Method
Pointwise ranking
Score each item independently with regression or classification loss.
10 cases from 8 companies apply Pointwise ranking. 7 were checked in an online A/B test.
Typical tasks: Rank search results, Retrieve and rank from millions of items, Balance relevance and diversity, Recommend to new users and new items and 4 more. Companies: Instacart, Pinterest, Whatnot, Airbnb, Yandex, DoorDash and 2 more. Often combined with: Negative sampling, A/B test, Gradient boosting, Embedding similarity matching, Position bias correction and 27 more.
| Identify User Journeys at Pinterest | social | 2025-10 | |
| Evolving Feed Ranking at Whatnot | marketplace | 2025-04 | |
| Improving search ranking for Maps | travel | 2024-12 | |
| How Instacart Uses Machine Learning to Suggest Replacements for Out-of-Stock Products | retail | 2024-11 | |
Machine Learning Service from Scratch, or Creating Frankenstein![]() | ![]() | general | 2024-11 |
| How We Relaunched Recommendations at Auto.ru | classifieds | 2023-05 | |
Deploying Transformers in Yandex Music![]() | ![]() | media-streaming | 2023-05 |
| How Instacart Uses Machine Learning-Driven Autocomplete to Help People Fill Their Carts | e-commerce | 2022-08 | |
| Machine Learning and Personalization in Auto.ru Search | classifieds | 2022-06 | |
| 3 Changes to Expand DoorDash's Product Search Beyond Delivery | food-delivery | 2022-05 |

