Ozon

Tuned e-commerce search
engine with fast scaling

Image
Accelerate your
e-commerce AI
View live demo
Accelerate your
e-commerce AI
View live demo

Client

Large online retailer
  • Russia's leading multi-category e-commerce platform
  • More than 9 million SKUs across 24 categories

Challenge

Improve search engine quality by creating reference samples
  • Reference samples are used for evaluating search relevance, identifying problems in the search algorithm, and choosing the best ranking model
  • In-house manual labeling of 100 search queries showed good results but was time consuming
  • Needed to quickly scale up labeling

Solution

Toloka search relevancy project allowed for fast scaling
  1. Tolokers were trained to evaluate how well certain products match search queries
  2. Five Tolokers were assigned to each task to improve quality

Business impact

Deep analysis of search results for ongoing improvements
  • 40,000 search relevance tasks completed in one month for $1500
  • Large crowd assigned to tasks provides granular evaluations for deeper analysis
  • Ongoing process works stably with large amounts of data to continually tune search algorithms

Similar success stories

  • Improved crowdsourced translations of product descriptions.

    Results:

    17%

    budget reduction while achieving optimal quality.

    Read the storyRead the story
  • Improved the accuracy of a predictive tool using local shopping patterns.

    Results:

    30%

    improvement in app accuracy after data collection, reaching 95%.

    Read the storyRead the story
  • Enhanced performance of a recommendation engine.

    Results:

    6x

    reduction of errors in the product recommender model.

    Read the storyRead the story
  • Expanded product coverage for e-commerce price matching.

    Results:

    2.5%

    better coverage of key products.

    Read the storyRead the story

Accelerate your e-commerce AI

Let's talk about the ideal solution for your data needs.

Talk to us
Fractal