A graph showing the percentage of viewers still watching at each point in a video — the primary signal algorithms use to rank video.
The retention curve is the X-axis-is-time, Y-axis-is-percent-of-viewers chart that YouTube Studio (and TikTok analytics, and Reels insights) shows for every video. A perfectly flat curve at 100% would mean nobody dropped off — impossible in practice. Real retention curves drop sharply in the first 3 seconds, then either decay slowly (good) or fast (bad).
Algorithms weight retention heavily. A video with 50% average retention beats a video with 30% retention at the same view count, every time. The shape matters too — spikes (rewatches) signal high-value moments; cliffs (mass drop-offs) signal where to cut next time.
The first-30-seconds retention number is the most predictive single metric. Above 70% three-second retention generally means the video will get pushed; below 40% generally means it won't.