The constant M in |f(b) − f(a)| ≤ M|b − a| is the worst-case sensitivity of the output to the input: it is the largest possible change in f per unit change in x. In applied modeling, M is often the only thing known about a system — a maximum thermal expansion coefficient, a maximum stretch of a spring, a tolerance on a sensor gain. The Mean Value Theorem converts that single number into a guaranteed band around the output, so a small input change can be shown to produce an output change no larger than M times that input change.
Rolle's Theorem: The Flat-Point Guarantee and Where It Pays Off
Error Bounds, Sensitivity, and Applied Uses
Reading the Bound as Sensitivity
2 / 5
That same inequality has a second reading. M is not just a slope ceiling — it is the worst-case sensitivity of the output to the input. It answers the question: how much can the output move for each unit the input moves?
0:00 / 0:00