The misclassification matrix offers a powerful way to establish how reliable are the measurements of a geotechnical property and how well defined are the categories in a classification scheme. Using soil strength in a case history, poor agreement was generally observed between the field description of soil consistency and the undrained shear strength as measured in quick undrained triaxial tests on ‘undisturbed’ samples. A misclassification matrix was used to show the disparity between description and measurement. Errors in the description of strength may have accounted for discrepancies of one class but not of two and three classes. This suggests that the sampling process tended to result in weakening probably due to stress relief and sample disturbance effects.