If you start a hike at 200 m above sea level and finish at 900 m, then at some moment you were at exactly 500 m. You cannot get from 200 to 900 without passing through every height in between, as long as you never teleport. A continuous function is a hike with no teleporting: on [a,b] its graph has no holes, jumps or asymptotes, so it takes every value between f(a) and f(b).
The most common use is with N=0. For f(x)=x3−x−1, f(1)=−1 and f(2)=5: the graph starts below the axis and ends above it, so it crosses somewhere in (1,2). Check the midpoint: f(1.5)=0.875>0, so the crossing is in (1,1.5). Then f(1.25)≈−0.297<0 moves it to (1.25,1.5). Each step halves the interval; this is the bisection method, and it closes in on the root x≈1.3247.
The guarantee has limits. It promises at least one c, not exactly one; a wavy graph can cross the axis several times between endpoints of opposite sign. It needs continuity: x1 has f(−1)=−1 and f(1)=1 but is never 0, because the asymptote at x=0 lets the graph get from negative to positive without crossing the axis. And endpoints of the same sign prove nothing either way: the graph might cross twice or not at all.
Think of it like this. A thermometer that warms from 15 °C to 25 °C during the morning must read exactly 20 °C at some instant, because temperature cannot skip numbers. The IVT says the same of every continuous function. The picture fails for quantities that do jump, like a price tag or a bank balance at the moment of a deposit, and that is exactly when the theorem does not apply.
Core idea. If a continuous function starts below a level and ends above it, it must hit that level somewhere between.