Average to Distribution

When “it usually works” clashes with someone’s actual experience, look behind the average. Show individual cases, the spread, and any meaningful differences between groups.

Keep the average → show the spread → inspect who or what differs.

An average can be accurate while hiding distinctions the decision depends on. Keep it, then add a view of the underlying cases: a dot plot, ranges, useful subgroups, or a timeline. Choose divisions that fit the question, and avoid treating a handful of cases as a stable pattern.

Try this when…

The average looks fine, but people are still complaining.

Some days this takes ten minutes and some days it takes an hour.

That typical customer does not sound like anyone I know.

We all got the same result, but for very different reasons.

What you can use it for

Service experience

See whether a reasonable typical wait hides a group with very long delays.

Planning capacity

Account for variability when a typical duration is not enough to protect a commitment.

Learning and performance

Distinguish broad results from particular question types or situations needing attention.