Getting a sense of the scale of social media platforms can be tricky. While tech companies often share self-serving metrics—like monthly active users or how likely users are to buy products after engaging with brands—they rarely offer a true sense of their platforms’ enormity.
But a new study published in Cornell University’s preprint server arXiv aims to change that by quantifying TikTok’s scale over a single day—claiming to be among the first to grasp the platform’s full scope. It also offers insight into what people are watching, how much content is being uploaded, and who is posting it.
“The motivation is using this social media data to better understand society,” says Jürgen Pfeffer, one of the study’s coauthors and chair of computational science at the Technical University of Munich. The researchers reverse-engineered how TikTok assigns video IDs to gather their data, eventually capturing what they believe to be at least 99% of all posts from a single hour on April 10, 2024—around five million videos. They also sampled one minute from every hour across a 24-hour span between April 9 and 10. The entire process took five months.
Among the team’s findings was the sheer scale of TikTok activity. They estimated 117 million videos were uploaded on that single day. Roughly one in five featured children, based on estimates from an age classification engine the team trained. Videos posted precisely at the 0th second of each minute significantly outperformed others—suggesting timing tricks might influence the algorithm. However, Pfeffer notes this could be due to professional creators scheduling their posts, who are also more likely to achieve success.
About a fifth of the videos were deleted within seven months, hinting at large-scale moderation or user regret.
The average video was approximately 20.5 seconds long and was viewed between 2,250 and 2,500 times—depending on whether the data came from the hour-long slice or the one-minute-per-hour sample.
Pfeffer was also surprised to find that TikTok engagement peaked twice daily. Around 4 a.m. UTC (11 p.m. EST), uploads surged to nearly six million videos per hour, reaching a similar peak again around 1 p.m. UTC (8 a.m. EST). The researchers attributed these spikes to global usage cycles, as morning users in Asia overlapped with late-night uploaders in the U.S., and vice versa later in the day.
The team also grouped popular videos into thematic clusters. Among the most common were “images with colorful backgrounds and graphics” and “selfies with fashionable outfits and watermarks.” Less frequent themes included “political commentary on police operations in Pakistan” (0.025% of all videos) and “camouflaged military soldier images,” which accounted for 0.06% of the content.
Zaloguj się, aby dodać komentarz
Inne posty w tej grupie
I’ve worked at the bleeding edge of robotics innovation in the United States for almost my entire professional life. Never before have I seen another country advance so quickly.
In
Restaurant industry leaders are excited for
Elon Musk’s anger over the One Big Beautiful Bill Act was evident this week a
Welcome to AI Decoded, Fast Company’s weekly new
When artificial intelligence first gained traction in the early 2010s,
