Measurement and Improvement
Histogram
The average may be telling the truth while hiding nearly everything useful.
What is it?
A Histogram groups numerical observations into ranges and shows how often values occur in each range. It reveals the shape, spread and concentration of the data.
What's it really for?
It helps a team notice when one process is really several different processes sharing the same average. Two peaks may point to different shifts, products or methods. A long tail may reveal the awkward cases everyone remembers but the monthly figure absorbs.
You know when…
The report says the average response time is four days. Customers describe experiences ranging from an afternoon to three weeks. Both statements are correct, which is not especially comforting. Plotting the distribution shows where those very different experiences are hiding.
Have a go
- Gather numerical data measured in the same way. ☐ Choose sensible, equal-width intervals. ☐ Count observations in each interval. ☐ Draw touching bars and label both axes. ☐ Examine shape, centre, spread, tails and possible groups.
Why it works
Averages compress a whole set of observations into one number. A Histogram puts the distribution back on view, making skew, multiple peaks, gaps, outliers and excessive spread easier to see.
Look for…
A single balanced mound, skewed tails, more than one peak, gaps, isolated values and a spread wider than the requirement allows. Compare apparent groups with what was happening in the work.
Watch out…
A Histogram loses time order. Two very different sequences can create the same shape. Avoid tiny samples, arbitrary bin widths and mixing unlike data simply to produce a fuller-looking chart.
Worth noticing…
A second peak is often a clue that the data has been waiting for someone to ask a better question about where it came from.
Try next…
Use a Run Chart to restore time order. Use a Pareto Chart for categories rather than continuous numbers. Use a Scatter Diagram to explore whether two measures move together.