A structured breakdown of how the YouTube algorithm actually works, covering three levels of depth. At the beginner level: the algorithm finds videos for people (not the reverse), and clarity of niche plus CTR and average view duration are the key metrics. At the intermediate level: traffic sources (browse, suggested, search) each signal different things, satisfaction surveys matter more than raw watch time, and format saturation can bury smaller channels. At the expert level: the two-stage ML pipeline (candidate generation via co-visitation, then ranking) explains why inconsistent channels become 'null candidates,' why videos can explode days after upload (explore-exploit tradeoff), and why your real competition is your own channel's historical baseline — not other channels.