This paper starts with reviewing the basis of civil and mechanical engineering practices, focusing on the need for engineers to know the maximum effects induced by applied loads, including seismic loads. In this context, the response spectrum appeared historically as the most convenient tool for representing the seismic input motion; especially in design standards. Pro and cons of this model are discussed, with an emphasis on industrial facilities, for which two major difficulties appear, for transferring the input motion through the supporting structure towards the equipment and generating realistic input motions when necessary for the analysis of cases with inherent non-linear response. Another option advocated in this paper is that a power spectral density format is used for seismic input motion modelling. Outlines of the approach are presented as well as candidate input motion; its pro and cons are discussed. Compared to the response spectrum, the method simplifies dramatically the generation of compatible time series and the calculation/computation of seismic motion transferred to equipment. It makes easy the calculation of seismically induced fatigue type damage. Calculation of the maximum of the response is less straightforward than with the response spectrum, but the case is resolved by using peak factors.