, or 3. These results are supported by secondary background
research discussed in Section 2.4.2.3. In summary, here are some strategies to influencers may
use to evoke emotional appeal in their social media advertisement posts: use empathy with

56
audience while promoting products, use sounds to evoke emotions within viewers, use personal
experiences to allow viewers to relate and create a personal connection to the video, and pinpoint
which emotion is best for the video content being created (Saif, 2019).
We recommend that further research be conducted on the following content message
characteristics: production quality, music appropriateness, video creativity, influencer
excitement, emotional appeal, trustworthiness, and authenticity. Data analysis on these content
message characteristics discussed on Section 4 suggest that these video features are significant
(p<0.1), however due to the lacking sample size of videos in the case study analyses, further
testing should be conducted.
We recommend that further research be conducted on hashtags. Table 5 and Section 4.1.3
suggest that hashtags related to general product categories have the most effect on the viewership
of a social media post. However, due to the skew in data with the lacking sample size of videos
for the case study analyses, the team suggests additional research be conducted to support these
findings.

5.3 Limitations
For this IQP Project, two case studies analyses, a survey, and secondary research were
conducted. Recommendations and deliverables were developed from analyzing and comparing
the results. The addition of a marketable PowerPoint presentation and mind map showcasing
video characteristics effects on different product categories were included in as part of a pack of
final deliverables. Although these goals were accomplished and the sponsor was satisfied, there
were limitations and challenges that our group needed to consider and overcome during our time
in Hangzhou, China. We reflect on these items.
When reflecting on the project and the timeline throughout the term, it became evident
that the project scope and direction changed frequently. During the preparation period prior to
IQP, we had meetings with the sponsor, outlining their expectations and WotoKOL’s resources.
They introduced WotoKOL’s database called the WotoHub to our team and claimed that a
spreadsheet associated with the database would be available for analysis. As it was supposedly
populated with data on videos from over 11 million creators, this analysis initially became the
project’s primary methodology. Upon arrival in China, the spreadsheet provided consisted of less
than 600 rows with limited features, most of showing minimal relevance to the project objective.

57
Adjustments to the methodologies were made given the situation and ultimately the
decision was made to remove the spreadsheet analysis method. The sponsor then introduced a
project that they believed would help create valuable and immediate deliverables. This new
project was the case study analysis on the products from two different companies. Knowing that
the sponsor prioritized the case study over the survey and interview and that this methodology
would require more time than the initial analysis of the WotoHub, the interview methodology
was removed as well.

After presenting initial findings, the sponsor then strongly suggested our team conduct
‘mini case studies’ for different industries. A group discussion then took place, and it was
decided this additional methodology did not fit well in our timeline. Furthermore, feedback on
our initial findings presentation shifted our team’s design choices to create presentation
deliverable that was more business-focused and marketable.

The second biggest limitation of the project involved time constraints. Like all other WPI
students, the IQP must be completed in eight weeks. With the shift in methodology in week one,
the team had to accommodate to the new project timing. Our sponsor then informed us that the
case studies were live case studies that would be used by influencers, meaning they would like
the two case studies to be completed by Black Friday (based on the U.S. holiday shopping
timing). As a result, our team had three weeks to conduct the research. To stay organized and
ensure analysis completion by the deadline, the project timeline was updated.

After video collection and spreadsheet population for the case study analysis, the team
was informed that the case study for Company A must be completed one week earlier than
previously discussed. The team was able to accommodate around this unexpected change of
deadline by prioritizing Company A’s case studies and completing the analysis before beginning
to analyze Company B. At the beginning of the term, the project timeline was created with extra
room in all areas; using Microsoft Project Management the project remained on track.
Additionally, the HDU students offered to analyze the survey responses, allowing more time to
draft the report. However, if given more time, the survey would have been able to be distributed
to a broader range of potential respondents, and additional video analysis for the case study
would have been conducted. Appendices H and I show the demographic distribution of survey

58
respondents, demonstrating this limitation through the skew towards the college aged and
educated portion of the population.

More videos would also be accounted for if viewable outside the office. The videos
utilized for the case studies are from YouTube, Instagram, and TikTok, each of which is banned
in China. To watch the videos from those three platforms, a VPN must be used. The team did not
have a reliable VPN to watch the videos outside the company. The days we went into the office
were Tuesday, Wednesday, and Thursday, so we had to adjust the timeline to accommodate the
population of the spreadsheet.

Another obstacle that was overcome was the language barrier between the HDU students
and our team. Our team’s Chinese language skills are at the beginner level; thus, our vocabulary
was not proficient enough to fully understand the HDU students. Similar to our Chinese level,
the HDU students’ English level were not advanced enough to fully understand the IQP team’s
language. There was miscommunication, and we had to use extra time to have a total grasp of the
meaning of the conversation to ensure that nothing was incorrectly translated. Fortunately, the
HDU students brought a friend with a proficient level of English to the initial meetings who
acted as a translator. This was very helpful and allowed the meeting to go more smoothly. As the
weeks went by, we began to understand each other and the others’ objectives and projects better,
so the friend was not necessary, and we changed methods of translation.

One of the reasons why the meetings were more difficult was that our sponsor informed
us that because the case studies were live, they would be considered confidential, meaning we
were not allowed to tell the advisor or the HDU students about the two companies’ upcoming
products, the reasons for the case studies, and other project details. Due to these reasons, the
HDU students would not be working on this portion of the project.

