This master thesis is devoted to analysis of click models for improving quality of ranking. From my perspective, this theme is actual in modern IT-industry.  Potentially results of this work could be applied to information retrieval systems as well as for online recommendation systems.   From formal point of view, this thesis is easy to read and clear for understanding. Work contains overview of the main modern click models and analysis of one approach for increasing quality of such models using clustering of the users via k-means. This idea was checked on some real dataset and results for improved approach were compared against the baseline algorithm. From my point of view, all used metrics were relevant and demonstrated that clustering of the users was really helpful for most of the click models (except for one). This result are interesting  from practical point of view and could be used for analysis of real systems.   However, there are few issues in this master thesis. First of all, there are no analysis for robustness of the clustering. This is especially important because k-means algorithm is not stable. In addition to that, it would be nice to have more detailed analysis of results for different clustering approaches. Secondly, all described metric could be considered as inner for the train set. It would be nice to have at least one train-test split for each cluster to secure results from positive bias and calculate metrics on test set. Of course, the best way would be use some relevant stratified k-fold for approach for evaluation.  So, from my perspective, this work deserves mark B (4). However, if author could fix problem with train-test split for experiment, this work should be assessed as excellent (5).  03 June 2018 Senior Data Analyst (Data Science), EPAM-systems               Sergei Smirnov