This paper concerns database quality in the Keystroke Dynamics domain. The authors present their own algorithm and test it using two databases: the authors’ own KDS database and Keystroke Dynamics - Benchmark Data Set online database. Following problems are studied theoretically and experimentally: classification accuracy, database representativeness, increase in typing proficiency and finally: time precision in samples acquisition. Results show that the impact of the database uniqueness on the experimental results is substantial and should not be disregarded in classification algorithm evaluation.