How to Calculate a Z-Score: Formula, Steps, and Reading the Table
Learn how to calculate z score with a simple formula, worked examples, how to read a z score table, and practical checks to verify your result.
Quick answer: how to calculate z score in one line
To calculate a z score, subtract the population mean from the raw score, then divide by the population standard deviation. That gives how many standard deviations the value is from the mean.
In formula form: z = (X − μ) / σ. Use the population mean (μ) and population standard deviation (σ) when you are comparing a single observation to a known distribution; use a sample standard error when your data are sampled and you plan to use a t-based approach instead.
Example in a single sentence: if the mean test score is 80, the standard deviation is 10, and a student scored 95, the z score is (95 − 80) / 10 = 1.5, so the student scored 1.5 standard deviations above the mean.
Quick checks: confirm the units match (raw score and mean in same units), check the standard deviation isn’t zero, and make sure you’re using population or sample formulas consistently. If you need to convert the z score to a percentile, read a standard normal table or use software to map z to cumulative probability.
z score interpretation: what a z-score tells you
A z-score expresses where a value sits relative to the center and spread of a distribution in units of standard deviations. Positive z-scores mean the value is above the mean; negative z-scores mean it is below. The magnitude (how large the absolute value is) indicates how unusual the value is relative to the distribution’s spread.
Interpreting the number depends on context: in a roughly normal distribution, about 68% of observations lie within ±1 standard deviation (z between −1 and 1), about 95% within ±2, and about 99.7% within ±3. So a z of 2.1 is uncommon but not impossible, whereas a z of 4.0 is extremely unusual and merits checking for data errors or outliers.
Z-scores also let you compare different measurements on a common scale. For example, converting SAT and ACT scores to z-scores shows which test performance was relatively stronger regardless of original scoring scales. Remember that similarity in z-scores does not remove differences in measurement reliability or test content.
Finally, a z-score is not a measure of effect size by itself unless you attach it to a sampling framework. For single observations, it describes relative position; for sample means compared to a known population mean, you should check whether the distributional assumptions and standard error used to compute the z are appropriate.
Key clues to check before trusting a z-score
Before you rely on a computed z-score, confirm the data and assumptions that went into it. Basic clerical mistakes—mixing units, using a sample standard deviation when the population SD is required, or mis-entering the mean—cause the most common errors. The bullets below are a compact checklist to run through.
These checks help you detect arithmetic errors and contextual mismatches quickly. If several clues flag a problem (for example, an implausibly large z combined with a tiny sample size), pause and re-evaluate whether the z-score is meaningful.
- Unit match: raw score and mean must be in the same measurement units (e.g., points, kilograms).
- Standard deviation source: confirm whether σ is a population SD or s is a sample SD and apply the appropriate formula.
- Arithmetic check: recompute numerator (X − μ) and denominator separately before dividing to catch sign errors or misplaced decimals.
- Distribution shape: if the underlying data are highly skewed, z-based probabilities from the normal table can be misleading.
- Sample size flag: a z derived from a small sample mean may need a t-distribution approach instead of a z-test.
- Outlier warning: an extreme z should trigger a review of the raw data point and the measurement process.
- Contextual plausibility: compare the z-score to realistic bounds for the measure (e.g., test scores with finite min/max).
- z score interpretation: consult how far the value is from the mean in standard-deviation units and whether that magnitude matters for your decision.
z score formula and step-by-step calculation
Use this sequence to compute a z score and check your work. The formal z score formula is z = (X − μ) / σ. Treat each element carefully: X is the individual score, μ is the distribution mean, and σ is the distribution standard deviation. If you only have a sample and intend to make inference about a population, consider whether a t-based approach is more appropriate.
Step 1 — Record the numbers: write down X, μ, and σ with units. Step 2 — Subtract the mean from the score to get the deviation: D = X − μ. Step 3 — Compute or confirm the standard deviation. Step 4 — Divide the deviation by the standard deviation: z = D / σ. Step 5 — Round sensibly and check sign and magnitude.
For a full self-check: compute D and σ using independent calculations or a calculator’s memory; then divide. Compare the result to the expected range. If you plan to use the z for a probability, map the z to a percentile using a standard normal table or software (see examples for how that mapping works).
Worked arithmetic example: suppose X = 85, μ = 100, σ = 15. Deviation D = 85 − 100 = −15. Divide: z = −15 / 15 = −1.0. That means the score is one standard deviation below the mean. If you need a probability, a z of −1.0 corresponds to about the 15.9th percentile on the standard normal cumulative scale.
- Write each number with units to avoid unit mismatch mistakes.
- Compute numerator and denominator on separate lines before the final division.
- If σ was computed from sample data, note whether you used n or n−1 in that calculation.
- When converting z to percentiles, use cumulative probabilities from the standard normal distribution.
