Mean or median: which should you use when data has outliers?

Use the median first when you need a typical observation and a few values are much larger or smaller than the rest. Use the mean when you need a total amount distributed across all observations.

Question	Start with
What is a typical salary, house price or response time?	Median
What is the total workload per person or average resource use?	Mean
Are measurements fairly uniform with no clear extremes?	Mean and median together
Is a small dataset dominated by one suspicious value?	Source values, minimum and maximum, then both measures

Paste the values into the mean and median calculator. It shows both measures alongside count, sum, minimum, maximum and range. That context is more informative than either result on its own.

How one extreme value changes the answer

Consider five task durations:

48, 52, 55, 57, 310

The mean is 104.4: the sum of 522 divided by five. The median is 55, the middle value after sorting. If these are minutes spent on routine tasks, the median describes an ordinary case more closely, while 310 looks like a separate long incident.

The mean is not wrong, though. The five tasks took 522 minutes in total, or 104.4 minutes per task on average. Budgeting the median of 55 minutes for each task would reserve only 275 minutes and underbudget the batch by 247 minutes. Capacity planning must include the long incident when it represents real work.

The useful distinction is:

• the median marks the centre of the ordered observations;
• the mean includes every value's contribution to the total;
• the right choice follows from the decision you need to make.

When the median is more useful

The median is resistant to a small number of extreme values. It is often useful for prices, income, waiting times, request durations and similar data where most observations cluster together while a long tail extends in one direction.

A median price of 600,000 means at least half the observations are no higher and at least half are no lower. It does not mean most items cost exactly 600,000, and it does not estimate the total value of every item.

With an even number of observations, sort the list and average the two central values. In 2, 4, 7, 9, those values are 4 and 7, so the median is 5.5.

When the mean cannot be replaced by the median

Use the mean when the sum has operational meaning:

• requests a team must process;
• material consumed by a production batch;
• total equipment runtime;
• a score defined by an established averaging rule.

An extreme value may be rare but valid. Do not delete it merely to make the summary look stable. First check the source: were units mixed, was an extra zero added, and does the observation belong to the same time period and group?

Can the mean-median gap identify an outlier?

A wide gap is a useful reason to inspect the data, not a complete outlier test. A skewed distribution can be natural: a duration cannot go below zero, but it can occasionally become very large. Mean and median can also be close while unusual high and low values balance each other.

Inspect the minimum, maximum, range and original list. Formal outlier detection requires an explicit method and subject context. The calculator reports descriptive statistics; it does not label individual observations as errors.

A quick workflow for pasted data

1. Copy one numeric column from a spreadsheet or document.
2. Paste it into the calculator. It accepts spaces, line breaks, tabs, semicolons and unambiguous decimal points or commas.
3. Check the accepted count and any ignored tokens. A column heading or stray note should not disappear into the calculation unnoticed.
4. Compare mean and median, then review minimum, maximum and range.
5. Name the measure in your conclusion: write “median response time,” not just “the average.”

Keep the source list and your reason for selecting the measure. That makes the result reproducible and prevents choosing whichever number looks better after the calculation.