Another conflict we encountered with the HDU students is because we go into the office
during school hours and they have busy schedules with classes and extracurriculars, leaving them
unable to go into the office. There was one instance where they were able to make a trip to the
office; however, only one out of 11 HDU students attended. One of their project goals is
assessing general advertisement videos on YouTube, Instagram, and TikTok. However, like our
situation, they cannot view these platforms while in China unless they have the VPN that only
the company is allowed. Hence, if HDU students needed any help like getting access to videos,
59
we would need to allot our time to assist them. They were very kind and took over our analysis
for the surveys, so we were happy to help them.

The final general limitation is that our team does not have experience in marketing, thus
we were unfamiliar with the process in which they assess videos and present their findings.
Instead, we approached it in a way we were familiar with; we began by looking at the methods
and problems academically and analytically. Our sponsor provided helpful feedback during our
initial and milestone presentations to guide us in a direction to make our presentation more
business-focused and marketable. We were able to implement it, creating a more applicable and
practical PowerPoint.

5.4 Concluding Remarks

This IQP project sought to provide insights into four research questions: “What content
features relate to social media virality the most?”, “What content messages relate to social media
virality the most?”, “What creator features relate to social media virality the most?”, and “What
creator history characteristics relate to social media virality the most?”. We found insights for
each one, with some questions having greater and more direct insight than others, but each
provided additional information that can be used for sponsor recommendations.
Recommendations from the primary and secondary evidence highlight humor and
emotional appeal as leading video characteristics that help support social media virality the most.
A “marketable” final presentation and mind map were created to encapsulate the results of the
case study analyses and survey; these were deliverables sought by the sponsor.
Despite the various challenges encountered throughout the IQP process, our team was
able to overcome them and successfully complete the project. Through this experience, the team
was able to gain exposure to practical business settings and partake in the real-life application of
the concepts and theories found through our secondary research and from classroom instruction.
With the obstacles our team faced, we learned to work through and overcome challenges as a
team, strengthening communication and project management skills. Through the perseverance of
our team, project goals were met, and the sponsor was satisfied.

60
6. References
About us. (n.d.). WotoKOL. Retrieved September 19, 2023, from
https://global.wotokol.com/elementor-1849.html
Alegro, T., & Turnšek, M. (2021). Striving to Be Different but Becoming the Same: Creativity
and Destination Brands’ Promotional Videos. Sustainability, 13(1), 139.
https://doi.org/10.3390/su13010139
Celestino, P. (2023). Council Post: Influencer Marketing In 2023: Benefits And Best Practices.
Forbes. https://www.forbes.com/sites/forbesagencycouncil/2023/03/10/influencer-
marketing-in-2023-benefits-and-best-practices/
Chekima, B., Chekima, F. Z., & Adis, A.-A. A. (2020). Social Media Influencer in Advertising:
The Role of Attractiveness, Expertise and Trustworthiness (SSRN Scholarly Paper
3739287). https://papers.ssrn.com/abstract=3739287
Chowdhry, A. (2018, September 18). Study: Relevant Video Content Drives More Engagement
And Revenue. Forbes. https://www.forbes.com/sites/amitchowdhry/2018/09/18/study-
relevant-video-content-drives-more-engagement-and-revenue/
Cunnane, P. V., & Corcoran, D. N. (2018). ECSM 2018 5th European Conference on Social
Media. Academic Conferences and publishing limited.
Dean, B. (2023, March 27). TikTok User Statistics (2023). Backlinko.
https://backlinko.com/tiktok-users
Definition of CON. (n.d.). Retrieved November 21, 2023, from https://www.merriam-
webster.com/dictionary/con
Dobele, A., Lindgreen, A., Beverland, M., Vanhamme, J., & van Wijk, R. (2007). Why pass on
viral messages? Because they connect emotionally. Business Horizons, 50(4), 291–304.
https://doi.org/10.1016/j.bushor.2007.01.004
Doman. (2015). Basic rules of good videography | UVic Video resources.
https://onlineacademiccommunity.uvic.ca/videos/2015/09/24/basic-rules-of-good-
videography/
Fisher, R. A. (1956). Statistical Methods for Research Workers. In S. Kotz & N. L. Johnson
(Eds.), Breakthroughs in Statistics: Methodology and Distribution (pp. 66–70). Springer.
https://doi.org/10.1007/978-1-4612-4380-9_6

