Z-Score Calculator · Practical guide
What Z Score Is Considered an Outlier?
Beyond ±2 standard deviations is unusual; beyond ±3 is the classic outlier cutoff used in statistics.
At a glance
The Common Cutoffs: ±2 and ±3
There is no single official line, but two thresholds dominate practice. A z-score beyond ±2 is "unusual" — in a normal distribution only about 4.6% of values land out there. A z-score beyond ±3 is the classic outlier definition — just 0.27% of values, roughly 1 in 370. This calculator flags anything beyond ±2 as "Far above/below average (outlier)."
Interactive tool
Z-Score Calculator
Enter your own values, then compare the result with the examples below.
Results
How Rare Each Threshold Really Is
| |Z| beyond | Percentile range | Frequency (normal data) |
|---|---|---|
| 1.0 | Below 15.9% or above 84.1% | 1 in 3.1 |
| 2.0 | Below 2.3% or above 97.7% | 1 in 22 |
| 2.5 | Below 0.6% or above 99.4% | 1 in 81 |
| 3.0 | Below 0.13% or above 99.87% | 1 in 370 |
A Worked Example
On an exam with mean 75 and standard deviation 8, a score of 95 gives z = (95 − 75) / 8 = 2.50 — far above average, at the 99.4th percentile. A score of 58 gives z = −2.125 — flagged far below average. Both cross the ±2 line; neither reaches ±3. In quality control and research, the ±3 rule is the usual trigger for investigating a data point as a genuine anomaly or error.
Choosing a Threshold for Your Situation
Context matters. Exam grading often treats ±2 as noteworthy; manufacturing quality control and outlier-detection rules typically use ±3. Also remember the thresholds assume roughly normal data — in skewed distributions like income, extreme z-scores are far more common and do not necessarily signal errors. When in doubt, look at the data point itself before discarding it: outliers are sometimes the most interesting values in the set.
Frequently Asked Questions
Is a z score of 2 an outlier?
It is the boundary. Many textbooks call |z| > 2 unusual, and this calculator labels anything beyond ±2 as an outlier. The stricter ±3 rule is standard for formal outlier detection.
What z score do scientists use to flag outliers?
±3 is the most common convention, corresponding to the outer 0.27% of a normal distribution. Some fields use 2.5 or rely on the interquartile range instead.
Should outliers be removed from a data set?
Not automatically. First check for data-entry errors. If the value is genuine, removing it can distort results — report it or use robust statistics like the median instead.
Can a z score be bigger than 3 without being an error?
Yes. In large or skewed datasets, z-scores beyond 3 occur naturally. A perfect SAT score or a billionaire's income is a real outlier, not a mistake.