The Average Looks Fine. So What Are We Missing?

Why one perfectly correct number may still give us the wrong impression.

Imagine you're looking at the salaries of employees in a small company and you're told that the average monthly salary is $6,000. That sounds reasonably healthy. If that were the only number on a management report, you might assume that a typical employee earns somewhere around $6,000 a month.

Now let's look at the six salaries behind that average:

$3,000 · $3,200 · $3,400 · $3,600 · $3,800 · $19,000

The average really is $6,000. There is nothing wrong with the calculation. Yet five of the six employees earn less than $4,000.

So while $6,000 is mathematically correct, is it really a good description of what employees in this company typically earn?

That's one of the interesting things about averages. They can be completely accurate and still leave us with a misleading impression of the data.

An average is a summary, not the whole story

We use averages everywhere in business: sales per customer, employee performance, waiting time, satisfaction scores, transaction values. And for good reason. When we're dealing with hundreds or thousands of observations, we need a simple way to summarise them.

The problem begins when we treat that one number as though it describes everything happening underneath it.

In our salary example, the $19,000 salary pulls the average upwards considerably. Another way to describe the same group is to look at the median, the middle of the data once the salaries are arranged from lowest to highest. Because there are six employees, the middle falls between $3,400 and $3,600, giving us a median salary of $3,500.

So the same six salaries give us:

Mean: $6,000 Median: $3,500

Neither number is wrong. They're simply answering slightly different questions.

The mean and median are both correct. They simply tell us different things about what is typical.

This doesn't mean the median is automatically better. If I were calculating the company's payroll cost, that $19,000 salary absolutely belongs in the calculation. But if I wanted to understand what a typical employee earns, the median might give me a more useful picture.

So rather than asking “Should I use the mean or the median?”, I think the better question is “What am I trying to understand?”

The same average can hide very different experiences

Now imagine two customer service teams. Both report an average response time of 10 minutes, so on a dashboard they appear to be performing exactly the same.

Look underneath the average, however, and we find this:

Team A: 8, 9, 10, 10, 11, 12 minutes

Team B: 2, 3, 4, 16, 17, 18 minutes

Team A is relatively consistent. Most customers wait somewhere around 10 minutes. Team B is far less predictable. Some customers receive extremely fast service while others wait much longer.

Same average. Very different customer experience.

Both teams average 10 minutes. Looking at the spread reveals a very different customer experience.

This is where looking at variation becomes useful. The minimum and maximum immediately show us the extremes, while the range, the difference between them, gives us a simple indication of how widely the observations are spread.

Team A ranges from 8 to 12 minutes, a spread of just 4 minutes. Team B ranges from 2 to 18 minutes, a spread of 16 minutes. We can go further with measures such as standard deviation, which tells us more about how much the values tend to vary around the average.

For many business users, however, the formula isn't the most important place to begin. First look at the data and ask whether the observations are tightly clustered or scattered widely. The statistical measure simply helps us describe more precisely what we're already trying to understand.

Sometimes the average hides the people

Suppose your organisation reports an overall customer satisfaction score of 82. That sounds good.

But when you look more closely, you discover that new customers have an average satisfaction score of 91, while long-term customers score only 73.

The overall figure hasn't become incorrect. It has simply blended together two groups having very different experiences.

That could matter enormously. Perhaps a new onboarding process is working extremely well while long-term customers are becoming frustrated. Perhaps one customer segment receives better support than another. Perhaps improvement in one group is masking deterioration somewhere else.

A management team looking only at 82 might conclude that things are going well. Looking underneath the average could lead to a very different conversation.

An overall average can look healthy while important differences between groups remain hidden underneath it.

This is why descriptive statistics are more useful when we think of them as different views of the same data, rather than a collection of calculations we have to perform.

The mean helps us understand the overall average. The median can give us another view of what is typical, particularly when extreme values are present. Minimum, maximum, range and standard deviation tell us more about how widely the observations vary. Breaking the data into meaningful groups can reveal differences that disappear when everything is combined.

No single measure gives us the whole picture. The skill lies in knowing which views help answer the question in front of us.

Let the question decide what you look at

It's tempting to turn descriptive statistics into a checklist: calculate the mean, median, minimum, maximum and standard deviation, then put everything into a table.

We certainly can do that. But more numbers don't necessarily give us more insight.

If you're interested in overall cost or performance, the mean may be exactly what you need. If you want to understand what is typical and a few extreme observations are pulling the average around, the median may tell you something useful. If consistency matters, look at the variation. And when your data contains different types of customers, employees, branches or products, check whether the overall number is hiding meaningful differences between them.

The statistic should serve the business question, not the other way around.

AI can calculate everything. That doesn't mean we need everything.

AI makes descriptive statistics almost effortless. Give an AI tool a spreadsheet and it can calculate the mean, median, minimum, maximum, range and standard deviation in seconds. It can create charts, identify unusual observations and point out patterns that may deserve investigation.

That's enormously useful, but it also creates a temptation. Because we can calculate dozens of statistics, we may assume that more statistics automatically mean better analysis.

They don't.

A table containing twenty measures isn't necessarily more insightful than one containing three. What matters is whether those measures help us understand the business question we're trying to answer.

AI can tell us that the mean salary is $6,000 and the median is $3,500. It can calculate both perfectly. Someone still has to notice that the gap between those two numbers is interesting and ask why.

That's where the analytical thinking comes in.

Look underneath the number

Let's return to the salary example. If someone tells you that the average salary is $6,000, you now know that the number may be perfectly accurate. You also know that it may not be enough.

You might ask for the median, look at the lowest and highest salaries, or examine how salaries are distributed across different roles and levels. Not because you distrust the average, but because you want to understand what it is summarising.

The same habit applies to customer satisfaction, employee engagement, waiting times, sales performance and many of the other measures we regularly put on dashboards.

An average is useful precisely because it simplifies a lot of data into one number. But simplification always comes with a trade-off: some of the detail disappears.

Sometimes that detail doesn't matter. Sometimes it's exactly where the most important insight is hiding.

So the next time an average looks reassuringly normal, perhaps the question isn't whether the number is correct.

It's whether the number is telling you enough.

The average may summarise the data. Understanding what it hides is analysis.

New to Descriptive Statistics? A Quick Reference

Mean

The mean is what we commonly call the average. Add all the values and divide by the number of observations. It is useful for summarising data, but unusually high or low values can pull it towards them.

Median

The median is the middle value when observations are arranged from lowest to highest. It is less affected by extreme values and can sometimes provide a more useful picture of what is typical.

Minimum and Maximum

The minimum is the smallest observed value and the maximum is the largest. Together, they give us a quick view of the extremes in our data.

Range

The range is the difference between the maximum and minimum values. It provides a simple indication of how widely the observations are spread.

Standard Deviation

Standard deviation describes how much values tend to vary around the mean. A smaller standard deviation generally indicates values clustered more closely around the average; a larger one indicates greater variation.

Outlier

An outlier is an observation that sits unusually far from most of the other values. It isn't necessarily an error. Sometimes the outlier is precisely the observation worth investigating.