Last time I was covering a Yandex ranking algorithm was back in 2009, when Yandex search team released Snezhinsk. Since then there were 2 more versions of the algorithm: Obninsk and the latest one, Krasnodar.
Each version on the algorithm made SEO more and more complex, while every new algo was aiming to improve a certain area of search. Obninsk was rolled out in September 2010 in Belarus, Ukraine and Russia and was mainly aiming to improve search results for geo-independent queries. Yandex claimed the search results became 70% better (whatever that means).
10 days after Obninsk Yandex announced a number of changes in the ranking formula aiming at de-valuing if “SEO-links”.
In December the changes were implemented for geo-dependent queries.
A bit later in December came Krasnodar, the latest version of Yandex ranking algorithm. The technology behind it is called Spektr (in Russian) and the goal of it is to guess users’ intention.
Yandex claims that 20% of all their search queries are not clearly formulated, e.g. a query like “sushi” does not explain if the user if trying to find a sushi restaurant or a recipe; or a query like Pushkin could mean that the user is looking for information about the Russian poet called Pushkin or the city called Pushkin in honor of the poet.
Spektr is a technology that analyzes search queries and divides them into categories, then selecting the most popular categories and giving them more exposure in the SERPs. At the moment Spectr has around 60 categories and is expanding itself. Yandex runs this analysis several times a week to keep the categories and their content fresh.
Spektr incorporates information from various dictionaries, directories and encyclopedias, including Wikipedia, to be aware of different meanings of words.