Client

• The platform offers 1.3 million products from 7,000 partners
• 6 million daily users
• The recommendation engine suggests accessories and related products to customers as a vital way to increase average purchase value and CTR in the online store

Challenge

• Needed to improve performance of the recommendation engine
• The recommendation model requires large amounts of labeled product data for training
• An automated solution for training the recommendation model wasn’t performing well

Solution

1. Automated pre-labeling sorted items into related products and accessories
2. Tolokers matched products to accessories and related items in separate datasets
3. Labeled data was reviewed and verified by Tolokers to ensure quality
4. The labeled dataset was used to train a gradient boosting model

“We had two goals: to get high-quality data for training the recommendation system for our ecommerce platform, and to measure the accuracy of our current recommendation algorithm. Toloka helped us improve our model with super fast labeling of tens of thousands of products from our store. Toloka makes the data problem easier so that we can focus on our algorithms. ”

– Ivan Lapitsky, Project manager, Yandex.Market