Twelve case stories were mailed to 273 general practitioners to study the clinical assessment of haemoglobin values. For each case the general practitioner was asked to fill in the haemoglobin value representing the minimal change from a given value considered necessary to take action. haemoglobin change corresponding to the median haemoglobin value stated was transformed to a so-called 'medically useful coefficient of variation' in order to relate clinical demands to analytical and biological variation. The average medically useful coefficient was calculated to 3.9% (range 2.3-7.8%). We found that general practitioners assess haemoglobin values rather uniformly, although their judgement varies substantially with the clinical situation; they are not fully aware of the consequences of analytical and biological variation, and the advantage of knowing a previous haemoglobin value is not recognized. The analytical imprecision of haemometers based on clinical demands should be 2.8%, and the analytical quality should be the same in primary and secondary care.