Mean vs Median
Average versus midpoint—compare two measures of centre and which one outliers can quietly distort.
Overview
Mean and median agree on tidy data and disagree exactly when the answer matters. Response times, salaries, and prices are usually skewed, so the mean sits above the typical case while the median stays where most observations are. Reporting only one number hides that shape, which is why performance work quotes percentiles instead of averages.
Pros
- Uses every value in the dataset
- Feeds standard deviation and variance
- Familiar to almost every audience
Cons
- One extreme value drags it far off
- Misleading for skewed distributions
Pros
- Robust against outliers by design
- Describes the typical case honestly
- Standard for latency and income data
Cons
- Ignores the magnitude of extremes
- Harder to combine across groups
Comparison table
| Aspect | Mean | Median |
|---|---|---|
| Outlier sensitivity | High | Low |
| Uses all values | Yes | Only the ordering |
| Typical reporting | Totals and rates | Latency, salary, house prices |
| Best fit | The data is roughly symmetric without extremes | The data is skewed or contains outliers |
Recommendation
Report the median for skewed data, add the mean when totals matter, and include a spread measure or percentiles so the shape is visible. State which statistic you used—“average” alone is ambiguous.
Related tools
Related articles
Related comparisons
Frequently asked questions
- Which should I use for response times?
- Median plus high percentiles such as p95 and p99. Averages hide the slow tail that users actually complain about.
- What does a large gap between them mean?
- The distribution is skewed or has outliers. Investigate the tail before summarising, because the extremes are usually the interesting part.
- What pushes someone toward Mean instead of Median?
- Mean wins when the data is roughly symmetric without extremes. The practical upside is that uses every value in the dataset, and feeds standard deviation and variance. The trade-off to watch is that one extreme value drags it far off.
- When does Median beat Mean for the same job?
- Reach for Median when the data is skewed or contains outliers. It gives you robust against outliers by design plus describes the typical case honestly, at the cost that ignores the magnitude of extremes.
- Can ToolHub help me try Mean and Median before I commit?
- Yes. The tools linked from each side of Mean vs Median run in your browser, so you can exercise Mean and Median with sample data without uploading anything.