61
Foster, D. (2020). Factors influencing the popularity of YouTube videos and users’ decisions to
watch them [University of Wolverhampton].
https://wlv.openrepository.com/bitstream/handle/2436/623742/Foster_Phd_thesis_Redact
ed.pdf
Ge, J., & Guo, J. (2022). The Synergistic Effect of Sales Discount and Mobile Advertising: How
KOL Influence Online Education Community Purchases. In G. Salvendy & J. Wei (Eds.),
Design, Operation and Evaluation of Mobile Communications (pp. 246–255). Springer
International Publishing. https://doi.org/10.1007/978-3-031-05014-5_20
Glover, R. (2023, September 18). How the YouTube Algorithm Works in 2023 (+14 Tips for
More Views). WordStream. https://www.wordstream.com/blog/ws/2023/09/15/youtube-
algorithm
Haenlein, M., Anadol, E., Farnsworth, T., Hugo, H., Hunichen, J., & Welte, D. (2020).
Navigating the New Era of Influencer Marketing: How to be Successful on Instagram,
TikTok, & Co. California Management Review, 63(1), 5–25.
https://doi.org/10.1177/0008125620958166
Han, Y., Lappas, T., & Sabnis, G. (2020). The Importance of Interactions Between Content
Characteristics and Creator Characteristics for Studying Virality in Social Media.
Information Systems Research, 31(2), 576–588. https://doi.org/10.1287/isre.2019.0903
Hangzhou Wotokol Network Technology Co., Ltd | LinkedIn. (n.d.). Retrieved September 19,
2023, from https://cn.linkedin.com/company/hangzhou-wotokol-network-technology-co-
ltd
Hill, C. (2022, October 11). What is TikTok: The complete platform guide for 2023. Sprout
Social. https://sproutsocial.com/insights/what-is-tiktok/
influencer noun - Definition, pictures, pronunciation and usage notes | Oxford Advanced
Learner’s Dictionary at OxfordLearnersDictionaries.com. (n.d.). Retrieved November
21, 2023, from https://www.oxfordlearnersdictionaries.com/definition/english/influencer
Isabella, G., Melo, A. F. de, & Gonzalez, M. C. (2023). Going Viral on Advertising YouTube
Video: Detecting the Influences. Revista de Administração Contemporânea, 27, e220300.
https://doi.org/10.1590/1982-7849rac2023220300.en

62
Joshi, A., Kale, S., Chandel, S., & Pal, D. (2015). Likert Scale: Explored and Explained. British
Journal of Applied Science & Technology, 7(4), 396–403.
https://doi.org/10.9734/BJAST/2015/14975
Kiros, H. (2022, September 20). Hated that video? YouTube’s algorithm might push you another
just like it. MIT Technology Review.
https://www.technologyreview.com/2022/09/20/1059709/youtube-algorithm-
recommendations/
Kowalczyk, C. M., & Pounders, K. R. (2016). Transforming celebrities through social media: the
role of authenticity and emotional attachment. Journal of Product & Brand Management,
25(4), 345–356. https://doi.org/10.1108/JPBM-09-2015-0969
Lead Generation. (2022, August 23). The Importance of Using #Hashtags in Social Media

Marketing. https://www.linkedin.com/pulse/importance-using-hashtags-social-media-

marketing-lead-generation

Lee, J., & Hong, I. (2016). Predicting positive user responses to social media advertising: The
roles of emotional appeal, informativeness, and creativity. International Journal of
Information Management, 36(3), 360–373.
https://doi.org/10.1016/j.ijinfomgt.2016.01.001
Lee, M. T., & Theokary, C. (2021). The superstar social media influencer: Exploiting linguistic
style and emotional contagion over content? Journal of Business Research, 132, 860–
871. https://doi.org/10.1016/j.jbusres.2020.11.014
Leighton, M. (2022, December 30). A TikToker with science Ph.D. begged viewers to stop asking
a man to “explain” things she’s already explained. Insider.
https://www.insider.com/tiktok-female-science-tiktoker-viewers-tagging-hank-green-
2022-12
Leonard, M., & Kurniawan, M. (2023). A beginner’s guide to video resolution. Adobe.
https://www.adobe.com/creativecloud/video/discover/video-resolution.html
Ling, C., Blackburn, J., De Cristofaro, E., & Stringhini, G. (2022). Slapping Cats, Bopping
Heads, and Oreo Shakes: Understanding Indicators of Virality in TikTok Short Videos.
14th ACM Web Science Conference 2022, 164–173.
https://doi.org/10.1145/3501247.3531551

63
Lourenço, C. J. S., Isabella, G., Verbeke, W., Vo, K., Dimoka, A., & Bagozzi, R. P. (2023). How
songs from growing up and viewers’ attachment styles affect video ads’ effectiveness.
Psychology & Marketing, 40(1), 209–233. https://doi.org/10.1002/mar.21725
Masuda, H., Han, S. H., & Lee, J. (2022). Impacts of influencer attributes on purchase intentions
in social media influencer marketing: Mediating roles of characterizations. Technological
Forecasting and Social Change, 174, 121246.
https://doi.org/10.1016/j.techfore.2021.121246
Moon, S., & Yoo, S. (2022). Are More Followers Always Better? The Non-Linear Relationship
between the Number of Followers and User Engagement on Seeded Marketing
Campaigns in Instagram. Asia Marketing Journal, 24(2), 62–77.
https://doi.org/10.53728/2765-6500.1589
Moss, S. (2023, September 12). 38 Instagram Statistics For 2023: The Definitive List. Startup
Bonsai. https://startupbonsai.com/instagram-marketing-statistics/
Mosseri, A. (2023, May 31). Instagram Ranking Explained.
https://about.instagram.com/blog/announcements/instagram-ranking-explained/
Omar, K. (2023). How to Choose the Best Influencer For Your Brand. Lightspeed. Retrieved
January 20, 2024, from https://www.lightspeedhq.com/blog/how-to-find-an-influencer/
Park, C. (2023, September 6). Complete guide to Instagram content types | Learn at Microsoft
Create. https://create.microsoft.com/en-us/learn/articles/complete-guide-instagram-
content-types
Patzer, T. (2023, August 7). Make Them Laugh: How Humor Can Elevate Your Social Media
Presence. ILLUMINATION. https://medium.com/illumination/make-them-laugh-how-
humor-can-elevate-your-social-media-presence-ca00260f5423
Reid, T. (n.d.). YouTube for Press. Blog.Youtube. Retrieved September 19, 2023, from
https://blog.youtube/press/
Saif, M. (2019, October 14). How to Use Video to Trigger Customer Emotions. MotionCue.
https://motioncue.com/how-to-use-video-to-trigger-customer-emotions/
Segev, S., & Fernandes, J. (2023). The Anatomy of Viral Advertising: A Content Analysis of
Viral Advertising from the Elaboration Likelihood Model Perspective. Journal of
Promotion Management, 29(1), 125–154.
https://doi.org/10.1080/10496491.2022.2108189

