Empower search and recommendations with human-labeled data for +15% revenue growth
Help shoppers find the right products easily with Toloka-powered search and recommendation systems. Increase sales by making sure customers find what they want in 2 clicks.
I want to find out how human labeled data can improve search relevance and revenue by 15%
Search matters*
99% of users
return to a website with
good search53% of users
buy additional items after
they find what they need$300B/year
estimated losses of US retailers
attributed to poor search
performanceCustomer expectations
Customers get easily frustrated if they can't find an item or there are no recommendations for related products
Today's search is losing 8–15%
of potential revenue
- Modern search and recommender systems rely heavily on historical and behavioral data (clicks, views, scroll time, time on item page).
- Quality of the search and recommendations is measured by scroll time, time on item page and conversions to purchase.
All that creates a good enough search, especially if organic data are high quality and conversions are happening. The illusion is that everything is fine, but in reality sales are being lost.
How search quality fails*
8.3%
of unsuccessful
search sessionsWrong category
prediction15%
of unsuccessful
search sessionsWrong sorting
of products10%
of unsuccessful
search sessionsIncorrect typo
detection18 million
unsuccessful search sessions resulting in lost sales
60 million monthly visits, 2 million products shipped per month.
Irrelevant search results — wrong
sorting order
Incorrectly defined product category
Relevant product results omitted
Synonyms
Misspelled queries
Toloka's approach
Real people add labels to a data item:
- Spot check search results to evaluate search quality and train the model
- Add recommendations for product discovery to avoid the "recommendations bubble"
- Create attributes for items (color, brand, type, new categories)
- Maintain glossaries of synonyms, slang, and local variations of product searches
- Compare competitor pricing and search quality
How human-labeled data benefits e-commerce systems
40%
improved accuracy of search
and recommendation systems70%
increased recall
on related products8-15%
additional revenue
from Search
Media about us
Industry compliance
We are dedicated to best practices in information security to meet and exceed industry standards.
ISO 27001
Toloka implements an information security management system (ISMS) that was audited and certified in compliance with the ISO/IEC 27001 standard. The international standard provides a framework for information risk management to guarantee confidentiality, integrity and availability of data.ISO 27701
Toloka's privacy information management system (PIMS) is certified for compliance with ISO/IEC 27701, an extension to ISO/IEC 27001. This standard ensures the secure collection, handling, storage, and destruction of personally identifiable information (PII) on the Toloka platform.GDPR
The Toloka platform strives to comply with key requirements of the General Data Protection Regulation (GDPR). We prioritize the privacy of our users and make every effort to protect personal data by following established internal processes.
You're in good company
Toloka supports a community of data scientists, ML engineers, researchers,
and AI innovators around the globe to accelerate machine learning with better data processes
Trusted by leading
ML & AI teams