Watching how early engagement, watch time, and rewatches shift a video from testing to full distribution would reveal what actually matters. Tracking where users scroll past, rewatch, or instantly swipe away would act like behavioral heatmaps showing interest without explicit feedback. Behind-the-scenes logs of failed recommendations and corrected signals would highlight where the algorithm misunderstands intent or context. These insights could help creators and platforms refine hooks, timing, and content structure for better retention.