Last month I speculated that the pace of system change evolves along the S-curve. Upon second thoughts, it appears that the S-curve alone cannot correctly reflect this evolution. Indeed, if the pace of a system's improvement evolves along the S-curve, it will eventually reach some top pace after which the system will continue to improve at that pace forever. Nothing like this is observed in practice. No system improves forever. At some point all improvements cease. Moreover, in improvement of any system there are always surges and recesses. Thus, the S-curve is only a portion of the complete picture, which looks like this:
Accordingly, the graph of evolution of the total number of changes accumulated in a system looks as follows:
The latter picture consists of several jointed S-curves. However, transition from one S-curve to another one does not mean transition to another system (as is the case in the Altshuller's theory). All these multiple S-curves reflect evolution of the same system. Transition will only occur when system is close to exhaustion of its potential.1
As for system performance, the graph of its evolution is most likely a superposition of the above graph and some random process:
Here the actual performance is shown by a contiguous line. The dashed line is just the first component of the aforementioned superposition.
Such a generic graph is in accord with the experimental data obtained for some systems. For example, the Nuvolari graph of steam engines performance evolution is of the above type.
1 Notes added on March 30, 2014: After this article was written I doubted that there is a proven correlation between evolution of parameters of a system and its transition to super-system (or to some other system). These doubts were expressed in my other article titled "On the Logical Fallacy of Altshuller's Law of Transition to Super-System". Regardless of these doubts, it is undoubtful that parameters of a system do not evolve along S-curve but rather along the graph presented above.