64
Seo, Y., Kim, J., Choi, Y. K., & Li, X. (2019). In “likes” we trust: likes, disclosures and firm
serving motives on social media. European Journal of Marketing, 53(10), 2173–2192.
https://doi.org/10.1108/EJM-11-2017-0883
Sherman, T. (2021, October 16). One year of YouTube Shorts — what we’ve learned so far.
Blog.Youtube. https://blog.youtube/news-and-events/one-year-of-youtube-shorts-what-
weve-learned-so-far/
Sohaib, M., Safeer, A. A., & Majeed, A. (2022). Role of social media marketing activities in
China’s e-commerce industry: A stimulus organism response theory context. Frontiers in
Psychology, 13. https://www.frontiersin.org/articles/10.3389/fpsyg.2022.941058
Tanwar, A. S., Chaudhry, H., & Srivastava, M. K. (2022). Trends in Influencer Marketing: A
Review and Bibliometric Analysis. Journal of Interactive Advertising, 22(1), 1–27.
https://doi.org/10.1080/15252019.2021.2007822
Tellis, G. J., MacInnis, D. J., Tirunillai, S., & Zhang, Y. (2019). What Drives Virality (Sharing)
of Online Digital Content? The Critical Role of Information, Emotion, and Brand
Prominence. Journal of Marketing, 83(4), 1–20.
https://doi.org/10.1177/0022242919841034
The Importance of Using #Hashtags in Social Media Marketing. (2022, August 23).
https://www.linkedin.com/pulse/importance-using-hashtags-social-media-marketing-
lead-generation
Thelwall, M., Stuart, E., Mas-Bleda, A., Makita, M., & Abdoli, M. (2022). I’m Nervous about
Sharing This Secret with You: Youtube Influencers Generate Strong Parasocial
Interactions by Discussing Personal Issues. Journal of Data and Information Science,
7(2), 31–56. https://doi.org/10.2478/jdis-2022-0011
TikTok CEO Shou Chew on Its Future — and What Makes Its Algorithm Different | Live at
TED2023. (2023, April 21). TED. https://www.youtube.com/watch?v=7zC8-06198g
Tribbey, C. (2016, June 29). Verizon: Online Viewers Expect TV-Quality Video. Broadcasting
Cable. https://www.nexttv.com/news/verizon-online-viewers-expect-tv-quality-video-
157690
Tucker, C. E. (2015). The Reach and Persuasiveness of Viral Video Ads. Marketing Science,
34(2), 281–296. https://doi.org/10.1287/mksc.2014.0874

65
Video Marketing Statistics | New Data For 2023. (n.d.). Wyzowl. Retrieved October 5, 2023,
from https://www.wyzowl.com/video-marketing-statistics/
Wei, Y., Cheng, Z., Yu, X., Zhao, Z., Zhu, L., & Nie, L. (2019). Personalized Hashtag
Recommendation for Micro-videos. Proceedings of the 27th ACM International
Conference on Multimedia, 1446–1454. https://doi.org/10.1145/3343031.3350858
West, C. (2021, October 18). 7 Video metrics you need to know to measure success. Sprout
Social. https://sproutsocial.com/insights/video-metrics/
What Does FYP Mean, And Why Is It All Over TikTok? (2020, May 20). Dictionary.Com.
https://www.dictionary.com/e/acronyms/fyp/
WotoKOL. (n.d.-a). WotoKOL. Retrieved September 19, 2023, from https://global.wotokol.com/
Yang, H. (2013). The Case for Being Automatic: Introducing the Automatic Linear Modeling
(LINEAR) Procedure in SPSS Statistics. Multiple Linear Regression Viewpoints, 39.
https://www.statwks.com/wp-content/uploads/2019/03/Introducing-the-Automatic-
Linear-Modeling.pdf
YouTube Statistics 2023 [Users by Country + Demographics]. (2023, August 17). Official GMI
Blog. https://www.globalmediainsight.com/blog/youtube-users-statistics/

