A histogram is a graphical tool that displays the frequency distribution of numerical data by grouping values into equal-width ranges called bins and drawing a bar for each. The height of each bar shows how many data points fall within that range, revealing where the data is centered, how much it varies, and what shape the distribution takes. In quality control, the histogram is one of the 7 Basic QC Tools and is used to assess process behavior and compare output against specification limits. It turns raw measurement data into an instant visual summary that guides decisions about process capability and improvement.
Almost every quality problem you will ever face can be analyzed with at least one of the seven basic quality control tools. These tools form the foundation of process improvement in Lean Six Sigma, and the histogram is where the series begins.
What Is a Histogram?
A histogram is a bar chart that shows how often values occur within a dataset. It takes a list of numerical measurements, groups them into ranges called bins, and displays how many data points fall into each range.
The result is a visual picture of your data’s distribution. Instead of staring at rows of numbers in a spreadsheet, you get an immediate answer to three fundamental questions:
- Center: Where is my data centered?
- Spread: How much variation does it show? (Quantified formally by the standard deviation.)
- Shape: What pattern does the distribution take?
These three questions are at the heart of process understanding, and the histogram answers all three at a glance.
How to Build a Histogram
In practice, software builds the histogram for you in seconds. But understanding the manual steps ensures you know what the tool is actually doing, which makes you a better analyst.
Step 1: Collect your data. You typically want at least 50 data points, and ideally 100 or more, for the shape to be meaningful. Small samples can produce misleading patterns.
Step 2: Find the range. Subtract the smallest value from the largest value. This defines the span your bins must cover.
Step 3: Decide how many bins to use. A common rule of thumb is to take the square root of the number of data points. For 100 measurements, that gives approximately 10 bins.
Step 4: Calculate the bin width. Divide the range by the number of bins. Every bin should be the same width so the chart is not visually distorted.
Step 5: Count values per bin. Tally how many data points fall into each range. This is your frequency count.
Step 6: Draw the bars. The height of each bar represents the frequency for that bin. The bars are adjacent, with no gaps, which distinguishes a histogram from a standard bar chart.
That is the full construction process. Once you understand it, you can let software handle the calculation while you focus on interpreting the result.
How to Read a Histogram: The Shapes That Matter
Building the histogram is the easy part. Reading the shape is where the real analytical value lives. Here are the five histogram shapes you will encounter most often in quality work.
Normal Distribution (Bell Curve)
A symmetric, bell-shaped histogram indicates that most values cluster around the center and taper off evenly on both sides. This is what a stable, predictable process typically looks like, and it matches the classic normal distribution. When you see this shape, it is a good sign.
Skewed Distribution
A skewed histogram has a long tail stretching to the left or right, meaning the data is not symmetric around its center. This can be perfectly natural for some process types — cycle time data cannot go below zero, for example, so it often skews right. The key is to determine whether the skew reflects the nature of the process or a real problem.
Bimodal Distribution (Two Peaks)
Two distinct peaks in a histogram are a significant clue. A bimodal shape almost always means you are looking at data from two different sources — two machines, two shifts, two operators, or two raw material lots mixed into a single dataset. The histogram is telling you to stratify your data before drawing any conclusions.
Truncated (Cut-Off) Distribution
When the histogram has an abrupt wall on one side instead of a natural taper, that is often a sign that parts were sorted or inspected before measurement. The defective units were removed, so the histogram only shows the surviving population. This shape should raise questions about what happened to the missing data.
Isolated Bars (Outliers)
Bars sitting away from the main distribution, separated by gaps, point to outliers or data entry errors. These individual points deserve investigation before any broader conclusions are drawn.
Using Histograms for Process Capability
This is where the histogram transitions from a descriptive tool to a decision-making tool.
When you overlay your specification limits on the histogram, you can immediately see whether your process is capable of meeting requirements:
- Draw a vertical line at the Lower Specification Limit (LSL) and another at the Upper Specification Limit (USL).
- If the entire histogram sits comfortably between those two lines, your process is producing within specification.
- If bars extend beyond either limit, you are generating defects. The histogram shows you how much and in which direction.
