Drop 1,000 marbles through a board of pegs. At each peg a marble bounces left or right at random. To end up far to the right it must bounce right almost every time, which is rare, so most marbles collect a mix of lefts and rights and land near the middle. The pile at the bottom is a bell: tall in the center, thinning evenly toward both edges. Any quantity built from many small, independent pushes, like adult height (many genes, diet, sleep) or the error in a measurement, piles up the same way.
Two numbers describe the whole bell. The mean μ is the center and the peak. The standard deviation σ is the distance from the center to where the curve changes from bending down to bending up; a small σ gives a tall, narrow bell and a large σ a low, wide one. The total area under the curve is always 1, meaning 100% of the data, and the share of values in a range is the area over that range. That is also why, for a continuous normal variable, the chance of one exact value such as 170.000 cm is 0.
Not every hump is normal. Waiting times at a clinic might peak near 10 minutes but stretch out to an hour on the right: one peak, but lopsided, so a normal model would underestimate long waits and could even predict negative ones. Before using the curve, check that a histogram is roughly symmetric with one peak and no long tail.
Think of it like this. A steady archer's misses behave like a normal curve: most arrows land near the aim point, big misses are rare, and misses to the left are as common as misses to the right. Where the archer aims is the mean; how steady the hands are is the standard deviation. Unlike real arrows, the model allows a tiny chance of a miss any distance away.
Core idea. When many small, independent effects add up, values pile into a symmetric bell fixed by its mean and standard deviation, and the area under that bell over a range is the probability of landing in it.