Your shop’s average order is forty dollars, a number you quote with a little pride. Yet almost every order that actually comes in is closer to twenty, with a small handful of big spenders quietly hauling that average up to a place where nobody really lives.
The average is not lying by accident, it is being dragged. The word for exactly why it keeps getting dragged is skewness. I run my own small businesses, and I spent years reading the average as if it meant typical, then getting ambushed by the long tail hiding off to one side. This is not probability theory, it is the plain habit of asking which way your data leans before you trust any average at all.
What skewness actually is
Skewness is a measure of how lopsided your data is, whether it leans to the low side or the high side rather than sitting evenly around its middle. A perfectly balanced shape, where the left half mirrors the right, has no skew at all.
The moment one tail stretches out longer and thinner than the other, that stretch is the skew, and the longer tail points in the direction your data is said to lean.
The common right skew
The kind you will meet most often in a small business is a right skew, sometimes called a positive skew. It happens because so many of your numbers clump hard against a floor of zero and then trail away into a long tail of large values.
Revenue does it, order sizes do it, time on page, days to pay an invoice, customers won from a single referral: they all pile up low and then reach out to the right with a thin tail of whales and stragglers. That long tail on the high side is the signature of a right skew, and it is the native shape of most business data.
The rarer left skew
A left skew, or negative skew, is the mirror image, where the long thin tail stretches toward the low side instead. You tend to see it when your values bunch up near a natural ceiling with only a few exceptions pulling downward.
Think of a product rated mostly four and five stars, a comfortable crowd of happy reviews clustered near the top, with a thin tail of one and two star ratings dragging the shape to the left. Most of the weight sits high, and the lonely low values form the tail that gives the lean its name.
The mean versus median giveaway
Here is the fastest way to catch a skew without drawing anything, and it is worth memorising. Put your average right next to your median, the true middle value with half your data above it and half below.
| Shape | Where the average sits | What it means |
|---|---|---|
| Right skew | Average above the median | A long tail of high values |
| Left skew | Average below the median | A long tail of low values |
| Symmetric | Average and median together | No meaningful lean |
The gap between mean and median, and the direction of that gap, is skewness you can read in a single glance.
Why the mean stops being typical
This matters because on skewed data your average quietly stops being typical. It describes a middle that almost nobody actually occupies, a point pulled toward the tail and away from the crowd.
When your orders are right skewed, the average sits up among the big spenders while most of your real customers live down near the median, spending far less. Quote that average as if it were the normal customer and you will overstock, overpromise, and misjudge who you are actually serving. The median is the honest answer to what a typical value really looks like.
See it with a histogram
If you want to see the skew rather than infer it, draw a histogram, which just sorts your values into buckets and shows how tall each bucket stands. A symmetric data set looks like an even mound with matching slopes on both sides.
Right skewed data looks like a wave that has broken to the left, a tall pile near the low end with a long shallow tail streaming off to the right. One honest picture like that tells you more about the shape of your business than a whole page of averages ever will.
Measure it as a number
There is also a proper number for it, and most spreadsheets and analytics tools will hand it to you. A skewness value near zero means your data is roughly symmetric. A positive number means a tail reaching to the right, and the larger it grows the heavier that high tail is. A negative number means a tail reaching to the left.
As a rough working guide, a value within about half of zero is fairly balanced, and once you drift past one in either direction you are looking at data with a serious lean that no symmetric rule should be trusted on.
If you would rather not hunt for that statistic, there is a simple proxy. Take the average, subtract the median, and notice both the sign and the size of what is left. A clearly positive gap leans right, a clearly negative gap leans left. It is not the exact figure a tool would print, but for the everyday job of deciding whether your average can be trusted, that little subtraction is very often all you need.
A founder sized example
Treat these as illustration, not benchmarks to copy. Say your median order is twenty two dollars but your average order is forty. That gap is not noise, it is a fat right tail of large orders quietly stretching the average to nearly double the value a typical customer actually spends.
If you had priced or planned around the forty, you would have been designing for a customer who barely exists, while the real crowd sitting at twenty two went unserved. The skew was not a rounding error, it was the whole story of who your buyers really are.
What skew quietly breaks
A skew also breaks the tidy rules you may have borrowed from the bell curve. The two thirds rule, where most values are supposed to sit within one standard deviation of the average, assumes a symmetric shape and falls apart on skewed data.
A target set two standard deviations out will flag heaps of perfectly ordinary values as strange. A forecast that treats next month as a neat balanced spread around its average will badly misjudge both the monster month and the dead one. Every one of those tools silently assumed no skew, and your lopsided data did not agree.
Which number to report instead
On skewed data, reach for the median first as your honest picture of the typical case, and pair it with a tail percentile to show where the extremes actually live. The average still has its uses, especially for money, where total revenue divided by number of orders is a real and meaningful figure.
Just stop asking the average to describe the normal customer when the data is clearly leaning, because that is the one job a skewed average cannot do without quietly deceiving you.
Do not scrub the skew away
Resist the urge to treat the skew as a defect to be scrubbed out. Very often the lean is the most important thing your data is telling you, because the tail is exactly where your biggest customers, your worst delays, or your most loyal fans actually sit.
Yes, there are honest ways to reshape a badly skewed number so that symmetric tools fit it better, and they have their place in serious analysis. But for the daily reading of a small business, you are far better served by tools that never demanded symmetry in the first place.
Skew versus a lone outlier
It is worth separating a genuine skew from a couple of freak values, because they can look alike at first. One enormous order landing in an otherwise tidy week can tug your average and mimic a right skew, but that is really a single outlier, not the shape of your data leaning.
A true skew is a persistent lean of the whole distribution, showing up month after month, not one strange point you could name and remove. When you see the pull, ask whether it is the entire shape tilting or just one loud value shouting, because the honest fix for each is completely different.
You need enough data for a lean
A skewness reading needs a decent pile of data underneath it to mean much. On fifteen orders, a single lucky whale can make your data look violently skewed one week and almost balanced the next, and the number will jump around far too much to trust.
The median holds up on thin data better than the skew statistic does, so on small samples read the mean against the median for direction, and treat any precise skewness figure as a rough hint until you have real volume feeding it.
The honest limit
Hold the whole idea in proportion. Skewness tells you which way your data leans and roughly how hard, and that is genuinely useful, but it stops there. It will not tell you why the tail exists, it will not explain the big spenders or the slow payers, and it will not straighten a lopsided business by being measured.
All it really buys you is the judgement to stop trusting an average the moment your data tilts. Every metric and method like this one is explained in plain English at dataresearchanalysiscollection.com, so you can read it again slowly with your own numbers open. No hype, no promises about your results, just the idea explained until you can tell a genuine lean from the balanced data you were quietly assuming.
Get new guides and videos first — join the Telegram channel.