The problem started with a spreadsheet I was proud of
A few years back I pulled customer order values for one of my businesses and reported the average order value to myself like it meant something solid. It was a healthy number. Then I sorted the same column low to high and found the median sitting well below it. Same data, two numbers, two different stories about the same customers. That gap is the whole subject of this article.
Neither number is wrong. They just answer different questions, and if you only ever look at one of them, you’re making decisions on half the picture.
What average and median actually calculate
The average, or mean, adds up every value and divides by how many values there are. Ten customers, ten order amounts, sum them, divide by ten. Every dollar every customer spent gets pulled into that one number, including the customer who spent ten times what everyone else did.
The median is different. Line every customer up from lowest value to highest value and pick the one sitting exactly in the middle. If you have an odd number of customers, it’s literally the middle record. If you have an even number, it’s the average of the two middle records. The median doesn’t care how big or small the extreme values are, only where they rank in the line.
That single difference, sensitivity to extreme values versus indifference to them, is what causes average and median to disagree.
A small example with real mechanics, not real customers
Say you run a small services business and look at nine customers’ monthly spend: $40, $45, $50, $50, $55, $60, $60, $65, and one outlier at $600 because a client bought a big one-off package.
Add those up and divide by nine and the average lands north of $110. But sort them and the middle value, the fifth one in line, is $55. Most of your customers are spending in the $40 to $65 range. The average is telling you something true about total revenue divided across the base. The median is telling you something true about what a typical customer actually pays. Report only the average and you’d walk into a pricing conversation thinking your typical customer spends double what they really do.
This is not a trick of made-up numbers. It’s just what happens any time one or two values sit far from the rest of the group. Retail, SaaS, services, it doesn’t matter what you sell. Wherever a small number of customers spend a lot more than everyone else, the average gets pulled toward them and the median doesn’t move.
Why this matters for the metrics you’re already tracking
Customer lifetime value. If you calculate average CLV, a handful of big accounts can make your whole customer base look more valuable than it is. That’s fine if you’re trying to estimate total revenue potential. It’s a problem if you’re using that average CLV to decide how much you can afford to spend acquiring a typical new customer, because that typical customer probably isn’t going to behave like your biggest ones.
MRR per customer. Same shape of problem. One enterprise-ish account paying five times the standard plan will drag your average revenue per user up. If you’re setting expectations for a new sales hire based on that average, you’re setting them up to chase a number most deals won’t hit.
Deal size in a sales pipeline. Average deal size gets quoted in forecasts all the time. If your pipeline has a mix of small self-serve deals and the occasional large contract, the average can make a quarter look stronger than the bulk of your actual deal flow supports.
Time to convert, time to churn. Averages here get skewed by the rare customer who took forever to convert, or the rare customer who stuck around for years past when most people leave. The median gives you a cleaner read on what a normal customer’s timeline actually looks like.
None of this means average is a bad metric. It means average answers “what’s the total divided evenly,” and median answers “what does the middle, typical case look like.” You need both questions answered, not one instead of the other.
How to check which one you’re dealing with
You don’t need a statistics course for this, just a sorted list. Pull the raw values behind whatever average you’re looking at, sort them, and find the middle one. If the median is close to the average, your data is fairly evenly spread and either number is a decent stand-in for the other. If the median is meaningfully lower than the average, you’ve got some high-value outliers pulling the mean upward. If the median is meaningfully higher than the average, you’ve got a cluster of low values dragging the mean down, which happens too, just less often in revenue data.
A quick way to see the shape without doing full quartile math: look at the minimum, the median, and the maximum side by side. If the distance from median to maximum dwarfs the distance from minimum to median, you’ve got a long tail of high spenders and your average is being pulled by them.
What median hides too
Median isn’t some purer, more honest number that beats average in every case. It hides things on its own.
It doesn’t tell you anything about your best customers. If you’re trying to understand what drives your top 10% of spenders so you can find more people like them, the median is useless to you, it’s built specifically to ignore that group. It also doesn’t tell you total revenue. You can’t multiply median order value by customer count and get your actual sales total, because the median deliberately excludes the size of the outliers. For that you need the sum, or the average times the count.
Median also gets less stable with small sample sizes. With nine customers, one new customer joining can shift which record sits in the middle. With nine hundred customers, one new signup barely moves it. If you’re a small business looking at monthly numbers with a handful of transactions, don’t treat the median as a fixed truth, treat it as a snapshot that will jump around until your sample gets bigger.
What I actually do with this
I stopped reporting a single number for anything that touches money per customer. When I look at order value, deal size, or revenue per account, I pull the average and the median side by side, and if they’re far apart I go look at the distribution itself, not just the two summary numbers. Usually that means sorting the list and eyeballing where the outliers sit and how many of them there are.
That’s it. It’s not a formula, it’s a habit of asking “am I looking at the total spread evenly, or am I looking at what’s typical,” and making sure I know which question I’m actually answering before I act on the number.
Neither average nor median is going to tell you why a customer behaves the way they do, and neither one replaces actually looking at your data broken into segments. They’re both compressions of a lot of individual customers into one number, and any compression loses information. The goal isn’t to find the one true metric, it’s to know what each one is throwing away so you don’t get surprised by it later.
If you want more of this kind of plain walkthrough on the metrics and methods that actually run a small business, come take a look at Data Research Analysis Collection.
Get new guides and videos first — join the Telegram channel.