In this article
An average is a compact answer to one question. It is not a miniature version of the whole process. Two sets of completion times can share the same average while describing very different experiences for the items moving through them.
Same mean, different spread
Fictional example: Five generic poster requests complete in 9, 10, 10, 10, and 11 minutes. Another five complete in 2, 4, 10, 14, and 20 minutes. Both groups total fifty minutes, so both have a mean of ten minutes.
The first group is tightly clustered; the second includes both much shorter and much longer journeys. If an imaginary requester needs a poster within fifteen minutes, the average alone cannot reveal that one request in the second group misses that mark. The range offers another view: two minutes in the first group, eighteen in the second.
That does not make the range a sufficient replacement for the mean. With only five invented observations, one value heavily influences it. The point is to ask what information the summary hides, not to crown one statistic as the universal answer.
IHI’s Histogram resource explains that displaying a distribution can reveal patterns missed by an average or median. A distribution groups values by size. It answers a different question from a chart that preserves when each value occurred.
The same values can tell different stories in time
Consider these fictional sequences: 14, 13, 12, 11, 10, 9, 8, 7; and 7, 14, 8, 13, 9, 12, 10, 11. They contain the same eight values. They therefore have the same mean and the same spread. Yet one descends steadily, while the other alternates around the middle.
Sorting the numbers erases that difference. A histogram would not retain it either. Keeping time order allows you to ask whether an apparent change emerged gradually, followed a specific event, or consisted of alternating conditions. It does not by itself establish the cause.
A change in mix can move the total
Now imagine two fictional kinds of poster request: a short format taking four minutes and a long format taking twelve. A day containing mostly short formats will have a lower overall average than a day containing mostly long formats, even if neither format becomes faster.
Before praising or criticizing a daily average, ask whether the included work changed. This is a question about comparability, not permission to collect personal characteristics or sort employees into rankings. An authorized analyst may need approved non-sensitive process categories to interpret a result. A reader practicing at home needs only the invented short and long requests.
Do not attach a story to every movement
A value rising from ten to eleven does not automatically mean yesterday’s change made things worse. A value falling to nine does not prove success. NIST’s engineering statistics glossary distinguishes variation associated with the continuing process from variation associated with particular conditions. Determining whether a pattern warrants a specific explanation requires an appropriate method and context.
In practical reading, that means resisting two temptations: explaining every ordinary fluctuation as someone’s action, and dismissing every unusual result as “just variation.” A concern may still need reporting even when the available chart cannot establish a statistical signal.
Read a result in layers
- Confirm the unit and the definition stayed consistent
- Look at the center and at how widely values differ
- Preserve time order rather than relying only on a sorted summary
- Check whether the mix of included work changed
- Distinguish an observed pattern from a proposed explanation
For the five-request examples, write two truthful sentences: one about the shared mean and one about the different spread. Then explain why neither sentence forecasts a real department’s next day. This small exercise builds a more useful habit than treating every headline number as a verdict.
Scope: The numeric sets and reading sequence are original editorial examples. IHI public page text and the NIST glossary’s definitions were checked October 8, 2026. We did not assess Northwell performance or apply statistical decision rules to any workforce data. Continue with reading a run chart to keep time in the picture.