Z-Score Analysis
Convert raw values into standard deviation units so they can be compared on a common scale. Use z-scores to find probabilities and percentiles from the normal distribution, screen for outliers, and express performance in sigma terms.
Run Z-Score Analysis →What is a Z-Score?
A z-score, or standard score, expresses how many standard deviations a value lies from the mean of its distribution. A z-score of +2 means the value is two standard deviations above the mean; a z-score of −1.5 means it is one and a half standard deviations below.
Standardizing to z-scores puts values from different distributions onto a single common scale, making them directly comparable. It also connects any normal value to the standard normal distribution, so a z-score can be translated into a probability (the area under the curve) or a percentile rank.
In quality work, z-scores link a measurement to the probability of falling beyond a specification limit, which is the foundation of sigma-level and defect-rate calculations. They are also a simple, transparent way to flag outliers.
In plain terms: A z-score answers 'how unusual is this value?' by counting standard deviations from the average. Because everything is on the same standardized ruler, you can compare a test score to a height to a cycle time, and you can look up exactly what fraction of values sit beyond any point.
What Z-Scores Give You
Probability & Area
A z-score maps to the area under the standard normal curve, giving the probability of a value below, above, or between points.
Percentile Rank
Converting a z-score to its cumulative probability gives the percentile: a z of 0 is the 50th percentile, +1.28 is roughly the 90th.
Outlier Screening
Values with |z| beyond about 3 are far in the tails and are common candidates for outlier review in approximately normal data.
Key Formulas
Interpreting Z-Scores
The sign shows direction (above or below the mean) and the magnitude shows distance in standard deviations. Because about 68, 95 and 99.7 percent of a normal distribution lie within one, two and three standard deviations, a z-score immediately conveys how typical or extreme a value is.
Z-scores only translate cleanly into probabilities and percentiles when the data is approximately normal. For strongly skewed data, the same z-score corresponds to a different tail probability, so verify normality before reading probabilities from z.
Assumptions & Validation
Known or Estimated Parameters
The mean and standard deviation are known or well estimated from adequate data.
If violated: With small samples and an estimated standard deviation, use the t-distribution instead of z.
Approximate Normality
Translating z-scores to probabilities and percentiles assumes a normal distribution.
If violated: Transform the data or use distribution-free percentile methods.
Meaningful Scale
The variable is on an interval or ratio scale where distances are meaningful.
If violated: Do not standardize ordinal or nominal data.
⚠️ Check assumptions first
A z-score is always computable, but its translation into a probability or percentile is only valid when the data is approximately normal. Applying normal-based tail probabilities to skewed data will misstate how extreme a value really is. Check normality first, and for small samples with an unknown population standard deviation, use the t-distribution rather than the standard normal.
When NOT to Use Z-Score Analysis
Small Samples, Unknown Sigma
When the population standard deviation is unknown and the sample is small, use the t-distribution, which has heavier tails than the normal.
Strongly Non-normal Data
For skewed distributions, z-based probabilities are wrong. Use empirical percentiles or a fitted distribution.
Comparing Group Means
To test whether means differ, use a z-test or t-test rather than raw z-scores of individual values.
Industry Applications
Sigma & Defect Rates
Convert the distance from the mean to a specification limit into a z-score, then into an expected defect probability.
Outlier Detection
Flag measurements with large absolute z-scores for investigation in approximately normal process data.
Cross-Metric Comparison
Standardize different metrics so performance can be compared or combined on a common scale.
Percentile Reporting
Express where an individual value falls relative to the whole distribution as a percentile.
Frequently Asked Questions
What does a z-score tell me?
A z-score tells you how many standard deviations a value lies from the mean. The sign gives the direction, above or below the mean, and the magnitude gives the distance. Because it is standardized, a z-score lets you compare values from completely different distributions and, for normal data, look up the exact probability of being beyond that point.
How do I convert a z-score to a probability or percentile?
For approximately normal data, the cumulative standard normal function converts a z-score into the probability of a value at or below it, which is also its percentile when multiplied by 100. For example a z-score of 0 corresponds to the 50th percentile and a z of about 1.28 to the 90th. Tail probabilities beyond a spec limit come from the complementary area.
When should I use a t-score instead of a z-score?
Use the t-distribution when the population standard deviation is unknown and estimated from a small sample. The t-distribution has heavier tails that account for the extra uncertainty in estimating the standard deviation, and it approaches the standard normal as the sample size grows, typically becoming very close beyond about 30 observations.
Can z-scores detect outliers?
Yes, for approximately normal data. Values with an absolute z-score beyond about 3 lie far in the tails and are common candidates for outlier review, since only about 0.3 percent of normal values fall there. For skewed data or small samples, however, z-based thresholds can be misleading, and robust methods such as the modified z-score or the IQR rule are preferable.
Do z-scores require a normal distribution?
Calculating a z-score does not require normality, it is just a rescaling. However, interpreting a z-score as a probability or percentile does assume approximate normality, because those readings come from the standard normal curve. For non-normal data the same z-score corresponds to a different tail probability, so normality should be checked before drawing probability conclusions.
How do z-scores relate to sigma level?
Sigma level is essentially a z-score measured from the process mean to the specification limit. A characteristic whose nearer spec limit sits three standard deviations away has a z of 3 on that side. Converting that z to a tail probability gives the expected fraction of nonconforming output, which is the basis of DPMO and sigma-level reporting.
Put Every Value on a Common Scale
Standardize values, find probabilities and percentiles, and screen outliers. Free during Beta.
Run Z-Score Analysis →