Skip to content

epfl-ada/ada-2025-project-shenanigans

Repository files navigation

How indie YouTubers became industry players

sheNaNigans

Analysing content, upload, and sponsorship trends of YouTube channels over time.

The Website

The final website can be found here: https://ri-ru.github.io/ADA-website/

The repository for the website can be additionally found here: https://github.com/ri-ru/ADA-website

Abstract

At the beginning of YouTube, creators mainly posted for fun - making something to entertain others. Back then, people didn't put much thought into making a living out of it. Today, we have channels competing to get viewership for monetization and partnering with big sponsors, turning content into profit. Clearly, the creator focus has evolved from indie content to establishing a profitable brand, which is a fantastic representation of professionalisation on social media.

This project investigated the professionalisation of YouTube channels - analysing their journey from indie to industry-like creators. We attempted to quantify when we can see clear transitions in adopting professional strategies, analyse how the creators were able to achieve this, and explore the effect of this on the channels and audiences.

Research Questions

1. How have channels evolved from indie to professional?

  • How did the YouTube ecosystem change as it grew bigger?
  • Did all categories grow in the same way?
  • How did different types of links (content, social, monetization) spread and change over time?
  • Which categories used monetization links the most?
  • Did channels become professional all at once or gradually?

2. What enabled indie channels to become professional-minded?

  • How long does one need to have a channel to obtain a sponsor?
  • At what thresholds (views, subscribers, videos, activity) do creators obtain their first sponsorship?
  • Are certain topic transitions common after the first sponsorship?
  • How does getting sponsored affect channel longevity?
  • Are there cohorts of channels with the same/similar sponsors?

3. How does a focus on professionalism affect engagement?

  • Are like/dislike ratios, video durations, view counts different between sponsored and unsponsored channels? Is it different between categories?
  • Do the number of views and the subscriber growth change after channels get sponsored?

Dataset

We use two datasets: YouNiverse and SponsorBlock

The YouNiverse dataset, produced by the ADA lab at EPFL, provides a large-scale overview of the English-language YouTube ecosystem. It incorporates metadata for more than 72.9 million videos and 153,550 channels with weekly time series, totally 18,872,499 observations.

SponsorBlock is an open-source browser extension enabling users to automatically skip chosen types of segments on YouTube videos (preview, credits, sponsor segments, ...) and it is crowdsourced: users submit the start and end times of these segments. For this project, only the sponsor segment category is selected, thus including only videos with paid promotional content for a brand or product.

Dataset enrichment

Sponsor segment analysis

We used the SponsorBlock crowdsourced dataset to label videos as sponsored. The crowdsourced dataset started in 2019, but we found pretty uniform labelling for all years in the YouNiverse dataset.

Unshortening URLs

Many URLs in the dataset corresponded to shortened URLs (e.g. bit.ly). We unshortened the subset of URLs from sponsored videos.

Channel and video activity tracking

The activity of channels in 2025 was verified using the YouTube API. This would enabled comparing of stability between professional and indie creators.

Methods

The bulk of the analysis is detailed in this notebook.

Main approach - R1 - Platform Evolution

Analysis of upload volumes across categories

Classification of links and adoption rate tracking

Cross-category comparisons of industry-like patterns

Main approach - R2 - Path to Sponsorship

Distribution fitting for time-to-first-sponsor

Log-normal distribution analysis for video count thresholds

Category transition analysis using 6-month windows before/after first sponsorship

Channel-sponsor network construction with Leiden community detection

Paired t-test comparing sponsored vs. non-sponsored channel longevity

Main approach - R3 - Engagement Analysis

Dumbbell plots comparing sponsored vs. non-sponsored video metrics

Time-series analysis of delta_views and delta_subs around first sponsorship

Engagement comparisons

Main takeaways

YouTube grew enormously in volume from 2008 to 2018. Creators professionalized in different ways: adding monetization, social, and content links. Different categories did it at different speeds.

On median, it takes ~3 years and ~148 videos to land your first sponsor. Sponsored channels are 16% more likely to still be active in 2025. Thresholds for views, subs, and videos vary with category.

Sponsored videos get more views, are longer, and have higher like ratios. But after your first sponsor, growth rate slows down!

Initial timeline

gantt
    title sheNANigans ADA Project Timeline
    dateFormat YYYY-MM-DD
    tickInterval 1week
    section Exploration & Analysis
        Temporal Evolution : 2025-11-5, 14d
        Research Question 1 : 2025-11-12, 14d
        Research Question 2 : 2025-11-19, 14d
        Research Question 3 : 2025-11-26, 14d
    section Website implementation
        Foundations        : 2025-11-5, 7d
        Add temporal analysis  : 2025-11-12, 7d
        Add RQ 1 analysis  : 2025-11-19, 7d
        Add RQ 2 analysis  : 2025-11-26, 7d
        Add RQ 3 analysis  : 2025-12-3, 7d
        Add story telling  : 2025-12-3, 10d
        Finish up          : 2025-12-10, 2025-12-17
Loading

Contributions

  • Ender Sari: Focus on research question 1 and YouTube data enrichment. Website implementation for question 1.
  • Kalan Walmsley: Performance-intensive data preprocessing. Focus on research question 2. Website implementation. WebAssembly graph visualisation. Cleaning and preparing P3 submission.
  • Danael Robert-Nicoud: Focus on research question 3.
  • Veronika Wannack: Website design and implementation, cleaning and preparing P3 submission.
  • Jan Tomasz Juraszek: Website implementation and story, cleaning and preparing P3 submission.

Releases

No releases published

Packages

No packages published

Contributors 5