The clinical decision to proceed with prophylactic platelet transfusions is widely based on trigger points for platelet counts being equal to 20, 10, or even 5 x 10(9)/L. But an increasing number of publications show evidence that the conventional automated platelet counting methods are unable to provide consistently accurate results in this lower thrombocytopenic range. These measurement errors are mainly associated with the most commonly used impedance principle; optical methods seem to be more precise. The problems of counting imprecision in the low thrombocytopenic range can be avoided with direct or indirect immunological counting methods using monoclonal antibodies or by time-consuming manual procedures. But how should new counting procedures be evaluated? Which method should be used as the "gold standard" for platelet counting? A way out of this apparent dilemma is the application of a statistical procedure as proposed by Gautschi et al. This mathematical model allows a reference method independent evaluation of new methods by calculation of the limits of detection (LD) and limits of quantification (LQ) based on the imprecision profile of the investigated method. Using this evaluation procedure, it can be shown that immunological automated counting methods can provide reliable, sufficient, and prompt platelet counts, especially in the thrombocytopenic range.