Quality assessment of coagulation tests is carried out within Russian Federal External Quality Assessment Scheme (FEQAS) throughout the 20 years of its existence, from 1995 onwards. In 2014 the EQA scheme "Coagulology" was attended by 2094 laboratories among 7478 participants of FEQAS. Each participant in "Coagulology" receives 3 times a year for the analysis sets of two lyophilized blood plasma control samples, each of the samples should be analyzed on different days. If both samples are identical, then the following estimates are assessed: 1) the relative bias of the mean of four measurements, obtained in duplicate from each control sample; 2) between-day range between the mean of the two duplicate measurements of two control samples, performed on different days; 3) the average within-day range between two duplicate measurements of control samples performed in one day. The bias is an indicator of the accuracy of techniques, between-day range - its precision, and within-day range - its repeatability. If the set contains two different control samples (normal and pathological), the bias and within-day range are estimated separately for each sample, and between-day range is not calculated. The results of external quality assessment have been statistically processed. The analysis of variance (ANOVA) allows to estimate the contribution of each source of error in the total analytical error. Besides intralaboratory errors the difference between the sets of reagents from different manufacturers has significant impact on result of a measurement. In case of INR determination the contribution of the last source of error makes about 25% in the total sum of squares of the uncertainties. But the main contribution (about 60-75%) is provided by interlaboratory variation, i.e. spread of systematic bias in different laboratories. Contribution of between-day variation ranges from 9 to 18%, and within-day variation (random error of the instrument) - from 2 to 4%. Thus, the most important source of analytical errors in INR determination is calibration of instruments in the laboratory and/or homogeneity of reagents/calibrators from the same manufacturer. And the interlaboratory coefficient of variation is typically less than 10% on normal samples, but increases to 20% on pathological samples with increased INR.