What Is a Histogram? Definition and Uses

A histogram is a bar chart that shows how a single measurement distributes across its range. Values are grouped into equal-width bins, and each bar’s height counts the observations that fall inside. One glance reveals what no average can: where a process centers, how widely it spreads, and what shape its variation takes.

How it’s used

Shape is the point. A single symmetric peak suggests a stable process with routine variation. A lopsided tail signals skew — common in time data, where work cannot finish early by much but can finish late by a lot. Two peaks almost always mean two processes wearing one label: two machines, two shifts, two suppliers, mixed into one dataset. And a distribution that stops abruptly at a specification limit is a question worth asking, because nature rarely produces cliffs — sorting and rounding do.

Histograms need a reasonable amount of data — a few dozen points at minimum — and an honest choice of bin width. Shape emerges only with volume; a dozen points can imitate almost any distribution. Drawn against the customer’s specification limits, the histogram becomes the first crude capability picture: how much of the distribution sits where it should.

A worked example: the two-humped delivery chart

A courier operation quotes two-day delivery, and its average delivery time backs the claim. Then someone plots the histogram. Instead of one peak near two days, there are two: a tall one at one day and a second at four. The average was arithmetically real, and it described almost no actual delivery. Stratifying the data traces the second hump to orders routed through an overflow depot that adds handling steps. The team now has a target it could never have found in a spreadsheet of averages — and a standing reminder that a mean can be an alibi. The fix that follows is ordinary rerouting; the point is that it was invisible until the shape appeared.

• Plot the distribution before trusting any average; the mean of a two-peaked process describes nobody’s experience.
• Try more than one bin width — too few bins hide structure, too many dissolve it into noise.
• Mark the specification limits on the chart so the spread has meaning.
• Treat a sharp cliff at a limit as a finding about sorting or measurement, not a comfort.
• When you see two peaks, stratify before you act; you are looking at two processes.

An average is a summary; a histogram is a confession.

Histograms enter the curriculum at Yellow Belt ($129, 14 hours) as part of the basic toolkit, and the Green Belt ($299, 35 hours) builds on them directly — process capability analysis is essentially a histogram with judgment attached. At Black Belt, the shapes acquire names and mathematics, but the habit starts here: never trust a number you haven’t seen drawn.