Determination of the energy input for the magnetospheric energy budget is a nontrivial matter. As no direct means to measure the input are known, various solar wind-derived proxies have been developed. In this article we discuss one of the most widely used energy input functions, the so-called epsilon parameter of Akasofu. While practice has shown it to be a very useful parameter, there is no convincing evidence that it is superior to all other coupling parameters. Furthermore, its somewhat unclear definition and lack of physical foundation sometimes lead to confusing interpretation of the parameter in practical studies of magnetospheric energy cycle. For example, the parameter is sometimes understood to describe the transfer of solar wind Poynting flux into the magnetosphere, whereas the actual physical energy transfer involves conversion of solar wind kinetic energy to magnetic energy measured inside the magnetopause. Another questionable interpretation is to relate the size of the energy transfer region to the length of the reconnection line, as the scale factor in epsilon has the physical unit of area. These confusions may partly result from mixing the concepts of energy source and energy transfer. In spite of these problems the present empirical formulation of the epsilon parameter appears, from the global energy budget point of view, to give a remarkably good estimate for the total energy input into the inner magnetosphere in substorm and storm timescales. This is even more remarkable as after the parameter was first formulated we have learned that the ionosphere is a major sink of storm and substorm energy, exceeding the ring current in importance as an energy output channel. An additional issue is the energy carried away by the plasmoids and outflow of the postplasmoid plasma sheet. One can argue that the application of epsilon should be restricted to the energy consumption in the inner magnetosphere. However, as the intermittent plasmoid releases are essential parts of the same complex of processes as the ring current enhancement and ionospheric particle injections, we argue that they should be included in the energy budget, even if that might result in rejection of the epsilon as a useful input parameter. The recent analyses of energy output suggest that we can still use epsilon by scaling the parameter up by a factor of 1.5--2. It should be noted, however, that this energy budget does not account for all energy passing through the magnetosphere but only that part which is consumed in the storm and substorm processes.