Right-Skewed Histogram: How to Read and Report One

What a right-skewed histogram looks like

A right-skewed histogram piles up on the left and thins out toward the right. Most of the data sits at the low end, and a small number of larger values pull a long tail out to the right side of the chart.

The example above is 60 customer wait times in minutes. Most people wait a couple of minutes, a handful wait much longer, and those long waits form the tail. The name follows the tail, not the peak, so a shape that peaks on the left is still called right-skewed.

Right skew, positive skew, and skewed to the right are three names for this one shape, and left skew is its mirror image. If the naming is what brought you here, the right-skewed vs left-skewed guide covers every variant of both side by side, along with the formula for putting a number on skew.

This page is about the next question: you have a right-skewed histogram in front of you, so what do you actually do with it?

Reading this histogram, number by number

Here is the full summary of the wait times above, and it is worth walking through because every one of these numbers says something different.

Measure	Value
Count	60 waits
Range	1 to 10 minutes
Mode	2 minutes
Median	4 minutes
Mean	4.12 minutes
Standard deviation	2.22 minutes
Q1 to Q3	2 to 5.25 minutes
Interquartile range	3.25 minutes
Skewness	0.74

Read that as a story rather than a table. The single most common wait is 2 minutes. Half of all customers are served in 4 minutes or less, and 48 percent are done inside 3 minutes. But 17 percent wait 7 minutes or longer, and one unlucky person waited 10.

The skewness of 0.74 is the formal confirmation. Positive means right-skewed, and anything past about 0.5 counts as a real lean rather than noise.

That is the honest summary of this counter: fast for most people, occasionally slow, and the occasional slowness is not a fluke in the data. It is a feature of how the queue behaves.

When the mean-median rule goes quiet

Most guides, including parts of this one, tell you that right skew means the mean prints higher than the median. That is true here. The mean is 4.12 and the median is 4.

But look at the size of that gap: 0.12 minutes. Barely six seconds. If you were checking for skew by eyeballing the mean against the median, you would almost certainly conclude this data was symmetric, and you would be wrong. The skewness statistic says 0.74, which is a clear right lean.

Two things cause that mismatch, and both are common.

The first is that the data is discrete. Wait times here are whole minutes, so the median gets pinned to a whole number and cannot drift the way the mean can. The second is that Pearson's second skewness coefficient, the mean-median shortcut, is a crude approximation. Run it on this data and you get 0.16, which reads as almost symmetric, against a true skewness of 0.74.

So treat the mean-median rule as a hint, not a verdict. When they are far apart, you certainly have skew. When they are close, you might still have it. The reliable checks are the shape of the tail and the skewness statistic itself.

Mean, median, and mode

Skew changes the order of the three averages. The long right tail drags the mean upward, so:

mode < median < mean

This data follows that exactly: 2, then 4, then 4.12. The mode sits under the tall bars on the left, the median is in the middle, and the mean is pulled right by the long waits.

Which average should you report

Once a histogram leans right, the mean stops describing a typical experience because the tail keeps tugging it upward. Here are four summaries of the same 60 waits.

Summary	Value	What it answers
Mean	4.12 min	The total wait divided evenly across everyone
Median	4 min	What the middle customer actually experienced
Trimmed mean (10 percent)	3.90 min	The mean after dropping the fastest and slowest tenth
Geometric mean	3.53 min	The typical value on a multiplicative scale

The median is the right default. It ignores how extreme the extremes are and stays with the bulk of the data, which is why economists report median income and why hospitals report median length of stay.

The gap between these four is small on this dataset precisely because the skew here is moderate. On strongly skewed data, incomes being the standard example, the mean can land above the 70th percentile, and reporting it as the typical value is close to misleading.

Skew is not the same as outliers

This is the distinction people most often get wrong, and this histogram is a clean demonstration of it.

Run the standard outlier test on this data. Q3 plus 1.5 times the interquartile range gives a fence at 10.12 minutes. The longest wait is 10 minutes. Nothing clears the fence, so by the textbook rule this dataset contains zero outliers.

And yet the skewness is 0.74. The shape is genuinely right-leaning.

Skew is a property of the whole distribution: the data thins out gradually on one side. An outlier is a single point sitting far from everything else. You can have skew with no outliers, as here. You can also have a symmetric dataset with one wild outlier in it. Removing points will not fix skew, because skew is not caused by a few bad values, and deleting real data to force a symmetric shape is how honest analysis goes wrong.

What to do with right-skewed data

Four practical moves, roughly in order of how often they are the right call.

• Report the median with the interquartile range, not the mean with the standard deviation. Here that is a median of 4 minutes with an IQR of 2 to 5.25. A standard deviation of 2.22 implies a symmetric spread in both directions, which is not what this data does.
• Quote a high percentile when the tail is what matters. For a service counter, "17 percent of customers wait 7 minutes or more" is more useful to a manager than any average, because the complaints come from the tail.
• Consider a log transform if you need to run a test that assumes symmetry. Logging pulls the long right tail in and often produces a roughly symmetric shape. The geometric mean of 3.53 minutes is what you get back when you average in log space.
• Do not delete the tail to make the shape behave. Those long waits are real customers.

Where you see it

Income, house prices, wait times, response times, file sizes, insurance claims, and hospital length of stay are almost always right-skewed. They share one trait: a hard floor at zero and no ceiling. Values cannot drop below zero, but a few can run far above the crowd, so the data can only stretch in one direction. That one-sided freedom is what builds the right tail.

Income is the textbook case. Most earners cluster in a normal band while a small number of very high earners stretch the top end, so average income lands above what a typical person makes. That is exactly why median income is the figure that gets reported.

Right-skewed vs the mirror image

Flip this shape and you get a left-skewed histogram, where the tail points left and the mean falls below the median instead. Everything on this page reverses: report the median still, but quote a low percentile rather than a high one, since the tail that matters is the short end.

For the two shapes laid out side by side, plus the naming variants and the skewness formula worked through step by step, see the right-skewed vs left-skewed guide.

Try it on your own data

Paste your numbers into the histogram maker and check the tail direction and the skewness figure together rather than relying on the mean-median gap. The frequency distribution table maker shows the raw count in each bin, which makes a thinning tail easy to see in the numbers themselves.