Rules for the Use of Recommendation Technologies on the Ivi Service

The Ivi Service uses Recommendation Technologies.
This document explains the principles behind the operation of Recommendation Technologies. We aim to provide every User with the best individual content offering. To this end, the Service uses recommendations that, through technical means, select content suited to the interests of a particular User.

1. Terms and Definitions

Administrator - the owner and rights holder of the Ivi Service.
Video Content - audiovisual works in digital formats (feature, documentary, popular science, educational, animated, film, TV, and video productions, and series).
Video Fragment - a small excerpt of Video Content used to promote Video Content in the Administrator's «Potok» system.
User - a person using the Ivi Service.
Rules - these rules on the use of recommendation technologies on the Ivi Service.
Viewing or watch - access to receiving a stream, over the Internet, of Video Content, Trailers, and Video Fragments.
Recommendation Technologies - information technologies for providing information based on the collection, systematization, and analysis of data relating to the preferences of Internet users located in the Territory.
Service or Ivi Service - the Ivi audiovisual service, which provides the ability to watch Video Content free of charge or for a fee.
System - software, the exclusive right to use which belongs to the Administrator, including a recommendation system that uses machine learning.
Territory - all countries where the Ivi Service is available and operates, except for the Russian Federation, Armenia, Azerbaijan, Belarus, Kyrgyzstan, Moldova, Tajikistan, Turkmenistan, Uzbekistan, and Georgia, as well as any other countries specified on the Service.
Trailer - a short video consisting of brief and the most striking excerpts of Video Content, used to promote Video Content.
Devices - various user devices (computers, laptops, mobile phones, Smart TVs, set-top boxes) that provide access to the Ivi Service via a website or dedicated software (an application) installed on the Devices.

2. Purposes and conditions for the Use of Recommendation Technologies

2.1 Recommendation Technologies are used on the Ivi Service, on all pages and on all Devices.
2.2 Recommendation Technologies help to:
• take into account the interests and preferences of Users when selecting Video Content and Video Fragments;
• improve the user experience;
• reduce the time spent searching for Video Content and Video Fragments in line with User preferences;
• increase User engagement when using the Ivi Service.
2.3 Recommendations are generated individually for each User. No special action is required to activate recommendations.
2.4 The Administrator collects, systematizes, and analyzes data relating to User preferences, and provides recommendations for Video Content and Video Fragments based on this data.
2.5 The Administrator does not collect, systematize, process, or use Users' personal data when applying Recommendation Technologies.

3. Types and sources of data taken into account when applying Recommendation Technologies

The System takes into account User actions on the Ivi Service.
The following types and sources of data are of key importance for building Recommendation Technologies:
• Views: information about the genres and type of Video Content, Video Fragments, and Trailers previously viewed by the User;
• Purchases: information about the genres and type of Video Content previously purchased by the User for a fee;
• Additions to favorites: the genre and type of Video Content added by the User to their favorites list;
• Ratings: ratings and reviews of Video Content left by the User on the Ivi Service;
• Search queries: the User's search queries on the Ivi Service are analyzed;
• Viewing speed: the speed at which Video Content is viewed may indicate the User's level of interest;
• Time of day and day of the week: taking into account the User's activity at different times of day and days of the week;
• Active feedback: Users may hide information about Video Content suggested by the System that does not suit them; this data is collected and taken into account by the System;
• Passive feedback: the System keeps track of which recommendations Users ignore, in order to offer more suitable options.
The System may also collect and take into account the following data: the User's identifier on the Ivi Service; pages visited; the number of page visits; any actions with Video Content; and the User's country.

4. Processes and Methods for Collecting Data

Data is collected using the Groot event collection system. Data is stored on the Administrator's servers.

5. Processes and methods for systematizing data

The System categorizes Video Content and Video Fragments based on genre, cast, director, screenwriter, creators, year, country of release, and other parameters.
The System organizes Users into profiles based on their preferences, additions to favorites, purchases, Views, ratings, feedback, action history, and other characteristics.

6. Processes and methods for analyzing data

6.1 Collaborative filtering

The actions of a specific User are analyzed to find similar User profiles. The User is offered Video Content and Video Fragments that are of interest to similar User profiles.

6.2 Content-based filtering

The genre, plot elements, creators, and cast of Video Content that interested the User are analyzed in order to recommend Video Content with similar content elements to the User.

6.3 Deep learning

Machine learning algorithms, including neural networks, are used for more accurate predictions and recommendations.

6.4 Ensemble machine learning meta-algorithms

The analysis methods listed in items 6.1-6.3 are combined to achieve the highest possible quality in generating recommendations.

7. Providing the User with Information based on data obtained through the Use of Recommendation Technologies

Once the System has collected, systematized, and analyzed the data obtained through the use of Recommendation Technologies, the User is shown the following recommendations:
On the Ivi Service home page: the System provides personalized recommendations of Video Content and Video Fragments on the Ivi Service home page.
Recommendations on Video Content viewing pages: on the Video Content viewing page, the System suggests other Video Content and Video Fragments with similar content elements.
Recommendations on the Video Content search page: for a text query entered in the "search" field, if there are several options matching the query, the System ranks Video Content according to the User's preferences.
Recommendations in the catalog: the catalog of Video Content and Video Fragments may be sorted by the System according to preferences, in order to reduce the time User spends searching for Video Content.

8. Model evaluation and updates

8.1. Performance evaluation

The System continuously evaluates how well recommendations match Users' actual preferences, using metrics such as CTR (click-through rate) and conversion to Viewing.

8.2. Model updates

The System is regularly updated to reflect new data and changes in User preferences.

9. Compliance with rights and legislation

Recommendation Technologies on the Ivi Service do not infringe the rights or legitimate interests of individuals or organizations. The Administrator also does not use Recommendation Technologies for the purpose of providing information in violation of applicable law.

10. Address for legally significant notices

Address for legally significant notices: legalreport@ivi.tv.