The attenuation provided by hearing protection devices (HPDs) is traditionally determined under laboratory conditions by means of a standardised method titled REAT (Real Ear Attenuation at Threshold). This performance is often overestimated compared to what has been measured so far in industrial reality. Concerning the earmuff, the formable earplug and preformed earplug type of HPD, there are numerous data related to the discrepancy between, on one hand, the laboratory measured attenuation which is labelled by the manufacturers and, on the other hand, the in situ estimated attenuation. But, few exist concerning custom-moulded earplug (CMEP) type protection devices. This paper presents the practical application of measurements intended to estimate the attenuation of this type of protection device while it is being worn by users during execution of their task. Three manufacturers of hearing protectors took part in this initiative; this participation involved the manufacturing of custom-moulded earplugs with a miniature microphone inserted. The measurements were carried out on 63 employees in nine industrial sites with diverse activities. The estimate of this in situ attenuation, in this study called level reduction (LR), was obtained from the difference between the noise exposure level measured close to the ear of the employee and the residual noise level measured under the earplug. The CMEP level reduction estimated in situ highlighted significant variability characteristics. The results of the measurements confirm the overestimation of the values indicated by the manufacturers. From 3 to 5 dB at high frequencies, the discrepancy can reach 8-10 dB at medium and low frequencies. The study also confirms that CMEPs which have been tested here, like other earplugs types, are unsuited to attenuate low frequencies.