How to Use Z Scores to Compare Different Data Sets

What Is a Z-Score? (The Basics)

A z-score, also called a standard score, describes exactly where a single data point falls compared to the rest of its data set. Instead of saying "this value is 82," a z-score says "this value is 1 standard deviation above average" — which turns out to be a far more useful way to think about numbers.

Standard deviation measures how spread out a group of values is. A small standard deviation means most values sit close to the average. A large one means values are scattered more widely. Once you know both the mean and the standard deviation of a data set, you have everything you need to turn any raw value into a z-score.

If you want the fuller picture of what a z-score represents on its own, our guide on what a z-score means and how to calculate it correctly covers the concept from the ground up. This article builds on that foundation and focuses specifically on using z-scores to compare two different data sets.

The Z-Score Formula

The formula for a population z-score is:

For a sample instead of a full population, the formula uses slightly different symbols but works exactly the same way:

What Each Part of the Formula Means

Here's what each symbol in the formula stands for:

Why Z-Scores Let You Compare Completely Different Data Sets

This is the real reason z-scores are worth learning. Once you understand this part, the rest of the guide is just mechanics.

The Problem With Comparing Raw Numbers

Raw numbers alone don't tell you much on their own. A score of 88 sounds better than a score of 82, but that assumes both numbers come from the same scale, with the same average and the same spread. Test scores rarely work that way.

The problem gets even more obvious once the units themselves are different. There's no meaningful way to compare a height of 170 centimeters to a weight of 65 kilograms — the two numbers aren't measuring the same thing, so lining them up side by side tells you nothing.

Z-scores solve both problems at once. They strip away the original units and scale, and replace them with a single, shared measurement: how far a value sits from its own group's average, in standard deviation units.

How Standardizing Solves It

Subtracting the mean centers every data set at 0. Dividing by the standard deviation rescales it so that "1 unit" always means "1 standard deviation," no matter what the original units were.

Once two data sets go through this process, they land on the exact same scale: a mean of 0 and a standard deviation of 1. Two completely different test scores become directly comparable — not because the underlying numbers are similar, but because both have been converted to answer the exact same question: how unusual is this value, relative to its own group?

How to Calculate a Z-Score, Step by Step

Calculating a z-score by hand takes four steps.

1. Find the mean of the data set. Add up every value and divide by how many values there are.
2. Find the standard deviation of the data set. This measures how spread out the values are around the mean. Our guide on mean absolute deviation vs standard deviation is a good refresher if you need to calculate this piece from scratch.
3. Subtract the mean from your value. This tells you the raw distance between your value and the average.
4. Divide that result by the standard deviation. This converts the raw distance into standard deviation units — your z-score.

Population vs Sample Z-Scores

Use the population formula (μ and σ) only when you have data on every member of the group you care about — every student in one specific class, for example. In most real-world situations, you're working with a sample instead — a smaller subset of a larger group, like a survey of 200 people out of a city of 200,000. In that case, use the sample formula (x̄ and s).

The good news: the math works the same way either way. The only thing that changes is where your mean and standard deviation come from.

Worked Example: Comparing Two Different Data Sets With Z-Scores

Here's the exact situation this guide opened with. A student takes two different tests and wants to know which one they actually did better on, relative to their classmates.

Step 1: Find the Mean and Standard Deviation for Each Data Set

Both tests are scored out of 100, but they clearly weren't graded the same way — the science class average is higher, and the scores in that class are more tightly clustered around the mean, since its standard deviation is smaller.

Step 2: Calculate Each Z-Score

Working through the formula for each test:

Step 3: Compare the Z-Scores, Not the Raw Scores

Looking only at the raw numbers, 88 beats 82. But that comparison was never quite valid in the first place — these are two different tests, graded on two different curves.

Once both scores are converted to z-scores, a clearer picture appears. The math z-score is 1.0. The science z-score is 2.0. Even though the raw science score is higher, the size of the gap above average is what matters here — and the science result is twice as many standard deviations above the mean as the math result. Relative to classmates, this student's science performance was more unusual, meaning stronger, than their math performance, even though the raw numbers might suggest something different at a glance.

How to Read and Interpret a Z-Score

Positive vs Negative Z-Scores

A positive z-score means the value sits above the mean. A negative z-score means it sits below the mean. A z-score of exactly 0 means the value is the mean.

The size of the number matters more than most people expect. A z-score of 0.2 is barely different from average. A z-score of 3 is dramatically different, rare enough that it stands out from almost everything else in the data set.

What Counts as a High or Low Z-Score

As a general guide:

The Empirical Rule: Understanding 68-95-99.7

For data that follows a roughly normal, bell-shaped distribution, statisticians rely on a simple pattern called the empirical rule, sometimes called the 68-95-99.7 rule:

• About 68% of values fall within 1 standard deviation of the mean (z between -1 and 1)
• About 95% of values fall within 2 standard deviations of the mean (z between -2 and 2)
• About 99.7% of values fall within 3 standard deviations of the mean (z between -3 and 3)

This is exactly why a z-score beyond ±2 tends to raise eyebrows, and a z-score beyond ±3 is often flagged as a likely outlier — fewer than half a percent of values in a normal distribution land out that far.