66
Appendices
Appendix A: Social Media User Survey Preamble and Confidentiality Notice

67
Appendix B: Social Media User Survey

68
69
70
71
72
73
74
Appendix C: Analysis Checklist

75
76
Appendix D: Data Field Structure of Case Study Database by “Type” and
“Measure” fields
Field Name Measure Type
Is YouTube (dummy of Platform) Nominal Numeric
Is Instagram (dummy of Platform) Nominal Numeric
Is TikTok (dummy of Platform) Nominal Numeric
Platform Nominal String
Title Nominal String
Date of Creation Scale Date
View Count Scale Numeric
Like Count Scale Numeric
Comment Count Scale Numeric
Share Count Scale Numeric
Is German (Dummy of Language in Video) Nominal Numeric
Is Arabic (Dummy of Language in Video) Nominal Numeric
Is Chinese (Dummy of Language in Video) Nominal Numeric
Is French (Dummy of Language in Video) Nominal Numeric
Is Japanese (Dummy of Language in Video) Nominal Numeric
Is Spanish (Dummy of Language in Video) Nominal Numeric
Is English (Dummy of Language in Video) Nominal Numeric
Language(s) in Video Nominal String
No Hashtags (Dummy of Hashtags) Nominal Numeric
Redacted Hashtag 1 (Dummy of Hashtags) Nominal Numeric
Redacted Hashtag 2 (Dummy of Hashtags) Nominal Numeric
Redacted Hashtag 3 (Dummy of Hashtags) Nominal Numeric
#SmartGlasses (Dummy of Hashtags) Nominal Numeric
#XR (Dummy of Hashtags) Nominal Numeric
#AR (Dummy of Hashtags) Nominal Numeric
#augmentedreality (Dummy of Hashtags) Nominal Numeric
#tech (Dummy of Hashtags) Nominal Numeric
#smartglass (Dummy of Hashtags) Nominal Numeric
Redacted Hashtag 4 (Dummy of Hashtags) Nominal Numeric
#PR (Dummy of Hashtags) Nominal Numeric
#clean (Dummy of Hashtags) Nominal Numeric
Redacted Hashtag 5 (Dummy of Hashtags) Nominal Numeric
#aspirateur (Dummy of Hashtags) Nominal Numeric
Redacted Hashtag 6 (Dummy of Hashtags) Nominal Numeric
#vacuumcleaner (Dummy of Hashtags) Nominal Numeric
#cleaningasmr (Dummy of Hashtags) Nominal Numeric
#ASMR (Dummy of Hashtags) Nominal Numeric

77
#cleaningmotivation (Dummy of Hashtags) Nominal Numeric
Redacted Hashtag 7 (Dummy of Hashtags) Nominal Numeric
#cleanwithme (Dummy of Hashtags) Nominal Numeric
#cleantiktok (Dummy of Hashtags) Nominal Numeric
#cleantok (Dummy of Hashtags) Nominal Numeric
#nettoyage (Dummy of Hashtags) Nominal Numeric
#menage (Dummy of Hashtags) Nominal Numeric
Redacted Hashtag 8 (Dummy of Hashtags) Nominal Numeric
Redacted Hashtag 9 (Dummy of Hashtags) Nominal Numeric
Hashtags Nominal String
Captions Nominal String
Time Length Scale Date
2160p4k (Dummy of Highest Possible Resolution) Nominal Numeric
2160p60 4k (Dummy of Highest Possible Resolution) Nominal Numeric
1080p HD (Dummy of Highest Possible Resolution) Nominal Numeric
1080p Premium HD (Dummy of Highest Possible Nominal Numeric
Resolution)
1080p60 (Dummy of Highest Possible Resolution) Nominal Numeric
2160p (Dummy of Highest Possible Resolution) Nominal Numeric
Highest Possible Resolution Nominal String
Music in the background? (y/n) Nominal Numeric
Influencer face shown? (y/n) Nominal Numeric
Does it have a hook? (y/n) Nominal Numeric
Is humor used? (y/n) Nominal Numeric
Rational Appeal (y/n) Nominal Numeric
Do they give cons about the product? (y/n) Nominal Numeric
Do they give opinions? (y/n) Nominal Numeric
Influencer asked viewers questions? (y/n) Nominal Numeric
Video Creativity (1-7) Scale Numeric
Music Appropriateness (1-7) Scale Numeric
Setting Appropriateness (1-7) Scale Numeric
Production Quality (1-7) Scale Numeric
Production Expertise (1-7) Scale Numeric
Influencer invited comment/feedback? (y/n) Nominal Numeric
Influencer Excitement (1-7) Scale Numeric
Influencer Passion (1-7) Scale Numeric
Emotional Appeal (1-7) Scale Numeric
Account name Nominal String
Race Nominal String
Nationality Nominal String

78
Male (Dummy of Gender) Nominal Numeric
Female (Dummy of Gender) Nominal Numeric
Gender Nominal String
Number of Followers Scale Numeric
Number of Posts Scale Numeric
WotoHub score Scale Numeric
Average Like Scale Numeric
Average views (in the last 60 days) Scale Numeric
Top audience region Nominal String
Top gender Nominal String
Top age range Nominal String
Average Creator engagement rate Scale Numeric
Commercial content engagement rate Scale Numeric
Attractiveness (1-7) Scale Numeric
Trustworthiness (1-7) Scale Numeric
Authenticity (1-7) Scale Numeric
Expertise (1-7) Scale Numeric
Video Engagement Rate Scale Numeric
Engaged Viewers Scale Numeric

79
Appendix E: Depicts the independent variables used in the linear regression
models.
Field Name Measure Type
Platform Dummies Nominal Numeric
Date of Creation Scale Date
Language(s) in Video Dummies Nominal Numeric
Hashtags Dummies Nominal Numeric
Time Length Scale Date
Highest Possible Resolution Nominal Numeric
Music in the background? (y/n) Nominal Numeric
Influencer face shown? (y/n) Nominal Numeric
Does it have a hook? (y/n) Nominal Numeric
Is humor used? (y/n) Nominal Numeric
Rational Appeal (y/n) Nominal Numeric
Do they give cons about the product? (y/n) Nominal Numeric
Do they give opinions? (y/n) Nominal Numeric
Influencer asked viewers questions? (y/n) Nominal Numeric
Video Creativity (1-7) Scale Numeric
Music Appropriateness (1-7) Scale Numeric
Setting Appropriateness (1-7) Scale Numeric
Production Quality (1-7) Scale Numeric
Production Expertise (1-7) Scale Numeric
Influencer invited comment/feedback? (y/n) Nominal Numeric
Influencer Excitement (1-7) Scale Numeric
Influencer Passion (1-7) Scale Numeric
Emotional Appeal (1-7) Scale Numeric
Race Nominal String
Nationality Nominal String
Gender Dummies Nominal Numeric
Number of Followers Scale Numeric
Number of Posts Scale Numeric
WotoHub score Scale Numeric
Average Like Scale Numeric
Average views (in the last 60 days) Scale Numeric
Top audience region Nominal String
Top gender Nominal String
Top age range Nominal String
Average Creator engagement rate Scale Numeric

