Predicting User-Engagement on YouTube

I choose this topic as part of my mandatory master’s thesis. Coming from a digital agency background, I wanted to explore the theoretical side of social media. During my time at the agency, clients and peers had always been complaining about social media engagement – why aren’t we getting results? That is where I thought to myself, why not dive deeper into this phenomenon? I did not know it would change the way I look at the world! Let’s start with a bit of what I discovered as I continued my masters.


Based on a lot of literature, I discovered that Social media, despite providing new marketing opportunities, has also created some complexities for brands. Brands face difficulties in achieving results for their social media marketing campaigns. As users can connect and communicate with a brand on a variety of social media applications, brands now require different social media strategies, with content that is relevant to the needs of users and such tactics may become hard to execute. YouTube, the platform containing millions of content creators has become the preferred destination for sharing and viewing videos online. Merrill Lynch, one of the world’s largest wealth management company forecasted a gross digital advertising revenue for YouTube to be 20.4 billion USD for 2018, amounting to 38% of the total digital ad share. YouTube allows users to create, view, post and share videos. Users who create and upload content on YouTube are known as vloggers. Some vloggers become “YouTube celebrities” when they reach a considerate number of followers or subscribers. Creative vloggers can achieve a high reach due to a large following, with a few having more than 100 million subscribers. Vloggers establish a high online following due to their ability to develop and contribute novel information and influence the attitudes and behaviours of others. Considering the influence and importance of vloggers, there is little research around YouTube engagement studying which type of YouTube videos and channels (vlogger or brand) generate more engagement.

YouTube contains videos from eighteen separate categories. Sports, one of the eighteen separate categories present on YouTube, is one of the most popular video categories and generates one of the highest numbers of views. Considering the prevalence and viewership of sports videos on YouTube, I saw an opportunity to study sports videos and channels, created by different sports brands and vloggers, to develop an understanding of what factors generate engagement for such types of channels. I decided to take make something of this opportunity.

Research Objectives and Qs

The main research objective of this research was to examine factors that may help predict user engagement with sports brand-related YouTube videos and investigate which type of YouTube channels generate increased user engagement. To achieve that, I derived the following questions:

So, what do I mean I write fancy words like User/YouTube Engagement, channel properties and video properties? Here is how I defined my variables (the fancy words):

Thinking of types of channels? I will have to explain my research framework and methodology first.

Research Framework

So basically so far, I am trying to see if we can use channel and video properties to predict user engagement on YouTube videos. Here is what my framework looked like (ignore the ‘Hs’ please):  

Research Methods

I kept the methodology and design as simple as possible.  The design comprised quantitative research that applied multiple linear regression to analyze data. The dataset comprised of YouTube videos containing elements of ‘Adidas Soccer.’ Why Adidas – because it is my favourite sports brand – period! The dataset was extracted using an online tool called Netvizz that is designed for researchers to extract data from YouTube using simple techniques like ‘keyword search’ – which I used as well. Keywords ‘Adidas Football’ and ‘Adidas Soccer’ provided me with a dataset of 707 entries that comprised videos uploaded by the brand as well as different vloggers. To further explore the type of channels posting videos on Adidas Soccer, I had to manually view content and classify each type of channel. Five different types of channels were explored. The definition and dataset can be found below which were used to carry out the research:








Brand channels are likely to generate 4x times more engagement than other channels. This is because brands carry a larger network and more firepower in terms of advertising spend and the ability to feature celebrities.

A positive relationship between the no. of channel subs and YouTube engagement was found. Many content creators preferred gaining new followers over video views or likes.

YouTube engagement is likely to increase as the channel ages. YouTube allogrithm play a massive role especially by featuring videos in different user interests.


That’s all folks…hope you enjoyed the interesting insights!

Predicting YT Engagements – References


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.