Histograms show the shape of a numeric column, which a mean and standard deviation alone cannot.
Shape includes where values cluster, how far they spread, whether they lean one way, and whether there is more than one peak.
Building one means cutting the number line into bins of equal width and counting the values in each.
Most tools pick a bin width automatically, which is a starting point rather than an answer.
Trying 2 or 3 widths shows which features are real, because real patterns survive a change of width.
A long tail to the right is common for times, prices and incomes, and shows as bars that taper off slowly.
2 peaks often mean 2 different groups mixed together, such as weekday and weekend deliveries.
A spike at one value, such as 0 or 999, often points to a placeholder rather than a real measurement.
Regular spikes at round numbers, such as every 5 minutes, often show values typed by hand and rounded.
Histograms with different row counts are hard to compare, so they are often scaled to show shares rather than counts.
Overlaying 2 histograms works only when they are partly transparent, or one hides the other.
With few values a histogram can look lumpy by chance, so its shape should not be read too closely.
In practice a histogram is the first chart drawn for nearly every numeric column during exploration.
It quickly shows whether the mean is a fair summary or whether the median would be more honest.
The common misconception is that a histogram is a bar chart, when its bars are ranges of 1 number and their order is fixed.