80
Commercial content engagement rate Scale Numeric
Attractiveness (1-7) Scale Numeric
Trustworthiness (1-7) Scale Numeric
Authenticity (1-7) Scale Numeric
Expertise (1-7) Scale Numeric

81
Appendix F: Summary statistics of each of the survey questions
No. Question Mean S.D. N
1 I enjoy comedy videos. 4.31 0.765 61
2 I enjoy sports videos. 2.87 1.284 61
3 I enjoy travel and events videos. 3.31 1.148 61
4 I enjoy gaming videos. 3.34 1.58 61
5 I enjoy lifestyle videos. 3.3 1.283 61
6 I enjoy auto and vehicle videos. 2.34 1.237 61
7 I enjoy entertainment videos 4.16 0.757 61
8 I enjoy education videos. 3.82 0.879 61
9 I enjoy beauty and cosmetics videos. 2.63 1.402 60
10 I enjoy political videos. 2.43 1.204 61
11 I enjoy economics videos. 2.54 1.205 61
12 I enjoy religious videos. 1.74 1.047 61
Summary statistics data for the first section of Likert scale questions in the social media user
survey.

No. Question Mean S.D. N


1 I find new social media content through a “For You Page.” 4.43 .741 61
2 I use hashtags to find new social media content. 2.07 1.031 61
3 I find new social media content through the accounts I 3.89 1.112 61
subscribed to.
4 I use the search bar to find new social media content. 3.66 1.196 61
5 I find new social media content through friends. 3.85 1.005 60
6 I find new social media content through family. 2.69 1.298 61
7 I find new social media content through the accounts I 3.93 1.031 61
follow.
Summary statistics data for the second section of Likert scale questions in the social media user
survey.

No. Question Mean S.D. N


1 Lower video quality makes me less likely to watch a video. 3.67 1.1 60
2 I like to watch videos with influencers who look like me. 2.93 1.219 60
3 I enjoy watching videos with familiar trends. 3.38 1.059 60
4 I enjoy watching videos that are at most 60 seconds. 2.98 1.172 60
5 I will watch an influencer who has values I don’t agree 2.68 1.097 60
with.
6 I enjoy watching videos from influencers that I am 3.28 .993 60
unfamiliar with.
7 I enjoy watching video genres that are unfamiliar to me. 3.13 1.033 60
8 I am more likely to click on a video with a high view count. 3.82 1.033 60

82
9 I don’t pay attention to hashtags in videos. 4.08 .979 60
10 I enjoy watching videos with opposing views. 2.85 1.071 60
11 I enjoy watching content that evokes strong emotions. 3.23 1.079 60
12 I seek out content that I find humorous. 4.2 .805 59
13 I stay away from content that has unverified information 3.28 1.035 61
14 I am more likely to watch an influencer that posts 2.16 0.778 61
unverified information.
15 I am more likely to watch an influencer with a bigger 3.43 .921 61
following.
16 I like to watch recently uploaded content. 3.95 0.902 61
17 I tend to stay away from influencers with fewer posted 2.43 0.921 61
videos.
18 I am more likely to watch influencers with high like count. 3.21 1.018 61
19 I am more likely to watch content from influencers that I 3.26 1.079 61
find attractive.
20 I seek out content from influencers who I think are good 4.08 .843 61
people.
21 I seek out content from influencers who I think are not good 2.1 .889 61
people.
22 I am more likely to watch content from an influencer if I 3.95 .973 61
think they are a credible source.
23 I am more likely to watch a video from an influencer with a 3.62 .969 61
wider variety of videos.
Summary statistics data for the third section of Likert scale questions in the social media user
survey

83
Appendix G: Table of Defined Terms and Acronyms
Table of Acronyms

Term Definition Citation Feature Determination


Type Method
Viral Video Describes a social media post reaching (Han, Lappas,
unusually high unique views, likes, and Sabnis, 2020)
shares.
Content Feature Objective characteristics of a post that (Han et al., CoF
(CoF) can be readily discerned by a viewer 2020)
instantly. This includes characteristics
such as the colors used in the video, the
images, hashtags, and video and audio
quality, etc.
Content Characteristics of a video describe the (Han et al., CM
Message (CM) features that viewers have to read or 2020)
watch to comprehend. Characteristics
such as affective cues, opinions, humor,
and facts are a few examples of
subjective characteristics that require
more cognitive effort to comprehend.
Creator Feature Describe the tangible features that are (Han et al., CrF
(CrF) evident to the viewer at the first glance. 2020)
This includes Numerical measurements
such as the average number of likes,
followers, and posts. Additionally, the
creator's demographic details, such as
their gender, race, and age also fall into
this category.
Creator History Traits that are known by their viewers (Han et al., CH
(CH) but are something that is not known at 2020)
first glance. This includes a creator's
credibility on a subject, topic propensity,
reputation, and a creator's attractiveness.
Immediate The feature can be determined Self-Defined IO
Observation immediately without watching the video.
(IO)
Watch Video A single person watched the video to Self-Defined WV
(WV) determine this feature.
Group Multiple members of the team watched Self-Defined GA
Assessment the video and agreed on a value for this
(GA) feature.
WotoHub (WH) This feature was taken directly from the Self-Defined WH
WotoHub.