Examples: worked z-score conversions and percentiles
This section shows common scenarios and how to proceed. Example A (single observation): a blood pressure reading is 140 mmHg, population mean 120 mmHg, population SD 10 mmHg. Compute z = (140 − 120) / 10 = 2.0. Interpretation: the reading is 2 SDs above the mean and is roughly in the 97.7th percentile for a normal distribution.
Example B (test comparison): two students take different exams. Student 1 scored 78 with mean 70 and SD 6; Student 2 scored 640 on a different test with mean 600 and SD 40. Compute z1 = (78 − 70) / 6 = 1.33 and z2 = (640 − 600) / 40 = 1.0. Student 1 performed relatively better compared to their cohort despite lower raw score because z1 > z2.
Example C (sample mean used for inference): a sample mean of 52 is compared to population mean 50 with population SD 5 and sample size 25. Use z for the sample mean’s standard error: SE = σ / sqrt(n) = 5 / 5 = 1, then z = (52 − 50) / 1 = 2.0. Interpret this as the sample mean being 2 SEs above the population mean; check assumptions before treating this as strong evidence.
Example D (percentile lookup): after computing z, you can find the percentile from the standard normal cumulative distribution. For z = −1.0, look up the cumulative probability ≈ 0.1587, or the 15.87th percentile. The next FAQ explains how to find that number in a printed z table and what the entries mean.
Limitations and when to double-check z-score results
Z-scores rely on correct inputs and assumptions. The primary assumptions are that the mean and standard deviation appropriately describe the distribution and that the normal approximation is reasonable for mapping z to probabilities. If the underlying data are skewed or heavily tailed, normal-table probabilities can be misleading.
Another limitation is sample information. When σ (population standard deviation) is unknown and estimated from the sample, using the z formula for inference underestimates uncertainty — in those situations a t-distribution is usually the correct choice. Small sample sizes increase this concern.
Rounding and measurement error also matter. A small arithmetic slip in the numerator or denominator can flip the sign or change the magnitude of the z-score. Always recompute the numerator and denominator separately and record intermediate results. If a z is extremely large (for example |z| > 4), inspect the original measurement and data collection process for errors or data entry problems.
When to get extra verification: if your decision depends on the z-score (grading, clinical thresholds, or publication), double-check using independent software, have another person verify the arithmetic, or use a tool to scan your working steps. Statikia can act as a verification pass, but you should still confirm numbers and assumptions manually before making important decisions.
Related guides
Verify your z-score steps with Statistics AI: Statikia
Want a quick verification? Use the Statikia app to scan or enter your numbers and check each arithmetic step and the final z-score. Treat the app’s result as a fast second check: always confirm your inputs, assumptions, and rounding by hand before making important decisions.
Frequently asked questions
How do I read a z score table when converting a z to a percentile for my sample?
A standard normal (z) table typically gives cumulative probabilities for z-values. Find the row for the z’s first two digits and the column for the second decimal place, or use the table’s format instructions. The table entry shows the area to the left of that z under the standard normal curve (the cumulative probability). For example, z = 1.25 corresponds to a cumulative probability around 0.8944, meaning about the 89.44th percentile. If the table lists tail areas instead, subtract the entry from 1 to get the cumulative left-area. When using a table, be careful with negative z-values: many tables list only positive z and require using symmetry (P(Z < −z) = 1 − P(Z < z)).
What is the z score formula and why does it look like that?
The z score formula is z = (X − μ) / σ. It standardizes a raw score X by measuring its distance from the mean μ in units of the distribution’s standard deviation σ. Dividing by σ converts raw deviations into a common scale so you can compare values from different distributions or compute probabilities using the standard normal distribution. The formula centers (subtract μ) and scales (divide by σ) the data; centering removes location differences and scaling accounts for dispersion.
How should I interpret a z-score of −2.5 in practical terms?
A z-score of −2.5 means the observation lies 2.5 standard deviations below the mean. In a roughly normal distribution, that corresponds to a cumulative probability near 0.0062 (about the 0.62nd percentile), so it is quite unusual. Practically, treat such an extreme value as a possible outlier: check the measurement, verify the calculation, and consider whether the observation represents a real rare event or an error. For decision-making, combine the z-score with subject-matter thresholds and the study’s sample size before drawing strong conclusions.
Can I use a z-score when I only have a small sample instead of the population?
You can compute a z-score from sample data using the sample mean and sample standard deviation, but using that z for inference (for example, to calculate p-values or confidence intervals) is risky for small samples. If the population standard deviation is unknown and the sample is small (commonly n < 30), a t-distribution approach that accounts for extra uncertainty is usually more appropriate. For descriptive comparisons, computing z-like standardized scores is acceptable, but label them clearly as sample-standardized values rather than population z-scores when reporting.