Success stories

In this section, we will tell you about successful projects implemented on the Yandex.Toloka platform.

Collecting a dataset

The benefit of the Yandex.Toloka crowdsourcing platform is that the requester can collect a dataset while spending a minimum amount of time and money.

Protecting biometric data

The company ID R&D collected a dataset that let them build algorithms to protect their data from malicious attacks.

SMS, passwords, and codes don't guarantee 100% protection against attackers. ID R&D develops security systems based on voice, facial, and behavioral recognition. Using Yandex.Toloka, the company collected a dataset of 112,000 sample photos and videos with people's faces.

For more information about how ID R&D created their database, see

Dataset in five weeks

Roman Kutsev, the founder of, created a dataset of more than 10,000 images using a chatbot and Yandex.Toloka.

The dataset is based on photos of users and their description by the chatbot. This dataset might be useful for online dating platforms, because it lets the algorithms select suitable matches for specific users more accurately.

How it works:

  1. The user uploads their photo to the chatbot.
  2. After a while, the bot gives the user a detailed review.

The owners of the bot transmit the image to Yandex.Toloka, where performers describe the characteristics of the person. Within five weeks, Roman managed to collect a dataset of more than 10,000 images.

For more information about how the process was set up, see

How to train artificial intelligence to understand the interlocutor

Sberbank trained the Event2Mind model to recognize the Russian language. After collecting the data for labeling, the Sberbank team used Yandex.Toloka to create a dataset. This dataset was used for training the Event2Mind model.

For more information about the final result, see

Dataset for 75 dollars

Neatsy's team was developing an app for ordering shoes online. The team collected and labeled 50,000 photos of feet of different sizes to teach the algorithm to automatically build a 3D model of a human foot and determine its size.

For this purpose, the company used Yandex.Toloka. By using task decomposition, the team collected a dataset in five days, spending $75.

For more information about how the team collected their dataset, see

Mass survey

Tens of thousands of performers are active in Yandex.Toloka every day. Requesters can quickly run mass surveys with a minimal budget.

Crowdsourcing and logo changes

The “Zenith” bank demonstrated an example of successful logo change. Using a focus group, the bank created two versions of their logo.

To make their final conclusions, they decided to run a survey among their target audience. With Yandex.Toloka, they managed to collect the necessary number of filled out forms within just a few hours. Based on the data from the forms, the bank chose the logo that was more memorable for the audience.

For more information about the survey and its results, see

How to update an app icon

Yandex showed an example of how to update an app icon without losing its recognizability.

Using Yandex.Toloka, the company compared how long it took performers to find the classic icon and the new icon on their phone screens. The performers needed more time to find the new icon.

This prompted the company to implement the changes gradually, only slightly changing the icon at different project stages. With such an approach, they managed to achieve a similar result.

For more information about the project, see

Which supermarkets people like

Digital Guru analyzed which supermarkets are preferred by people. For this purpose, the company identified the most popular subjects in user reviews. They analyzed more than 75,000 reviews.

To analyze such a huge number of reviews, they decided to use Yandex.Toloka. Several thousand performers analyzed the reviews. As a result of the analysis, two categories were identified. Based on the categories, the company created a table of the “best” and “worst” supermarkets.

For more information about the results, see

Yandex.Toloka blog

In our blog, we regularly publish successful projects from our requesters, post useful tutorials, and inform you about new features in Yandex.Toloka.