Table of Definitions
Term Definition Citation Feature Determination
Type Method

84
Title The text that is found directly below Self-Defined CoF IO
videos found on YouTube. (YouTube
only)

Date of The date in which the video was made Self-Defined CoF IO
creation public
View count The number of times the video has Self-Defined CoF IO
been watched
Like count The number of users that have liked the Self-Defined CoF IO
video
Comment The numbers of comments found on a Self-Defined CoF IO
count video
Share count The number of times a video has been Self-Defined CoF IO
shared (Data only publicly available on
TikTok and Instagram Reels)
Language(s) in The language(s) that are spoken or in Self-Defined CoF IO/WV
video captions in the video
Hashtags The hashtags that are used in a video’s Self-Defined CoF IO
caption (Wei et al.,
2019)
“They may consist of any arbitrary
combination of characters led by a hash
symbol ‘#’ (e.g. #Puppy
and #thegoodlife).
Captions A brief description of the video found Self-Defined CoF IO
below the video
English An English translation of the video’s Self-Defined CoF IO
Caption caption.
Time length The exact duration of the video Self-Defined CoF IO
(Tellis,
Macinnis,
Tirunillai,
Zhang, 2019)**
Vertical vs Whether the video is filmed vertically, Self-Defined CoF IO
horizontal or filmed horizontally
Highest The highest quality resolution available Self-Defined CoF IO
possible on a YouTube video
resolution
Music in the Whether or not the video contains Self-Defined CoF WV
background? music at any point.
(y/n)
Influencer A person with the ability to influence (Influencer
potential buyers of a product or service Noun -
by promoting or recommending the Definition,
items on social media Pictures,
Pronunciation
and Usage
Notes | Oxford
Advanced
Learner’s
Dictionary at

85
OxfordLearners
Dictionaries.Co
m, n.d.)
Influencer face Whether or not the influencer of a Self-Defined CM WV
shown? (y/n) video shows their face
Hook “Opening section of your video that (6 Tips for
captures your audience's attention and Creating an
encourages them to keep watching.” Attention-
Grabbing Video
Hook, 2023)
Does it have a Whether or not the video has a hook in Self-Defined CM WV
hook? (y/n) the video
Humor The quality of being amusing or comic, (Influencer
especially as expressed in literature or Noun -
speech Definition,
Pictures,
Pronunciation
and Usage
Notes | Oxford
Advanced
Learner’s
Dictionary at
OxfordLearners
Dictionaries.Co
m, n.d.)

Is humor used? Whether or not humor is contained at Self-defined CM WV


(y/n) least once throughout the video.
Money- The ad emphasizes saving money by (Segev & CM
Off/savings using coupons, premiums, or Fernandes,
appeal promotions or the low price of the 2022)
brand/product
Health/well- The ad emphasizes a sense of well- (Segev & CM
being appeal being, typically by emphasizing the Fernandes,
health benefits of using the advertised 2022)
brand/product or the healthy
features/characteristics of the
brand/product.
Social The ad emphasizes the socially (Segev & CM
responsibility responsible characteristics of the Fernandes,
appeal corporate sponsor. The company is 2022)
committed to social welfare or a cause.
The ad promotes the image of a
company as “doing something good for
others” (society in general or groups in
society). The ad emphasizes that it will
allocate a percentage/amount of a sale
to purchase or fund a social cause
(cause subsidy).

86
Comparison The ad presents a comparison between (Segev & CM
with alternatives the advertised brand with an equivalent Fernandes,
appeal brand/competitor. 2022)

Product The ad conveys that the advertised (Segev & CM


superiority brand is the best, number one, superior Fernandes,
appeal but no comparison is made with 2022)
alternative/ equivalent brands.
Attributes and The ad communicates specific features, (Segev & CM
benefits appeal components or characteristics of the Fernandes,
brand/product and/or its 2022)
benefits/advantages for consumers.
Rational Whether or not the video contains at Self-Defined CM WV
Appeal (y/n) least one of the following appeals: (Segev &
Money-Off/savings Appeal, Fernandes,
Health/well-being Appeal, Social 2022)
responsibility Appeal, Comparison
with alternatives Appeal, Product
superiority Appeal, Attributes and
benefits Appeal
Con An argument or evidence in opposition (Definition of
CON, n.d.)

Do they give The influencer provides downsides or Self-Defined CM WV


cons about the negatives about the product.
product? (y/n)

Do they give The influencer provides their personal Self-Defined CM WV


opinions? (y/n) opinion and views about the product.
Influencer The influencer prompted the viewers Self-Defined CM WV
asked viewers with questions in their video for the
questions? viewers to reply to.
(y/n)
Video How creative the concept of the video (Lee & CM GA
Creativity (1-7) is. Theokary,
2021)**
Music How appropriate the music in the video (Lee & CM GA
Appropriatene is. Theokary,
ss (1-7) 2021)**
Setting How appropriate the background of the (Lee & CM GA
Appropriatene video is. Theokary,
ss (1-7) 2021)**
Production The overall quality of a video including (Lee & CF CM GA
Quality (1-7) cinematography, lighting, editing, and Theokary,
video resolution* 2021)
Production Production expertise is the average (Lee & CF CM GA*
Expertise (1-7) calculated average of Video Creative, Theokary,
Appropriate Music, Appropriate 2021)
Setting, and Production Quality

