To achieve this mission, we are building search products that provide a personalized discovery and search experience based on a consumer’s past search and order history with DoorDash. This article details our approach for personalization in our search ecosystem, which has provided a significant lift in conversion from search to checkout.
Search and Recommendation Challenges
The three-sided nature of DoorDash platform (involving consumers, dashers and merchants) presents a lot of interesting and unique search challenges in addition to the general search and recommendation related problems. Some challenges include:
- Sparsity: not every consumer can see every store, making this different from a typical e-commerce recommendations problem
- Cold-start problem: cases when new stores or consumers enter the system
- Tradeoff between relevance versus diversity
- Including accurate driving distance in search selection