Anatomy of a viral video
Every video gets an audition.
TikTok, YouTube Shorts and Instagram Reels do not show a new video to everyone at once. They show it to a few people, watch what those people do, and decide whether to show it to more.
Illustrative model Room sizes of 300, 3,000 and so on are round numbers for the example. Platforms do not publish them.
Room 1 · the test audience
First, 200 strangers see it.
A test audience is a small group the platform picks because they tend to like this kind of video. Each square is one person. Its bar fills to how much of the video they watched.
Illustrative model Completion rate is the share who watched to the last second. Share rate is the share who sent it to someone. The score mixes average watched (45%), completion (35%) and shares (20%). Seeded, so every run is the same.
Test score 0 of 100. Benchmark: 40+ expand to a bigger room, 25 to 39 hold and keep testing, under 25 stop.
How long people stay
The retention curve tells the story.
A retention curve shows how many viewers are still watching at each second. The steep drop at the start is people swiping away. The area under the curve is the average watch time.
With loops on, people who reach the end and start again are counted a second time, so the curve can finish above the 100% line and average watch time can be longer than the video itself.
Illustrative model A 20 second example video. Curve shapes are made up to show the pattern, not taken from any platform.
What viewers tell the system
Some signals push harder than others.
A signal is anything a viewer does that the platform can count. Tap a card to send that signal from one viewer and watch the combined push move.
What TikTok has published. Its For You feed weighs what you interact with, details of the video such as captions, sounds and hashtags, and, more lightly, device and account settings. Finishing a longer video is called a strong indicator of interest. Follower count and an account's past hits are not direct factors. Tapping Not Interested shapes what you see next.
What YouTube has published. Recommendations use your watch history, searches, subscriptions, likes, dislikes and Not Interested feedback. Surveys ask viewers to rate videos they watched, so the system learns about satisfaction and not only watch time. The Shorts feed is personalised in the same way.
Expansion
Bigger rooms, harder crowds.
Pass one room and the video moves to the next, ten times bigger. Each new room is less matched, meaning fewer people in it already like this kind of video. The same video now has to work on strangers.
Press Post 50 videos to run the model.
Illustrative model Room sizes, match levels (95% down to 48%) and a pass mark of 50 are invented. Score = video quality × match + luck.
Who sees it next
It follows interests, not followers.
Viewers cluster by what they watch. A cooking video goes first to people who watch cooking, then hops to neighbouring interests where some viewers overlap. TikTok says follower count is not a direct factor.
What makes a topic clear: words in the caption, the sound, hashtags and what is on screen. TikTok lists captions, sounds and hashtags among the details it reads.
Illustrative model Hit rate is the share of people shown the video who watched most of it. Cluster layout and odds are invented.
Your side of the audition
What a creator controls.
You cannot pick who sees a video. You can change how it performs in each room. Switch on a habit, then post 100 videos with the same model as Bigger rooms.
Even with every habit switched on, luck still decides a lot. A share of good videos stop in the first room, so posting often matters too.
Illustrative model Each habit adds a few points to the room score. The sizes are invented to show direction, not exact effect.
In one line
Going viral is passing one audition after another.
- 01Test audienceA few hundred matched viewers see it first.
- 02RetentionA strong hook keeps people past second two.
- 03SignalsWatching to the end, rewatches and shares push hardest.
- 04Bigger roomsEach room is ten times bigger and less matched.
- 05InterestsIt travels by topic to neighbouring interests.
Sources: TikTok Newsroom, “How TikTok recommends videos #ForYou” (newsroom.tiktok.com); YouTube Help, “How YouTube recommendations work” (support.google.com/youtube/answer/16089387). Room sizes, weights, pass marks and every simulation in this deck are an illustrative model.
Made by Jaw Dropping Presentations