Recently one of the readers, Alex Acosta, drew my attention to the Performance Curve Database containing graphs of various system characteristics vs time. None of them resembles S-curve. All have very irregular shape with wild fluctuations. A question naturally arises: why ? Here is an explanation.
Suppose that one decided to plot a graph of cars' fuel economy vs time. He collected a statistics on all cars that went into production during a certain period of time and plotted the graph. The result would have been a random curve with wild fluctuations. It is because he mixed various types of cars: subcompact, compact, sedans, SUVs, trucks, etc. To see it suppose that in January of 2014 a truck went into production with the fuel economy 20 mi/gallon and next to it a subcompact car went into production in February of 2014 with fuel economy of 40mi/gallon. Putting these two cars on the same graph would result in wild and random fluctuations in the graph as production dates of various types of cars are interleaved and their fuel economy differ drastically.
Authors of graphs in the Performance Curve Database made the same mistake.