In this video I walk you through some of the specifics of the YouTube Algorithm. My background as a data scientist can help you better understand this algorithm and increase your chances of having it pick your videos up.
The algorithm wants you to watch YouTube for as long as possible. It prioritizes videos that have a high click through rate (CTR) and high watch time. When you combine these two metrics you get Expected watch time. This is the variable that the algorithm is trying to optimize for with each recommendation.
The model uses a two phased deep neural net architecture. The first phase takes millions of videos and filters them to less than 1000 relevant videos. The next phase ranks the smaller group of videos based on more specific metrics that youtube has access to.
Comments