Oral anticoagulation therapy is monitored by the use of international normalised ratio (INR). Patients performing self-management estimate INR using a coagulometer, but studies have been partly flawed regarding the estimated precision and accuracy. The objective was to estimate the imprecision and accuracy for two different coagulometers (CoaguChek S and XS). Twenty-four patients treated with coumarin were prospectively followed for six weeks. INR's were analyzed weekly in duplicates on both coagulometers, and compared with results from the hospital laboratory. Statistical analysis included Bland-Altman plot, 95% limits of agreement, coefficient of variance (CV), and an analysis of variance using a mixed effect model. Comparing 141 duplicate measurements (a total of 564 measurements) of INR, we found that the CoaguChek S and CoaguChek XS had a precision (CV) of 3.4% and 2.3%, respectively. Regarding analytical accuracy, the INR measurements tended to be lower on the coagulometers, and regarding diagnostic accuracy the CoaguChek S and CoaguChek XS deviated more than 15% from the laboratory measurements in 40% and 43% of the measurements, respectively. In conclusion, the precision of the coagulometers was found to be good, but only the CoaguChek XS had a precision within the predefined limit of 3%. Regarding analytical accuracy, the INR measurements tended to be lower on the coagulometers, compared to the laboratory. A large proportion of measurement of the coagulometers deviated more than 15% from the laboratory measurements. Whether this will have a clinical impact awaits further studies.