• Creator 3x3
  • Posts
  • How YouTube's algorithm works, Beehiiv's State of the Newsletter, and what everyone gets wrong about social media.

How YouTube's algorithm works, Beehiiv's State of the Newsletter, and what everyone gets wrong about social media.

Welcome to Creator 3×3!

This week we’re talking about platform algorithms.

And by the end you’ll understand why YouTube (and everyone else) is a content platform, not a channel platform - and why this means that platforms like YouTube and TikTok sometimes have conflicting goals with Creators.

Here are 3 algorithmic facts from the past, 3 notes on the present, and 3 thoughts about the future of the Creator economy...

Note: For us, an “algorithm” means the platform decisions, code logic, machine learning, data science, and all the other secret magic that content platforms use to decide what kind of content to show in your feed.



YouTube’s algorithm is a secret.


YouTube shares that they “match each viewer with the video they’re most likely to watch and enjoy” - which they base on factors like:

  • What videos users choose to watch

  • What videos they ignore or dismiss

  • How often that person watches a channel or topic

  • Average view duration

  • Average % of video viewed

  • Likes, surveys, and hundreds of other engagement metrics

But the reality is that you and I can’t go in and look at YouTube’s code.

And whenever there’s a black box I’m not allowed to look into I get a little…


I don’t really think YouTube is lying about how they rank videos, but that codebase is huge, man.

YouTube is waaaay too big for any one person to completely know what’s going on anymore - so there’s definitely going to be quirks.

It’s also one of the oldest social platform algorithms on the internet.

In its 18 year history, YouTube has had 4 major algorithm changes (and thousands of smaller ones):

  1. In 2012, YouTube changed from a “view count” focused ranking to a “watch time + session duration” focused ranking to combat clickbait titles dominating the site.

  2. Around 2016, YouTube shifted again from a “video performance” ranking to a “user satisfaction” recommendation engine based on that user’s YouTube history. It still used the existing metrics like watch time to analyze what videos were best for a user, but the change allowed for much better recommendation customization than ever before.

  3. In 2017, YouTube was hit with Elsagate. Elsagate is best understood as the period of time when really weird videos dominated the platform because kids got hooked on watching popular cartoon characters in insanely inappropriate scenarios. This very public scandal kicked off a huge response from YouTube to crack down on channels creating (and profiting) off of manipulating child dopamine. It also launched the YouTube Moderation Juggernaut which has continued to evolve to algorithmically suppress “borderline” content like alt-right propaganda, science denial, and election misinformation in waves in recent years.

  4. In 2021, YouTube launched Shorts in the U.S. to compete with TikTok. YouTube hasn’t shared a ton of detail how the Shorts algorithm is different, but they have said it is different than the “longs” (lol) algorithm in some key ways because swiping through a feed is a different user behavior than clicking through a thumbnail. Anecdotally, Creators report whiplash from Shorts getting over-emphasized in feeds, the cannibalization of longer video views, and the rapid testing of different ways to rank short form videos.

Today, YouTube explicitly says they focus on “individual video and audience level signals” rather than a channel’s history as a whole, although they mention if you have a series of videos with poor metrics in a row that can impact your video’s rankings.

YouTube’s biggest piece of advice is:

“Focus on what your audience likes. If you do that and people watch, then the recommendations will follow.”

And as frustrating that vague advice is… MrBeast basically agrees.

He then extends the idea to recommend that Creators get a lot more reps in making content because that’s how you actually get better, but ultimately lands on YouTube’s side that the algorithm probably isn’t wrong about that video.

Our content just sucks 🤷‍♂️

Fortunately, users don’t just find videos in their feed.

YouTube is also the world’s second largest search engine.

Behind Google of course.


So how does Google Search work?

They’ve had even more time to experiment with shoving content into people’s faces to generate ad revenue so they must be pretty good by now.

How good?

$160 Billion good in 2022 (or about 6 Icelands worth of value).

To stay on that grind, Google releases thousands of changes to their algorithm every year.


Not all of these are so-called “core updates”, but for everyone trying to rank for niche keywords even the smallest changes can have huge impacts on their search performance.

