This paper presents evidence on how the variability of machine parts can affect the throughput of an assembly line. For this purpose, a novel model based on mini-terms and micro-terms has been introduced as a machine subdivision. A mini-term is a cycle time subdivision that can be selected by the user for several reasons: the replacement of a machine part or simply to analyse the machine more adequately. A micro-term is a miniterm subdivision and it can be as small as the user wishes. Therefore, the cycle time of a machine is the sum of mini-terms or the sum of the micro-terms. This paper focuses its attention on a welding line in a Ford Factory located in Almussafes (Valencia) where a welding unit was isolated and tested for some particular pathologies. This unit is divided in three mini-terms: the robot motion, the welding motion and the welding task. The cycle time of each mini-term is measured by changing the deteriorated components for others in the time. The deterioration of a proportional valve, a cylinder, an electrical transformer, the robot speed and the loss of pressure are tested within a range that cannot be detected by alarms and maintenance workers, that is, the range of normal production. The real welding line is modelled and a novel simulation algorithm is created based on mini-terms. The experimental measurements are introduced in the simulation model and the effect of the pathologies in the production rate is computed. As a result, the pathologies with greater variability have a deeper impact in the production rate mainly due to the bowl phenomenon effect. On the contrary, the pathologies with low variability have a low effect in the production rate. In fact, this paper demonstrates that the maximum rate capacity can be achieved if the machine variability is near zero.