Z-Scores vs Percentiles

Z-scores and percentiles both describe where a value falls relative to a group, but they answer slightly different questions. A z-score measures distance from the mean, in standard deviations. A percentile tells you what share of the data falls below a given value.

For data that's roughly normally distributed, the two are directly connected:

A z-table, or standard normal table, is what statisticians traditionally used to look up these conversions before calculators made it instant.

Common Mistakes When Using Z-Scores to Compare Data

Even with the right formula, a few habits can quietly throw off a comparison. Here's what to watch for:

• Comparing raw scores instead of z-scores. If two data sets use different scales, the raw numbers alone can't tell you which value is relatively stronger.
• Mixing population and sample formulas inconsistently. Use the same approach, population or sample, for both data sets you're comparing, or the comparison won't be apples to apples.
• Assuming a positive z-score always means "better." A high z-score means "further from average" — whether that's good or bad depends entirely on what's being measured. A negative z-score on a race time might mean a faster finish.
• Treating percentile conversions as exact for skewed data. The z-score to percentile relationship assumes a roughly normal distribution, and heavily skewed data breaks that assumption.
• Forgetting that small samples produce shaky results. A single unusual value in a small data set can swing both the mean and the standard deviation, which throws off every z-score calculated from it.

Real-World Uses for Z-Scores

Z-scores aren't just a classroom exercise. They show up anywhere someone needs to judge a value against a typical range, rather than looking at it in isolation.

Standardized Testing

Z-scores are part of why standardized tests can be compared across different versions, question sets, and even different tests entirely. Once individual scores are converted to a shared standardized scale, admissions offices and researchers can compare an applicant's relative performance, not just their raw number.

Growth Charts in Healthcare

Pediatric growth charts often use z-scores to compare a child's height, weight, or head circumference to other children of the same age and sex. A z-score close to 0 suggests a measurement close to typical for that age group, while a z-score far from 0 in either direction is something a doctor may want to look into further.

Finance and Investment Analysis

In finance, z-scores help compare how unusual a return or price movement is relative to an asset's own typical behavior. A single-day move that looks large in raw percentage terms might be completely ordinary for a volatile asset, while a small move might be unusually large for a normally stable one. A z-score puts both in context.

Quality Control and Manufacturing

Manufacturers use z-scores to flag products that fall outside expected specifications. A part with a measurement several standard deviations from the target value is a strong candidate for closer inspection, even if the raw number alone wouldn't obviously stand out.

Data Science and Machine Learning

Before training many machine learning models, data scientists standardize their features into z-scores. This keeps a variable measured in the thousands, like income, from dominating a variable measured in single digits, like age, purely because of its scale.

Hypothesis Testing and Statistical Significance

Z-scores are also a building block of hypothesis testing, where they help determine whether a result is likely due to chance or reflects a real effect. This connects closely to p-values — our guides on what a p-value really tells you about significance and how to interpret a p-value in hypothesis testing dig deeper into that relationship, and our P Value Calculator can help once you're ready to apply it.

Z-Score vs Standard Deviation vs Raw Score: What's the Difference?

These three numbers work together, but they answer different questions. The raw score tells you what happened. The standard deviation tells you how much variation is normal for that group. The z-score combines both to tell you how that particular result compares to everyone else's.

When Z-Scores Don't Work Well

Z-scores are a genuinely useful tool, but they come with a few real limitations worth knowing about.

• They assume a roughly normal distribution. The empirical rule and percentile estimates lean on that assumption. For heavily skewed data, those specific interpretations become less reliable, even though the z-score itself is still mathematically valid.
• They're sensitive to outliers. Since both the mean and the standard deviation come from the same data, one extreme value can distort the whole scale, which in turn skews every z-score calculated from it.
• They need a reasonably sized data set. Very small samples produce mean and standard deviation estimates that can shift a lot from one data point to the next. Our guide on how to calculate mean absolute deviation step by step walks through a related way of measuring spread that runs into the same sample-size issue.
• They don't explain why a value is different. A z-score tells you how unusual a result is, not what caused it. That part still requires context and judgment.
• Comparability still has to make sense. Standardizing a height and a test score both produce valid z-scores, but deciding what a side-by-side comparison of the two would even mean is a separate question the math can't answer for you.

Quick Reference: Using Z-Scores With Confidence

A short recap before you put this into practice:

• Find the mean and standard deviation of each data set you want to compare
• Subtract the mean from your value, then divide by the standard deviation
• Compare the resulting z-scores, not the original raw numbers
• Remember that a higher z-score means "further from average," not automatically "better"
• Use the empirical rule as a quick gut-check for how unusual a z-score really is
• Double-check that both data sets used the same approach, population or sample, before comparing them directly

Once your numbers are ready, 100 Calculator's Z Score Calculator takes care of the arithmetic, so you can spend your time interpreting the result instead of double-checking your subtraction.

More From Our Math & Statistics Guides

Still building your statistics toolkit? These related guides dig deeper into the concepts covered above.

Frequently Asked Questions