Histogram

A histogram is the simplest way to look at a pile of measurements and learn something. Drop the numbers into a few bins, draw bars over them, and the process starts to talk back. A bell centered on target says the process is healthy. A bell pushed off center says it is drifting. Two humps say two different things are happening and being averaged into one chart.

"The shape of the bars is the voice of the process. Listen to what the shape is telling you."

How a histogram works

A histogram is built from a sample of measurements taken from the same process. Each measurement is sorted into a bin, a small range of values, and the height of the bar above that bin equals the number of measurements in it. With thirty or more data points, the bars start to form a recognizable shape.

The shape carries the diagnostic information:

• A symmetric bell centered on the target value means the process is in control and capable.
• A bell shifted off center means the process is producing on average but in the wrong place.
• A flat or rectangular shape means too much variation, no central tendency to count on.
• Two distinct humps mean two different conditions are mixed in the data, two operators with different methods, two material lots, two shifts.
• A long tail means rare outliers are present, often the most expensive defects in a run.

A histogram is a one-time snapshot. To know whether the shape stays the same over time, you also need a control chart. The histogram is descriptive, not predictive. Read it for shape, not for trend.

Where a histogram fits on the shop floor

Imagine a 30-person plastics injection molding shop running a household clip on three identical presses. Customer complaints have been creeping up over the past month about a critical hinge dimension being too loose. The owner has assumed it is one operator or one press, but every quality check looks fine on average.

A histogram makes the problem visible in an afternoon. The shift collects fifty measurements from each press over two days and plots three histograms side by side. Two of the presses show tight, centered bells. The third shows a flatter, slightly shifted distribution with a hump on the high side. That third press has a worn cooling channel that is producing parts inconsistently, and the average had been hiding it because the high outliers were balancing the low ones.

That is the work a histogram does. It exposes mix, drift, and spread that any single measurement or average will quietly bury. Forty minutes of paper and pencil pays off in weeks of avoided returns.

Common mistakes with histograms

• Too few data points. Anything under thirty measurements is noise. Wait until you have at least fifty before drawing conclusions.
• Wrong bin width. Bins that are too wide blur a bimodal shape into a single hump. Bins that are too narrow break the data into spikes. A reasonable starting point is the square root of the number of measurements.
• Confusing histogram with control chart. A histogram has no time axis. It cannot tell you whether the process is stable, only what it looked like during the sample.
• Reading the histogram as the answer. A bimodal shape is a question, not a conclusion. The work is to investigate what two conditions are being mixed.
• Ignoring the spec lines. Drawing the upper and lower spec limits on the histogram is the move that turns the chart from descriptive to diagnostic. Without spec lines, you cannot tell if a healthy-looking bell is actually inside tolerance.

Histogram and related Lean tools

A histogram is one of the seven basic quality tools and pairs naturally with a control chart, which adds the time dimension a histogram lacks. The data that feeds a histogram usually comes from a check sheet, the simple form operators fill in as they collect measurements. When the histogram suggests that one variable might predict another, the next step is often a scatter diagram to test the relationship.