87
Influencer Influencer directly asked viewers to Self-Defined CM WV
invited comment or give them feedback within
comment/feedb their video.
ack? (y/n)
Influencer Influencer showing strong positive (Lee & CM GA
Excitement (1- emotion toward the audience. Theokary,
7) 2021)**
Influencer Influencer showing strong emotion (Lee & CM GA
Passion (1-7) toward the product Theokary,
2021)**
Emotional Message-based peripheral cues that (Segev & CM GA
Appeal (1-7) evoke emotions, usually positive Fernandes,
toward the advertised object 2022)
Account name The username of the creators for the Self-Defined CrF IO
given social media platform
Race The race of the influencer Self-Defined CrF
Nationality The country in which the influencer Self-Defined CrF
resides in
Age The age of the influencer Self-Defined CrF
Gender The gender of the influencer (limited to Self-Defined CrF IO
male and female)
Number of The number of followers (or Self-Defined CrF IO
followers subscribers in YouTube) that the
influencer has on a given social media
platform
Number of The number of posts that the influencer Self-Defined CrF IO/WH
posts has on a given social media platform
WotoHub The rating of the creator by WotoKOL WotoHub CrF WH
score calculated using six key indicators Definition
(Health Score, Activity Score, Growth
Score, Business Score, Intimacy Score,
and Intent Score)

Health Score The coverage rate of videos on WotoHub WH


followers reflects the overall influence Definition
of the channel
Activity Score The publishing situation of influencers WotoHub WH
reflects the channel’s activity level Definition
Growth Score The growth trend of influencer’s WotoHub WH
followers reflects the speed and Definition
potential of their follower growth
Business Score Analyzing the influencer’s past WotoHub WH
business collaborations showcases their Definition
commercial performance and value
Intimacy Score The engagement rate of followers WotoHub WH
reflects the level of interaction and Definition
intimacy between the influencer and
their followers

88
Intent Score The level of interest in business WotoHub WH
collaborations reflects the influencer’s Definition
attitude towards cooperation
Average Like The average number of likes of all the WotoHub CrF WH
creator’s videos Definition
Average views The average number of views per day WotoHub CrF WH
(in last 60 in the past 60 days on the given social Definition
days) media platform
Top audience The country that interacts with the WotoHub CrF WH
region creator’s the most Definition
Top gender The top gender of the creator’s WotoHub CrF WH
audience (limited to male and female) Definition
Top age range The top age range of the creator’s WotoHub CrF WH
audience Definition
Average The average engagement rate across all WotoHub CrF WH
Creator of a creator’s videos Definition
engagement
rate
Commercial The engagement rate of a creator’s WotoHub CrF WH
content sponsored videos Definition
engagement
rate
Topic The topic of the videos on the Self-Defined CH WV
propensity influencer’s account
Similarity The supposed resemblance between the (Chekima,
receiver (social media followers) and Chekima, and
source Adis, 2020)
Familiarity familiarity is measured knowledge of (Chekima,
the source based on their exposure Chekima, and
Adis, 2020)
Likeability likeability is liking for the source as a (Chekima,
result of the source’s physical facial Chekima, and
beauty and behavior Adis, 2020)
Attractiveness Attractiveness largely depends on the (Chekima, CH GA
(1-7) source’s Chekima, and
similarity, familiarity, and likeability to Adis, 2020)
the receiver
Trustworthines Trustworthiness is characterized as the (Chekima, CH GA
s (1-7) apparent eagerness of the source of Chekima, and
(Previously influencer to make lawful statements in Adis, 2020)
“Credible”) the perspective on the consumer or
followers and gives truthful and honest
information about a product
Authenticity Authenticity is defined as a post or (Kowalczyk & CH GA
(1-7) photo that demonstrates some aspect of Pounders,
(Previously the celebrity’s true self 2016)
“Genuine”)
Expertise (1-7) The endorser is viewed as capable (Chekima, CH GA
(Previously enough to convey legitimate and exact Chekima, and
“Qualified”) information or conscious of a specific Adis, 2020)
subject

89
Video The percentage of all the viewers who Self-Defined CF IO*
Engagement liked, shared, and/or commented.
Rate
Engaged The sum of the number of likes, Self-Defined CF IO*
Viewers comments, and shares for a videos
Video Age (in The number of days since the date the Self-Defined CF IO*
days) video was posted
For You Page “An individual landing page for users (Ling et al.,
which showcases curated videos that 2022)
TikTok thinks they might watch or (“What Does
like.” FYP Mean,
And Why Is It
All Over
TikTok?,”
2020)

* Value was calculated based on other features of this type


** Source was not used to define the feature but includes research regarding the use of the
feature.

90
Appendix H: Demographic of Survey Respondents

Age range distribution of survey respondents.

Ethnicity distribution of survey respondents.

91
Country of residence distribution of survey respondents.

92
Annual family income level distribution of survey respondents.

Education level distribution of survey respondents.

93
Appendix I. Social Media Usage of Survey Respondents

Frequency with which respondents reported using the social media and video sharing platform
YouTube.

Frequency with which respondents reported using the social media and video sharing platform
Instagram.

94
Frequency with which respondents reported using the social media and video sharing platform
TikTok.

Distribution of how long respondents reported they use social media per day.

95

You might also like