But what is Google even doing?

They’re essentially ranking pages using an algorithm cleverly named…

According to Google itself:

“PageRank works by counting the number and quality of links to a page to determine a rough estimate of how important the website is. The underlying assumption is that more important websites are likely to receive more links from other websites.”

Pretty simple to be honest?

PageRank is the idea that started Google.

They still use PageRank today as part of their algorithm, but, similarly to YouTube, Google’s algorithm has undergone multiple major changes in their 25 year journey to help organize the world’s information.

You can find a complete list of those changes here, but unless you’re going to enter the mysterious world of search engine optimization (SEO) hardcore-style the big takeaway is that Google runs code to scan every web page on the internet and tries to use links between websites as a signal for how useful that web page is for users.

So whenever an important website (like a news agency) links back to your site, Google interprets that as a positive signal for your search rankings.


YouTube… Google…

I wonder what’s the third most important social algorithm of the past decade?

Probably TikTok.

TikTok’s parent company Bytedance is notoriously quiet about how TikTok’s algorithm actually works, but that hasn’t stopped everyone from Creators to politicians to journalists from speculating on how they think it works.

As far as I can tell, the last time TikTok publicly published anything (at least in English) about their algorithm was way back in June 2020 in response to increasing U.S. pressure about Chinese ownership of how TikTok’s feed is managed.

They share some very unhelpful information like how their recommendations are influenced by things like “video information”, “language preference”, and “user interactions”.


A year later, the New York Times published a fairly hostile analysis of an internal ByteDance explainer document that they claim was leaked to them by a concerned employee.

Their article had a little bit more information for Creators, but it is still fairly straight forward in its explanation that “more likes + comments” and “more playthroughs” generally equal more airtime in the For You page.

A breakdown of the “For You” page basics

But ByteDance is known as “the algorithm company” - we’ve got to be able to do better than just “more likes” right?

The best deep dive about the TikTok algorithm that I’ve found is from Eugene Wei.

He’s an outsider looking in, but the case he makes is pretty convincing and it’s centered around the idea of TikTok’s “Algorithm Friendly Design”.

This can be understood to basically mean 3 things.

  1. TikTok’s app design flow with it’s “one video at a time” for users lets TikTok capture a lot more data about a user than an infinite scroll feed like Twitter or pre-Shorts YouTube.

  2. TikTok’s native content (short form video with added sounds) is inherently easier to analyze and map to user interest graphs (ie. they can tell you like “Baby Fox” videos but not “Baby Cat” videos in lieu of just “Baby Animal” videos)

  3. TikTok’s in-app production tools make the creation loop much simpler for Creators.

Taken together, this means that TikTok’s algorithm is just an extremely ruthless and well-informed version of YouTube’s advice to:

“Focus on what your audience likes. If you do that and people watch, then the recommendations will follow.”

Except the platform started with short form video, so the emphasis on things like the first frame visuals, engaging sounds, and compelling intro hooks are much more critical for Creators to think about because they don’t have additional communication tools like a thumbnail or title to pique the interest of a potential viewer.



More recently, Instagram’s CEO published a blog post about their ranking algorithms.

Note the plural!

Every section of Instagram has a different algorithm apparently.

By section, the most important data for each of Instagram’s algorithms are:

  • Feed

    • A user’s likes, shares, saves, and activity

    • Post information including likes, shares, and comments

    • Creator information (ie. is the user gonna like with this person)

    • Interaction history (has the user ever engaged with their posts)

  • Stories

    • User viewing history

    • User engagement history

    • Social graph closeness (is the user their brother or just a friend?)

  • Explore - “Focused on helping users discover new things”

    • Post information including likes, shares, and comments** Instagram says this matters much more here vs Feed

    • User activity in Explore

    • Interaction history

    • Creator information

  • Reels - “Discovery focused on entertainment”

    • User activity

    • Interaction history

    • Post information

    • Creator information

The most interesting takeaway here is that it’s all basically the same data, but the ranking priorities are different across each section of the app.


Why do we even care about algorithms anyway?

The obvious answer is that if we do what the algorithm wants

Then as Creators we’ll make more money.


