An Analysis of Female Faces in Mr. Beast Thumbnails


*For the best experience, please view this site on a desktop!

Mall Picture
Jimmy Donaldson is the world's greatest YouTuber. In this analysis, I take a look at his channel's performance over the past 10 years. In particular, I take a deep dive into his usage of female faces in his thumbnails.

A Look at MrBeast's Long Form Videos:

Below I built a live dashboard that pulls in all long form videos on Mr. Beast's channel from the YouTube API.

Some insights worth calling out:

  • There was a dramatic pick up in views from mid-2018 onwards.
  • Long form videos tend to have more comments but fewer views than shorts.
  • It might seem like views are leveling off, but in reality, it just takes time for videos to accumulate views.
  • Meaning, recent videos can't be compared point blank to older videos.
  • His squid games video gained over half a billion views 🤯
  • A Look at MrBeast's Short Form Videos:

    Below is a live dashboard I built that pulls in all short form videos on Mr. Beast's channel from the YouTube API.

    Some insights worth calling out:

  • YouTube shorts began in September 2020 to compete with the rise of TikTok.
  • However, we can see below that Jimmy didn't begin making shorts until more than 2 years later.
  • He produced one short per month for his first nine months, but has increased production to 4 shorts/month these last few months.
  • Shorts tend to have more views but fewer comments than long form videos, which likely translates to lower ability to build community.
  • His short about the Paris baguette is his most viewed video all time, with nearly a billion views as of today 🤯
  • Video Production Over Time

    Below is a live dashboard that displays the number of (long form) videos that Jimmy has published each year for the past decade.

    Some insights worth calling out:

  • Presumably he has deleted several videos, but the general trend is that he started in gaming and then hard pivoted into entertainment.
  • For context, YouTube categorizes videos into 22 distinct categories, with "Gaming" and "Entertainment" being 2 of the 22.
  • We can also see that the overall number of videos has steadily dropped over the last 5 years.
  • This is likely because of increasing production value. Quality has increased whereas quantity has decreased.
  • A Deep Dive into Thumbnails

  • I was curious whether having a woman in the thumbnail increased the number of views on Mr. Beast's videos.
  • So I pulled all of his thumbnails over the past 10 years from YouTube's API.
  • I then used a couple machine learning algorithms to scan the thumbnails for faces and to categorize their genders.
  • In the end, I wasn't able to find any correlation, but I did still discover interesting trends which I've laid out below.
  • Woman Image
    Woman in Thumbnail (2021)
    Jimmy Image
    Man in Thumbnail (2023)


    First Chart
    The first thing worth noting is that the vast majority of MrBeast's thumbnails show a face. However, this wasn't always true. The percentage of videos with a face in the thumbnail has increased steadily over the past 10 years. Today, over 90% of Jimmy's thumbnails now show a face. (Facial recognition algorithm used: FairFace)
    Second Chart
    When we compare the number of views for these videos we see that there might be a slight advantage to thumbnails with a face, but the difference seems negligible. (Source: YouTube API)
    Third Chart
    Jimmy is much less likely to include women on his thumbnail images. Above we see the percentage of thumbnails that display at least one woman. On average, only about one in every 10 videos will show a woman in the thumbnail. (Facial recognition algorithm used: FairFace)
    Fourth Chart
    When we examine the views of videos with women in the thumbnail vs. those without, we again do not see any meaningful trends. (Source: YouTube API)


  • With just the data above, I cannot determine whether adding a woman to a thumbnail will increase view count.
  • To do that, I would need to build a substantially more robust model using much more data.
  • In particular, I would need to gain access to MrBeast's audience segment breakdowns, retention rates, and click-through rates.
  • With enough data though, I'm confident that I could predict the impact of female faces on view count.
  • If I could show an effect, then Jimmy should consider using more women in his thumbnails to increase his views.
  • In the future, I'd like to check whether "attractive" women do better than less "attractive".
  • I'd also be curious to see whether the same trends hold for male faces.
  • I'm also curious whether using female faces leads to a correlated decrease in brand value or brand trust.
  • Ultimately though, any model I build can only be predictive. To understand a causal effect, we'd have to run A/B tests.