Engagement-based ranking is the practice of ordering the items in a feed by how likely a user is to engage with them - to click, watch to completion, like, comment, share, or otherwise react - instead of showing them in simple reverse-chronological order. Most large social platforms moved to this approach because a feed sorted by predicted engagement keeps users on the platform longer than one sorted by recency.
The mechanics are recommender-system mechanics. A model is trained on a user’s past behavior and the behavior of similar users to predict, for each candidate post, the probability of each kind of interaction; those predictions are combined into a score, and the highest-scoring items go to the top. TikTok’s official description of its For You feed lists user interactions, video information, and device settings as inputs, and notes that a strong signal such as finishing a longer video is weighted more heavily than a weak one such as the viewer and creator sharing a country. Facebook’s feed evolved from the fixed EdgeRank formula to machine-learning ranking over many signals.
The approach is also the focus of public debate. Critics, including former Facebook employee Frances Haugen, have argued that optimizing for engagement can amplify divisive or sensational content because such content draws strong reactions. Meta’s own 2020 US election studies, published with academics in 2023, found that experimentally switching consenting users to a chronological feed reduced time on the platform but did not measurably reduce affective polarization or improve political knowledge - a result the company highlighted and some outside researchers disputed.
Why business readers should care: engagement-based ranking is the economic engine of the attention economy. Understanding that a feed optimizes for predicted engagement, not for relevance or accuracy, clarifies both why platforms are so effective at holding attention and why their effects are contested.