The signal in an NMR experiment is highly sensitive to fluctuations of the environment of the sample. If, for example, the static magnetic field B 0 , the amplitude and phase of radio frequency (rf) pulses, or the resonant frequency of the detection circuit are not perfectly stable and reproducible, the magnetic moment of the spins is altered and becomes a noisy quantity itself. This kind of noise not only depends on the presence of a signal, it is in fact proportional to it. Since all the spins at a particular location in a sample experience the same environment at any given time, this noise primarily affects the reproducibility of an experiment, which is mainly of importance in the indirect dimensions of a multidimensional experiment, when intense lines are suppressed with a phase cycle, or for difference spectroscopy techniques. Equivalently, experiments which are known to be problematic with regard to their reproducibility, like flow experiments or experiments with a mobile target, tend to be affected stronger by multiplicative noise. In this article it is demonstrated how multiplicative noise can be identified and characterized using very simple, repetitive experiments. An error estimation approach is developed to give an intuitive, yet quantitative understanding of its properties. The consequences for multidimensional NMR experiments are outlined, implications for data analysis are shown, and strategies for the optimization of experiments are summarized.