Creators talk about "the algorithm" as if it were one gatekeeper deciding who wins. In reality YouTube runs separate ranking systems for the home feed, suggested videos, search, the Shorts feed, and notifications. Each one answers a different question, and your video can succeed on one while failing on another.

The four surfaces that matter

  • Home feed — answers "what will this specific person want to watch right now?". Driven mostly by personal watch history and by how similar viewers responded.
  • Suggested — answers "what should play after this video?". Topical and stylistic similarity plus session length matter most here.
  • Search — answers "which video best matches this query?". The only surface where titles, descriptions and spoken words carry heavy weight.
  • Shorts feed — answers "will this person keep swiping?". Judged in the first two seconds, then by loops and swipe-away rate.

What the system actually measures

YouTube does not score "quality". It scores observable behaviour, then uses that behaviour as a proxy for quality. The main signals, roughly in order of influence:

  1. Click-through rate on impressions — of the people shown your thumbnail, how many clicked? A video that is never clicked is never recommended further.
  2. Average view duration and percentage viewed — short videos are judged on percentage, long videos on absolute minutes. Both matter; neither alone is enough.
  3. Session contribution — did the viewer keep watching YouTube afterwards, or close the app? Videos that end sessions get quietly suppressed.
  4. Satisfaction signals — likes, shares, subscriptions from the video, and survey responses. These separate "watched because I was stuck" from "watched and enjoyed".
  5. Negative signals — "not interested", "don't recommend channel", quick swipe-aways. A small number of these costs more than a large number of likes gains.
Subscriber count is not a ranking factor. It affects how often your video is shown to people who already subscribed — nothing more. A 500-subscriber channel and a 500,000-subscriber channel compete on identical terms in the browse feed.

Why your video "flopped" even though it was good

Every upload gets a test batch of impressions — usually to your subscribers and to viewers whose history matches your topic. If CTR and retention in that batch beat the alternatives YouTube could have shown instead, the batch grows. If not, distribution stops. That is the whole mechanism.

So a "flop" almost always has one of three causes: the thumbnail and title did not earn the click, the first thirty seconds lost people, or the topic had no existing audience to test against. Fixing the third requires a different video; the first two are editable.

The first thirty seconds

Retention graphs almost always show a cliff in the opening. Reduce it by cutting the intro, restating the promise of the title in the first sentence, and showing the most visually interesting moment early. Long channel intros, slow logo animations and "hey guys, welcome back" cost more views than anything else creators voluntarily add.

Shorts are a separate game

The Shorts feed does not care about your channel history nearly as much. Each Short is tested almost from scratch, which is why a new channel can get a million views on its third upload. The flip side is that Shorts success transfers poorly to long-form: a Shorts subscriber is worth far less in browse-feed terms than a long-form subscriber.

If your long-term goal is watch time and RPM, treat Shorts as top-of-funnel: use them to find which topics resonate, then build long-form around the winners.

What consistency actually does

Uploading frequently does not "please the algorithm". What it does is give the system more test batches, which means more chances to find an audience, and it teaches YouTube what your channel is about. A channel with a clear, repeated topic is far easier to recommend than one that posts a cooking video, then a tech review, then a vlog.

This is why scheduled, automated publishing works: not because frequency is magic, but because it removes the human bottleneck that causes six-week gaps.

Things that do not matter as much as people claim

  • Tags — useful for misspellings and disambiguation, essentially irrelevant for ranking.
  • Upload time of day — matters for your notification wave, not for long-term distribution.
  • Video length — there is no magic number. Make it as long as it stays interesting.
  • Deleting underperforming videos — does not "clean" your channel's score. There is no channel-wide penalty score.

A practical checklist before every upload

  1. Can a stranger understand the promise from the thumbnail alone, at phone size?
  2. Does the title complete the thumbnail rather than repeat it?
  3. Is the payoff of the title visible or promised within the first fifteen seconds?
  4. Is there a reason to keep watching at the one-third and two-third marks?
  5. Does the ending point at another video rather than at nothing?

Understanding the algorithm is not about tricks. It is about recognising that YouTube is constantly asking "should I show this instead of something else?" — and making sure the answer, for your target viewer, is obviously yes.