Instagram improves feed ranking by separating baseline from incremental boosts
Baseline and Incremental: User, Item, and Request Debiasing in Instagram Feed Ranking
Information RetrievalMachine Learning
Summary
Instagram recommends posts by predicting how much users will engage with them. The authors found that these predictions mix up a base level of interest with the extra attention a post might get. By keeping these two parts separate and weighting them differently, Instagram’s feed ranking performs better. This new approach was tested through experiments and has been used on Instagram for a year, improving user engagement. Meta has also adopted this method across its apps to help keep people using them longer.
What this means in practice
- •For mobile app developers: Improve content recommendation accuracy by separating baseline user/item effects from incremental engagement signals in ranking models.
- •For advertising platform engineers: Enhance feed ranking systems by applying parameterized debiasing operators to better balance habitual and contextual engagement predictions.
Authors
Hsin-Chiang Chen, Hao Wang, Gedi Zhou, Mingze Gao, Abhishek Kumar, Yiben Yang, Haoyu Wang, Hongjing Xia, Bo Chen, Zhengyu Su, Xian Sun, Chang Xu, Kevin Han, Weiwei Li, Zhongchang Sun, Yao Zhang, Qifei Wang, Peggy Yang, Yihan Wu, Haoran Liu, Siyang Sun, Xingyu Pan, Shannon Phu, Haoming Chen, Kevin Zhao, Chen Liang
Abstract
A feed ranker scores candidates with a value model, a weighted sum of engagement predictions. Each prediction conflates a baseline, what this viewer, item or request would produce anyway, with the incremental amount by which this candidate exceeds it. The literature mostly removes the baseline. We argue that a multi-signal value model should keep the raw prediction and add the incremental part as a separately weighted term. On Instagram Feed, three debiasing families deployed independently, for viewer habit, item popularity and request context, are one parameterised operator, and a design procedure picks a point in it. Media-type (format) bias goes instead to a debiased label chosen by the same procedure. In a controlled four-arm ablation of deployed Feed terms (neither term, each alone, both), the two-term arm beats both single-term arms at production weights on point estimates in 11 of 13 judged (term, metric) pairs. Of the 26 direct contrasts, 17 are significantly for it and only 1 against it, on impressions. The ablation also suggests that the offline diagnostic should screen signals, not rank them. Such terms have served Instagram Feed production for most of a year. Beyond Feed, Meta, whose apps serve billions of people, has adopted the operator across surfaces, apps and both ranking stages. Its launches there drive sessions and daily active users, the core visitation and retention metrics.