This visual comparison is a fast, intuitive first pass at capability. When you follow it with a formal capability calculation — Cp and Cpk, or Pp and Ppk — you get a precise, quantified answer to go alongside the picture. For a deeper treatment of capability indices, see the Process Capability Analysis guide.
Practical Tips for Better Histograms
Use enough data. A histogram built from 20 or 30 points can mislead you about the true distribution shape. Collect at least 50 data points, and aim for 100 or more when possible.
Be thoughtful about bin count. Too few bins collapse real patterns into a flat, uninformative shape. Too many bins fragment the data into noise that looks like random spikes. The square root rule is a reasonable starting point, but adjust based on what you see.
Never analyze a histogram in isolation. A histogram describes the distribution of your data, but it does not tell you whether your process is stable over time. Use it alongside a control chart to confirm stability before drawing capability conclusions. A capable-looking histogram built from an unstable process is not reliable.
Histogram vs. Other QC Tools
| Tool | What It Answers |
|---|---|
| Histogram | How is my data distributed? |
| Control Chart | Is my process stable over time? |
| Pareto Chart | Which defect categories matter most? |
| Cause-and-Effect Diagram | What are the root causes of variation? |
| Scatter Diagram | Is there a relationship between two variables? |
Build Histograms Instantly in SigmaDesk
If you want to build histograms and run a full process capability analysis without manually setting up bins or typing a single formula, SigmaDesk handles all of it automatically. Upload your data, and the platform generates the histogram, overlays your specification limits, and computes capability indices in one step.
Frequently Asked Questions About Histograms
What is a histogram used for in quality control?
A histogram is used to visualize the distribution of process data so you can assess its center, spread, and shape at a glance. In quality control, it helps identify variation patterns, detect mixed data sources, and compare process output against specification limits before running formal capability calculations.
What is the difference between a histogram and a bar chart?
A histogram displays continuous numerical data grouped into bins, with adjacent bars and no gaps, while a bar chart compares separate categories with gaps between bars. In a histogram, the horizontal axis represents a measurement scale rather than distinct categories.
How many bins should a histogram have?
A common rule of thumb is to use approximately the square root of the number of data points, so 100 measurements would use about 10 bins. Too few bins hide real patterns, while too many fragment the data into noise, so adjust the count based on what the shape reveals.
How many data points do you need for a histogram?
You need at least 50 data points for a histogram to show a meaningful distribution shape, and 100 or more is ideal. Histograms built from small samples of 20 to 30 points can display misleading patterns that do not reflect the true process behavior.
What does a bimodal histogram tell you about a process?
A histogram with two distinct peaks usually means the data actually comes from two different sources that have been mixed together, for example two machines, two shifts, two operators, or two suppliers with different average outputs. The fix is to separate the data by source and build a histogram for each, rather than treating the combined data as one process.
What is the difference between a histogram and a box plot?
A histogram shows the full shape of the distribution: where the peak sits, whether the data is skewed, and whether there are two humps rather than one. A box plot compresses that same data down to a median, quartiles, and outliers, which makes it far better for comparing several groups side by side but hides shape entirely — two very different distributions can produce almost identical box plots. For a capability study use the histogram, and for comparing five suppliers at a glance use box plots.
What should you do if your histogram shows a skewed, non-symmetric shape?
First check whether the skew is real, such as a natural boundary like zero or a process that can only go one direction, or caused by mixed sources like two shifts or machines. If it’s a genuine feature of the process, avoid using indices like Cpk that assume normality without adjustment, and consider a transformation or a distribution-based capability method instead.
One thing I’d add to the bin count tip: try two or three bin counts before you conclude anything about shape. The same dataset can look bimodal at 20 bins and perfectly normal at 8. The square root rule is a starting point, not an answer, and it’s worth knowing how fragile the picture is before you show it to anyone.
The 50 to 100 data point guidance is right but hard to live with in some settings. On destructive testing at several hundred pounds a part, nobody is giving you 100 samples for a histogram. Worth saying explicitly what to do instead, because in practice people just build the histogram from 15 points and read the shape anyway, which is worse than not drawing it.