Stop Sending Your CEO the Average. Send the Shape Instead
Why a single mean hides the story in most business data, and three simple alternatives that lead to better decisions.
A few months ago a founder asked me why their average order value had jumped 18 percent in a quarter when nothing about the product had changed. Everyone in the room had a theory: a new pricing page, a better ad campaign, a seasonal bump. The real answer took me four minutes to find. Two corporate customers had placed bulk orders worth more than the previous month of regular sales combined.
The average had moved. The business had not.
Why averages mislead
The arithmetic mean is a fine summary when data is roughly symmetric and free of extreme values. Most business data is neither. Order sizes, session lengths, time to close a deal, salaries, support ticket durations: these are typically skewed, with a long tail of large values dragging the mean upward.
In that founder's data, the median order was €38 and barely moved all year. The mean swung between €58 and €70 depending on whether a big account happened to buy that month. Reporting the mean meant reporting the purchasing schedule of two procurement teams.
Three things I report instead
1. The median, with a range around it
The median tells you what a typical customer does. I pair it with the 25th and 75th percentiles so the reader sees spread without needing a statistics lecture. A line like typical order €38, middle half between €25 and €58 is something a CEO can act on.
2. A split by segment
Often the right move is to stop pretending the population is one group. When I separated retail and corporate orders, each segment told a clean story. Retail was flat. Corporate was lumpy but growing. Those are different problems with different owners.
3. A simple histogram
I resisted charts in executive updates for a long time because they felt like decoration. Then I watched a leadership team stare at a histogram of delivery times and immediately spot a second bump around day nine. It was one courier partner. Nobody would have seen it in a single number.
When the mean is still right
I am not anti-mean. If you need totals, such as forecasting revenue or estimating total inventory, the mean is exactly what you want, because the mean multiplied by the count gives the total. The mistake is using it to describe what a typical case looks like.
A habit worth building
Before any metric goes into a deck, I ask one question: if two unusual events happened this month, would this number lie to us? If the answer is yes, I send the shape rather than the summary.
Most decisions do not need more data. They need the same data described honestly.
Cover: Photo by Nick Brunner on Unsplash.
