A histogram is skewed when its bars trail off unevenly to one side, creating a longer tail in one direction than the other. The quickest way to spot it: look at which side has the tail, not where the bulk of the data sits. If the tail stretches to the right, the histogram is right-skewed (positively skewed). If it stretches to the left, it’s left-skewed (negatively skewed). A symmetric histogram has roughly equal tails on both sides.
The Tail Tells You the Direction
The single most reliable visual rule is to find the tail. The tail is the thinner, stretched-out side of the histogram where bars get shorter and shorter as they trail away from the peak. Whichever direction the tail points is the direction of the skew.
This trips people up because it feels counterintuitive. In a right-skewed histogram, most of the data is bunched on the left side, near the lower values. The peak is on the left. But the tail extends to the right, toward higher values, so it’s called right-skewed. Your eye naturally goes to where the tall bars are, but the skew is named for the opposite end, where the data thins out.
Left-skewed histograms work the same way in reverse. The bulk of the data clusters on the right (higher values), and the tail stretches to the left toward lower values.
Right-Skewed Histograms in Real Life
Right skew is extremely common in real-world data. Income distribution is the classic example: most people earn in a moderate range, but a small number of very high earners pull the tail far to the right. Other familiar examples include the number of pets per household (most people have zero or one, but a few have seven or more), movie ticket sales (most films sell few tickets while a handful of blockbusters sell millions), and scores on a very difficult exam (most students cluster around a low-to-middle score, with a few high performers stretching the tail).
The pattern to notice: right skew shows up whenever there’s a natural floor but no real ceiling. Income can’t go below zero, but there’s no hard upper limit, so the data spreads out to the right.
Left-Skewed Histograms in Real Life
Left skew appears when most values cluster near the high end with a few trailing off toward lower values. Retirement age is a good example: the majority of people retire in their mid-60s or later, but some retire much earlier due to disability, early savings, or other circumstances, pulling the tail to the left. Scores on an easy exam also tend to be left-skewed, with most students scoring high and only a few earning very low marks.
Using Mean, Median, and Mode
If you can’t tell by looking, the relationship between the mean, median, and mode gives you a second check. In a right-skewed distribution, the mean is typically the largest of the three, the mode is the smallest, and the median falls in between. The order is: mode, then median, then mean, reading from smallest to largest.
In a left-skewed distribution, the order reverses. The mean is the smallest, the mode is the largest, and the median sits between them.
This happens because the mean gets pulled toward extreme values. A few very high data points drag the mean to the right; a few very low data points drag it to the left. The median resists that pull because it only cares about position, not magnitude. So when you notice the mean is noticeably higher than the median, that’s a sign of right skew. When the mean is noticeably lower than the median, that suggests left skew. These relationships hold in most cases, though occasional exceptions exist in unusual distributions.
Checking Skewness With a Number
Visual inspection works well for obvious cases, but borderline histograms need a more precise tool. The skewness coefficient gives you a single number that quantifies how lopsided your data is. A value of zero means perfectly symmetric. Positive values mean right skew, and negative values mean left skew.
How far from zero counts as “actually skewed” depends on your sample size. With a small sample (around 25 data points), random variation alone can produce a skewness coefficient anywhere from about -0.73 to +0.73, even if the underlying data is perfectly symmetric. With 100 data points, that expected range narrows to roughly -0.39 to +0.39. With 500 data points, anything outside -0.18 to +0.18 is likely reflecting real skewness rather than random noise. So larger datasets demand less skewness to be considered meaningfully skewed.
A practical rule of thumb that many introductory courses use: values between -0.5 and +0.5 suggest roughly symmetric data, values between -1 and -0.5 or +0.5 and +1 indicate moderate skew, and values beyond -1 or +1 point to heavy skew.
How to Calculate It in Software
You don’t need to compute skewness by hand. In Excel, type =SKEW(A1:A100) where the range covers your data cells. This treats your data as a sample. If your data represents an entire population rather than a sample, use =SKEW.P() instead. In Python, the scipy.stats.skew() function does the same thing. In R, use skewness() from the moments package.
All of these return a single number you can interpret using the guidelines above: check the sign for direction, then check the magnitude against expected ranges for your sample size.
Skew vs. Outliers
One common mistake is confusing a skewed distribution with a distribution that simply has a few outliers. If your histogram looks symmetric except for one or two bars far off to one side, that’s not necessarily skew. True skewness means the data gradually tapers in one direction, creating a smooth tail. A couple of isolated extreme values sitting far from an otherwise normal-looking shape may just be outliers, and removing or investigating those values could reveal a symmetric distribution underneath.
Both situations pull the mean away from the median, so you can’t rely on that relationship alone to distinguish them. Looking at the histogram’s overall shape is the most direct way to tell the difference. A skewed distribution shows a gradual, continuous tapering. Outliers show a gap between the main body of data and a few distant values.