What an outlier does to the mean

Here are six commute times in minutes: 22, 25, 24, 23, 26 and 95. The last one was a day with an accident on the road.

• Total: 22 + 25 + 24 + 23 + 26 + 95 = 215
• Mean: 215 ÷ 6 = about 35.8 minutes

That figure is higher than five of the six commutes, and it would make your usual trip look far longer than it is. Without the outlier, the total is 120 and the mean is 120 ÷ 5 = 24 minutes. One value changed the result by almost 12 minutes.

With only six values, the effect is dramatic. In a data set of six hundred commutes, one bad day would hardly move the mean at all, which is why outliers matter most when your sample is small.

The median resists extremes

The median is the middle value of the sorted list: 22, 23, 24, 25, 26, 95. With six values, take the mean of the third and fourth, (24 + 25) ÷ 2 = 24.5 minutes. That is close to the typical journey, and it barely changed because the median only cares about position, not size.

This is why reports on salaries and house prices so often quote medians. A handful of very high values would pull the mean upward and mislead most readers.

How to spot an outlier

Look at the data before averaging. If one value sits far from the rest, ask three questions:

1. Is it a mistake? A misplaced decimal or a duplicated entry is a data error, and it should be fixed.
2. Is it real but rare? An accident, a holiday or a promotional bonus really happened but does not represent normal life.
3. Is it real and important? Sometimes the extreme value is the story. A sudden sales spike may be worth investigating, not removing.

A rough rule some people use is to flag anything far outside the range of the middle half of the data, but no formula replaces judgement about what the numbers represent.

Options for dealing with outliers

You rarely have to choose between ignoring a value and letting it dominate. Here are the usual options:

• Report the median alongside the mean. If they are far apart, say so.
• Calculate the mean with and without the outlier. Show both, and explain why they differ.
• Use a trimmed mean. Drop the highest and lowest values, then average the rest. In the commute example, dropping 22 and 95 leaves 23, 24, 25 and 26, with a total of 98 and a trimmed mean of 24.5 minutes.

Whatever you choose, be open about it. Quietly deleting inconvenient numbers is a way of distorting data, not cleaning it.

An outlier in your rewards

Cashback payouts show the same effect. Suppose your monthly payouts were $8, $9, $7, $8 and $40, where the $40 included a one-off welcome bonus. The total is $72, so the mean is $72 ÷ 5 = $14.40, but the median is $8.

If you want to plan what a normal month brings, $8 is the better guide. To check the regular figures, a cash back calculator can confirm what the usual spending should have earned, and you can treat the bonus as a separate, non-recurring item.

Outliers when you average over dates

Time-based averages have their own outliers. Averaging daily spending over a month that includes a holiday trip will give a number that does not reflect an ordinary month. If you are dividing a total by a number of days, confirm that figure first. A day counter can show the exact length of the period, and you can then calculate a separate average for the unusual days.

Using a tool without losing judgement

Once you have decided which values belong, an average finder can handle the arithmetic quickly. Run it twice if needed, once with all values and once without the outlier, and compare. The difference between the two results is itself information: it tells you how sensitive your conclusion is to that single point, and how cautious you should be when you quote it.

Common mistakes

1. Removing a value because it is inconvenient. Have a reason you can explain.
2. Ignoring a genuine signal. A single extreme value may point to a real problem.
3. Reporting only the mean. Without context, readers cannot tell whether one value is driving it.
4. Treating small samples as reliable. With only five values, one outlier has a very large effect.

Final thoughts

An outlier does not mean your data is broken. It means the mean needs context. Look at the numbers, decide whether the extreme value is an error, a rare event or an important signal, and report your average in a way that lets readers see what you did.