Enzyme immunoassays with optical detection are amongst the most widely used bioanalytical tools. We defined seven parameters for the quality assessment of immunoassays that were addressed in a systematic study of direct and indirect immunoassays, using the enzymes horseradish peroxidase (HRP) and alkaline phosphatase (AP), the chromogenic substrates 3,3',5,5'-tetramethylbenzidine (TMB) and para-nitrophenyl phosphate, and the fluorescent substrates 3-(4-hydroxyphenyl)propionic acid and 4-methylumbelliferyl phosphate. The same monoclonal antibody against caffeine was used throughout the study. The four quality parameters regarding the standard curve were the test midpoint (sensitivity), the measurement range, the relative dynamic range of the signal, and the goodness of fit of the adjusted four-parameter logistic function. All HRP immunoassays showed a higher sensitivity compared to the AP assays. On the basis of all four criteria, it was established that the direct assay format is superior to the indirect format, the immunoassay using HRP TMB fulfilling all requirements best. In a second step, caffeine concentrations in 24 beverage and cosmetics samples were determined and three more quality parameters were assessed with this application. The direct HRP TMB assay showed one of the best intra- and inter-plate precisions and the best accuracy, defined by the correlation of results with those from the chosen reference method liquid chromatography tandem mass spectrometry (LC-MS/MS). Considering all criteria, HRP TMB seems to be the enzyme substrate system of choice preferably used in the direct assay format.