The Science of YouTube

How the Algorithm Really Works

“The algorithm isn’t mysterious. It has one job: keep people on the platform. Once you understand that, everything else follows.” — Varun Mayya

One Goal: Session Time

YouTube’s business model is simple: show ads to viewers. To show more ads, they need viewers to watch more. The algorithm’s entire purpose is to maximize how long people stay on the platform.

This means the algorithm promotes content that:

The Key Metrics YouTube Watches

The Algorithm’s Testing Process

When you publish a video, YouTube shows it to a small sample of your subscribers and people with relevant interests. Based on how that sample responds, YouTube decides whether to push it wider.

The testing stages:

  1. Stage 1: Shown to ~100-500 people (often subscribers)
  2. Stage 2: If CTR and AVD are good, pushed to ~1,000-5,000
  3. Stage 3: Good performance = pushed to broader audience segments
  4. Stage 4: Strong videos get browse/suggested placement

CTR × AVD

The simplified formula for algorithmic success

Traffic Sources Explained

Common Mistake: Optimizing for One Metric

Clickbait thumbnails boost CTR but tank retention. Boring intros protect watch time but kill CTR. The algorithm wants both — click-worthy AND watch-worthy. Optimizing one at the expense of the other doesn’t work.

The Real Truth About “Beating” the Algorithm

You don’t beat the algorithm — you align with it. The algorithm wants to surface content viewers will love. If you make content viewers genuinely love, you and the algorithm are on the same team.

Stop asking “what does the algorithm want?” and start asking “what does my audience want?” The answers are usually the same.

Pro Tip

Check your Analytics → Audience → When Your Viewers Are On YouTube. Publishing when your audience is online gives you a better first-hour performance, which helps the algorithm’s initial testing.

Action Steps

Key Takeaways

← Previous: Chapter 6 Next: Chapter 8 →