What Is a Box Plot? Definition and Uses

A box plot — box-and-whisker plot, formally — compresses an entire distribution into a five-number picture: the median, the two quartiles that bound the middle half of the data, and whiskers reaching toward the extremes, with outliers plotted individually beyond them. Its particular genius is comparison. Line up one box per machine, shift, or supplier, and differences in center, spread, and strangeness become visible in seconds.

How it works

The box plot was popularized by the statistician John Tukey in the 1970s as part of exploratory data analysis — the school of looking hard at data before theorizing about it. The box spans the interquartile range, the middle fifty percent of values. The line inside it is the median. The whiskers conventionally extend to the most distant points within one and a half times the interquartile range, and anything beyond earns its own dot: a candidate outlier with a story worth hearing. The format also travels well into reports, because it carries its own summary statistics on its face.

Where histograms excel at showing one group’s shape in detail, they turn cluttered when five groups need comparing. Five boxes side by side stay legible, which is why the box plot is the default first picture whenever a Six Sigma team suspects that some factor — line, lot, location — separates the data.

A worked example: three suppliers, one surprise

A manufacturer buys the same component from three suppliers and plots each one’s delivery lead times. Supplier A: a median of five days, short box, short whiskers. Supplier B: the same median of five days, a box twice as tall, and a scatter of high outliers. Supplier C: a median of seven days, tight box. On the average-lead-time report, A and B looked identical. The plot says otherwise — A is dependable, B is a planning hazard with the same average, and C is slower but keeps its promises. Purchasing stops asking who is fastest and starts asking who is predictable.

• Reach for box plots to compare groups; reach for a histogram when one group’s exact shape matters.
• Chase every outlier dot individually — some are the project’s best clues, and some are typos.
• A median line shoved to one end of its box signals skew.
• Distrust quartiles computed from a handful of points; small samples make unstable boxes.
• Equal medians with unequal boxes is a genuine finding — variation costs money even when averages match.

Averages tell you where a process aims; box plots tell you whether to trust it.

Box plots belong to Green Belt territory. Our $299, 35-hour program uses them throughout Measure and Analyze, and its 100-question proctored exam expects you to read one fluently. Black Belt training ($499, 60 hours) takes the next step, testing whether the differences between boxes are statistically real or merely noise.