Yandex launched a rating of Russian sites: the audience is calculated by a mathematical model on machine learning

After several months of beta testing, today Yandex has officially launched the Top Internet Projects in Russia - a list of 10,000 of the most popular resources among Russian users. All sites are divided into thematic categories. For example, “Habrahabr” was defined in the “Communities” category (21st place) and “Engineering and technology” (19th place).

Each resource has a weekly audience. For projects that transmit Metrics and AppMetrica data, the exact number of visitors per month, the daily audience, the average time that visitors spent on the site and in the application, and the percentage of users of the mobile application (if AppMetrica data is transmitted). In addition, the following segmentation is supported for them:

• by region;
• device types;
• user income;
• age;
• the floor.

“The rating is based on a mathematical model based on aggregated data from Yandex Browser and browser extensions,” the press service said. “If desired, sites can connect data from Metrics to enrich their statistics, for example, detailed data are already displayed on RIA Novosti, Kommersant, NTV, RBC and others.”

“The share of Yandex users in the audience of each Internet project is different - therefore, a mathematical model created using machine learning is used to calculate the audience. It determines the range of attendance of Internet projects. Next to each of them in the rating indicated the minimum score of attendance, the maximum can be seen by pointing the cursor. For the projects that allowed the use of these Metrics and AppMetrica, the exact values ​​are given, as well as daily attendance, application audience share and average time people spend on the site.

For each project in the ranking, two attendance indicators are calculated. In the first column - the number of cross-device visitors, and in the second - the number of "normal": this indicator is used in classical web analytics systems. The cross-device audience is defined by the Crypt technology, when the same user, who logs in from different devices and browsers, is counted as one person, not several.

Detailed instructions for working with a rating for advertisers are published here .