But the answer isn’t really so simple.

Broadly, of course it’s true that more views means more Adsense dollars rolling into your account every month.

But counterintuitively, Adsense isn’t really where the money is for most Creators.

Dynamic Creator duo, Colin and Samir (two guys who clearly do understand the YouTube algorithm) recently shared a pie chart of their income streams.

Eyeballing their earnings breakdown that’s:

  • Consulting - 10%

  • Speaking - 14%

  • Adsense - 16%

  • Brand Partnerships - 60%

(If you have more accurate numbers, please email me at [email protected] so I can update this post)

The vast majority of their total earnings comes from everything outside of direct advertising revenue

And that feels weird the first time you hear it.

But really it’s an open secret in the Creator Economy that the most viewed Creators aren’t necessarily the highest earning Creators.

What does that mean for you?

Understanding your platform’s algorithm is important, but understanding how Creators in your niche actually make money might be even more important.


I’m (obviously) a fan of newsletters.

They’re a crucial relationship platform for Creators that want to build a long term relationship with their community and they force clarity of thought in ways a lot of other content mediums do not.

I use Beehiiv (Affiliate Link 🫡) to manage my newsletter and they dropped a really useful State of Newsletters report that has a bunch of practical information for newsletter writers.

For example, what day of the week should you send your newsletters out?

Apparently Sunday/Monday! (whoops)

What time of day should you send it? 

10:00am GMT! (also whoops)

To be honest, I’m not gonna implement these changes just yet, but it’s always good to explicitly know when you’re doing something sub-optimal.



Every social platform more or less wants the same thing.

Get more users to engage with more content more of the time.

Each platform has a different audience with different consumption behavior so individual priorities can be different at different points in time, but they all operate in pursuit of that same goal.

That means while your tactics may need to be different from platform to platform, your overall strategy can remain remarkably similar even across completely different content types.

My strategic question almost always is:

How can I deliver the most value per unit of content to my audience with this next post?


You’re not the only one who has to deal with changing algorithms!

Every single Creator on the internet is subject to platform shifts and fickle audience tastes.

Take Gary Vee’s (4.3M subs) ridiculously well distributed channel.

This is his cumulative view count over time.

Notice how shallow the growth rate under the arrow is compared to 2022?

He’s still growing, but I bet he wishes he was back in his “hyper-growth era”.

He has to do something different if he wants to kickstart that faster growth again.

In a way, the content game doesn’t ever really change because change is the game.



Creators and the Platforms on which they post usually have the same goals.

They both generally want users to enjoy their content, they both want to increase grow view counts, and they both want to understand why some shorts, tweets, and videos get a lot of views (or none at all).

But Creators and Platforms are in direct conflict in one very important area.

Creators want fans to engage with their channel, while platforms want fans to engage with individual pieces of content.

Said another way…

Apps like YouTube are Content Platforms not Channel Platforms.

They ultimately care about how content performs, not how Creators perform.

It’s great for YouTube when Creators do well because they’ll create more content, but what if they can’t generate enough views support themselves?


There’s always somebody else out there who will still keep making funny videos.

So the next time your favorite platform makes a decision you pisses you off…

Remember this conflict.

In a way, algorithms are the gatekeepers.

They are what stand between us and that sweet, sweet engagement.

But when we understand what they’re designed to do

(Match the best content to the the user most likely to enjoy it)

We can start to recognize that algorithms our partner too.

Even if our goals aren’t entirely the same.

I want as many views as I can get, but also…

Do I really want to force someone who isn’t going to get any value from my video to watch it?

Getting in front of the right audience is infinitely more useful and enjoyable (and ultimately lucrative) for the both of us 🫡

Want to share this issue of Creator 3×3 via text, social media, or email? Just copy and paste this link:


Please subscribe and if you have any feedback or requests for Creator 3×3 just respond to this email or reach out at [email protected] 

→ I reply to every email!

See you next week,

p.s. if you’re looking to put a little pep in your step check out neuroscience PhD and Buddhist Chaplain Christin Chong’s latest newsletter - it’s a great reminder about health + self-reflection…

Join the conversation

or to participate.