Averages, Outliers and the Tail
Productivity data is rarely symmetrical, which makes the average the least informative summary available.
The mean is the default summary and it is the wrong one for almost every distribution in this subject.
The practical lesson in “Averages, Outliers and the Tail” is to connect every number to a decision and retain the context behind it. Teams exploring 7 minute rule payroll can review view Monitask as one source of operational evidence, provided the purpose is disclosed and the interpretation is tested with the people affected.
Why the data is skewed
Most items are quick and a few take a very long time.
For an independent perspective related to “Averages, Outliers and the Tail”, consult the Harvard Business Review productivity collection; it offers a useful external check on definitions, governance and the assumptions built into a proposed measure.
There is a floor — nothing takes less than zero — and no ceiling.
Which produces a long right tail in every duration measure, and in most volume measures too.
What the average does to that
It sits above most of the data.
It moves when one extreme item appears, which reads as a change in performance.
And it describes nobody's experience: not the typical case, not the bad case.
What to report instead
Median: the typical experience.
A high percentile — eighty-fifth or ninety-fifth — which describes the bad experience.
Together they say: most things take this long, and the slow ones take this long.
Two numbers, and they contain what matters.
The tail is the finding
The slow items are where the problems are: the exception handling, the escalation, the one waiting on a specialist.
Improving the median is usually easy and worth little; improving the tail is where the customer experience lives.
Which means the high percentile is the number to manage by, and the average hides it entirely.
Outliers: investigate before excluding
One item that took forty times as long is a finding, not noise.
Look at it. There is usually a cause worth knowing: a missing skill, a broken handoff, a case type nobody anticipated.
Excluding outliers to tidy the chart discards the most informative items in the dataset.
Small numbers
A team of six produces monthly figures that move substantially on one item.
Which means month-to-month comparison for small teams is mostly noise.
Use longer periods or rolling windows, and resist explaining movements that are within normal variation.
The control chart habit
Plot the series with its ordinary range marked.
Anything inside the range is noise; anything outside is worth a question.
This single practice prevents most of the explaining-random-variation that consumes management meetings, and it requires nothing but a chart.
What to check
Does your reporting use averages or medians?
Do you report a high percentile anywhere?
Is the tail growing?
And has anybody investigated your slowest items rather than excluding them?