One of the current methods to detect fouling in heat exchangers is based on the analysis of the evolution of the effectiveness over time. The present results show that, using this kind of analysis, a 6% decrease of the overall heat transfer coefficient cannot be detected when measurement noise (±0.5 °C) is taken into account. On the contrary, the present study shows that the analysis of the evolution of the modulus of the variable computed using the lock-in technique is simple and leads to a very sensitive detection when it is well tuned. In this study, the excitation needed for the application of the lock-in technique is a periodical variation of the inlet temperature of one fluid. It is shown that a 4 million sample sliding observation window is necessary to obtain an accurate value of the modulus. This corresponds to a time span from a few hours up to 5 days, depending on the sampling period. The latter can then be adapted to the expected fouling rate. One important finding is that, in order to detect fouling, the excitation has to be done on the side where fouling is expected.