Session 9: Timely Personalization at Peloton:System and Algorithm for Boosting Time Relevant Content
RecSys 2022
by Shayak Banerjee (Peloton Interactive, Inc., United States), Vijay Pappu (Peloton Interactive, Inc., United States), Nilothpal Talukder (Peloton Interactive, Inc., United States), Shoya Yoshida (Peloton Interactive, Inc., United States), Arnab Bhadury (Peloton Interactive, Inc., United States), Allison Schloss (Peloton Interactive, Inc., United States), Jasmine Paulino (Peloton Interactive, Inc., United States)
Peloton has a subscription-based service offering access to a rich catalog of high-quality fitness classes. Not only is there a wide diversity amidst these classes but this inventory is constantly changing. This dynamic inventory introduces a new challenge for our recommender systems – surfacing timely content in addition to relevant content. We are often faced with a set of classes that are filmed and timely only during a narrow window, for example, holiday-themed classes. During this time, they need to reach a sizable audience to satisfy business goals, while also preserving user engagement. However, naïvely surfacing timely content has the potential to hurt user engagement goals. We have to factor in individual interests when choosing who to display this timely content. Our recommender system, which is already aware of users’ interests, is best placed to balance relevance and timeliness.
In this talk, we will show how we have created a system and algorithms for artificially increasing the impressions of selected sets of classes, which we call boosting. We open up control over our recommender systems for our marketing and production partners, who are able to entered timed boosts for selected classes. We show how these boosts are then honored by both our batch and real-time recommendation engines to selectively display the boosted classes higher in rankings so as to get them more visibility. We will discuss a naïve boosting algorithm, which produced high lifts in impressions but at the cost of user engagement. We will then demonstrate a batch optimization approach that leads to a better balance between engagement and impressions. We will look at results of several A/B tests on our user base, and end with a summary of benefits of this system as well as emerging challenges.
Видео Session 9: Timely Personalization at Peloton:System and Algorithm for Boosting Time Relevant Content канала ACM RecSys
by Shayak Banerjee (Peloton Interactive, Inc., United States), Vijay Pappu (Peloton Interactive, Inc., United States), Nilothpal Talukder (Peloton Interactive, Inc., United States), Shoya Yoshida (Peloton Interactive, Inc., United States), Arnab Bhadury (Peloton Interactive, Inc., United States), Allison Schloss (Peloton Interactive, Inc., United States), Jasmine Paulino (Peloton Interactive, Inc., United States)
Peloton has a subscription-based service offering access to a rich catalog of high-quality fitness classes. Not only is there a wide diversity amidst these classes but this inventory is constantly changing. This dynamic inventory introduces a new challenge for our recommender systems – surfacing timely content in addition to relevant content. We are often faced with a set of classes that are filmed and timely only during a narrow window, for example, holiday-themed classes. During this time, they need to reach a sizable audience to satisfy business goals, while also preserving user engagement. However, naïvely surfacing timely content has the potential to hurt user engagement goals. We have to factor in individual interests when choosing who to display this timely content. Our recommender system, which is already aware of users’ interests, is best placed to balance relevance and timeliness.
In this talk, we will show how we have created a system and algorithms for artificially increasing the impressions of selected sets of classes, which we call boosting. We open up control over our recommender systems for our marketing and production partners, who are able to entered timed boosts for selected classes. We show how these boosts are then honored by both our batch and real-time recommendation engines to selectively display the boosted classes higher in rankings so as to get them more visibility. We will discuss a naïve boosting algorithm, which produced high lifts in impressions but at the cost of user engagement. We will then demonstrate a batch optimization approach that leads to a better balance between engagement and impressions. We will look at results of several A/B tests on our user base, and end with a summary of benefits of this system as well as emerging challenges.
Видео Session 9: Timely Personalization at Peloton:System and Algorithm for Boosting Time Relevant Content канала ACM RecSys
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
![Building public service recommenders: Logbook of a journey](https://i.ytimg.com/vi/zMLWxT9Om2E/default.jpg)
![Tops, Bottoms, and Shoes: Building Capsule Wardrobes via Cross-Attention Tensor Network](https://i.ytimg.com/vi/4r_xCgmTPnk/default.jpg)
![PS 5: Latent Factor Models and Aggregation Operators for Collaborative Filtering in Reciprocal Recom](https://i.ytimg.com/vi/IAWS92nUYt8/default.jpg)
![PS2: Translation-based factorization machines for sequential](https://i.ytimg.com/vi/IP7F7UcVhWM/default.jpg)
![Workshop on Context-Aware Recommender Systems](https://i.ytimg.com/vi/qAr9YJlMqGk/default.jpg)
![Paper Session 4: Domain Adaptation in Display Advertising: An Application for Partner Cold-Start](https://i.ytimg.com/vi/7llvCGTBXWE/default.jpg)
![Mitigating Confounding Bias in Recommendation via Information Bottleneck](https://i.ytimg.com/vi/pqKthq2JEsc/default.jpg)
![PS 7: Eliciting pairwise preferences in recommender systems Saikishore Kalloori](https://i.ytimg.com/vi/YgsaliIluII/default.jpg)
![PS 6: Judging similarity: a user-centric study of related item recommendations Yuan Yao](https://i.ytimg.com/vi/QQ8lLU_CYiY/default.jpg)
![RecSys 2015 Session 4b: Algorithms](https://i.ytimg.com/vi/CTAj-g-Hfz0/default.jpg)
![Boosting Local Recommendations With Partially Trained Global Model](https://i.ytimg.com/vi/z9HKZC8Vr64/default.jpg)
![Pessimistic Reward Models for Off-Policy Learning in Recommendation](https://i.ytimg.com/vi/CA6UrVNo7Wo/default.jpg)
![Learning a voice-based conversational recommender using offline policy optimization](https://i.ytimg.com/vi/b6AswbiUodo/default.jpg)
![RecSys 2020 Session P2B: Evaluating and Explaining Recommendations](https://i.ytimg.com/vi/t2hbVNyWUrc/default.jpg)
![RecSys 2020 Session P5B: Real World Applications II](https://i.ytimg.com/vi/juWT-_aK_gw/default.jpg)
![Privacy Preserving Collaborative Filtering by Distributed Mediation](https://i.ytimg.com/vi/FQ90qgq0EUU/default.jpg)
![RecSys 2016: Paper Session 7 - Behaviorism is Not Enough: Better Recommendations](https://i.ytimg.com/vi/AjcmnxM1m08/default.jpg)
![RecSys 2016: Paper Session 5 - Mechanism Design for Personalized Recommender Systems](https://i.ytimg.com/vi/zrWss_421yY/default.jpg)
![RecSys 2020 Session P5A: Real World Applications II](https://i.ytimg.com/vi/9NA5SMu0TdI/default.jpg)
![Debiased Off-Policy Evaluation for Recommender Systems](https://i.ytimg.com/vi/YOPO2u81mvg/default.jpg)