In a survey, results of more than 100 assay determinations of various analytes (active pharmaceutical ingredients, methylparabene, chlorthalidone, and proteins) at different concentration levels were collected with signal-to-noise ratio (S/N) levels between 2 and higher than 10?000. It must be concluded that without having a S/N of at least 50, repeatabilities (combination of injection, separation process, and integration) of 2% cannot be achieved, in contrast to the general assumption that a S/N of 10 is sufficient for analytical HPLC. These data now confirm an earlier assumption with a much broader fundamental of measurements. The empirical functions %RSD = 58/(S/N) + 0.30 (for HPLC data) and %RSD = 73/(S/N) + 1.07 (for CE data) could be derived. It was shown that a S/N of greater than 100 is necessary for optimal precision. Before optimizing HPLC or CE methods, for example, for sample pretreatment, S/N > 100 should be the prerequisite, else optimal precision will not be achieved. Only after the detection-related scatter is sufficiently reduced, other sources of variation can be successfully tackled.