I spent the holiday eating and conversing with family on various topics this Christmas. One of the more interesting points I noticed was the widespread usage of YouTube among family members, both young and old.
This led me down a rabbit hole of searching How Big is YouTube? Luckily, I stumbled across Ethan Zuckerman's article detailing that exact question.
In his article, Ethan Zuckerman delves into the vastness of YouTube and the challenges of quantifying its content. He explains the "denominator problem" in social media research, where the focus is often on specific, harmful content types without considering the overall content volume. For platforms like Facebook, a seemingly large number of views can be misleading without this context.
Zuckerman highlights that platforms like Reddit and Twitter, previously more accessible for research, have now restricted access, making it harder to study their total content. However, YouTube, used by a vast majority of internet users, presents a unique challenge due to its difficulty obtaining a random, representative sample of its content.
He discusses a novel approach to estimating YouTube's size, coined "drunk dialing," where random URL generation is used to find videos. This method, improved by Jason Baumgartner and Kevin Zheng, led to the collection of over 10,000 random videos. Analysis of these videos, led by Ryan McCarthy, provides insights into YouTube's language distribution and view count.
Their research estimates the size of YouTube to be around 13.325 billion videos, growing rapidly each year. It also reveals that most YouTube content belongs to the "long tail" category with fewer views. Zuckerman argues for the importance of such data for understanding large user-generated media platforms and their role in the digital public sphere. He plans to maintain and update this data on Tubestats.org, acknowledging the ethical considerations of handling such data.
Overall, the article is a worthwhile read, and it's fascinating that the median YouTube video only has 39 views.