Averages vs Medians: The Number That Describes Nobody

What you'll learn

Tell the mean from the median without hesitating
Spot skewed data — the situations where the average quietly lies
Know when to ask "what's the median?" and why it changes decisions

Someone in a meeting says “our average deal size is $48k,” and the room starts planning around $48k deals. Here’s the problem: it’s entirely possible that not one single deal looks like $48k — that most deals are $10k and one whale was $500k. The last few lessons covered where data comes from and what kind of number you’re even looking at. Now we start on the other half of the skill: what the numbers actually say once they reach your screen — beginning with the most common number in corporate life, and the most commonly misleading one: the average.

Two ways to summarize a pile of numbers

Both the mean and the median answer “what’s typical?” — but they answer it in very different ways, and skewed data pulls them apart.

The words that matter

Mean: What people usually mean by “average” — add everything up, divide by the count. Every value moves it, including the crazy ones.
Median: Line all the values up in order and take the middle one. Half are above, half below. Outliers barely touch it.
Outlier: A value wildly different from the rest — the $500k whale deal, the CEO’s salary, the one ticket that took 40 days.
Skewed data: Data with a long tail on one side — most values small, a few huge. Salaries, deal sizes, house prices, wait times.
Distribution: The overall shape of the data — where values cluster, how far they spread. The thing a single number can never fully capture.
Percentile: The value below which X% of the data falls. The median is the 50th percentile; dashboards love the 95th (“p95”).

Picture a ten-person company. Nine people earn between $45k and $55k. The CEO earns $1M. Add it all up and divide by ten: the mean salary is $145k. The median — line everyone up, take the middle — is about $50k. Now, which number is “the truth”? If you’re a recruiter writing a job ad, $145k is a fantasy. If you’re describing what a typical person earns there, the median is the honest answer. The mean isn’t wrong — it’s just answering a different question (“if we split total payroll evenly, what would each person get?”), and nobody in the room is asking that one.

One outlier drags the mean to $145k while the median stays home at $50k — and only one of those numbers describes a real person.

Text description of this diagram

A horizontal salary line runs from $0 to $1M. Nine blue dots appear one by one, stacked in a tight cluster near the $50k mark — the nine ordinary salaries. Then a red dot lands far to the right at $1M: the CEO. A green marker labeled median ≈ $50k stands in the middle of the cluster and never moves. A blue marker labeled mean = $145k starts at the cluster and slides visibly to the right, dragged by the single outlier, until it stands in empty space where no actual person’s salary sits. The amber banner makes the point: $145k describes nobody in the room.

When the average lies — and when it’s fine

The mean misleads exactly when data is skewed — a big cluster of ordinary values plus a long tail of extreme ones. And an uncomfortable amount of business data is shaped like that: deal sizes, salaries, customer spend, house prices, support resolution times, page load times, insurance claims. In all of these, a few giants drag the mean up and to the right of where “typical” lives. That’s why economists report median household income and realtors quote median home prices — and why your “average deal size” deserves suspicion.

The mean is perfectly fine when data is roughly symmetric (heights, test scores, daily temperature) — mean and median land close together and the argument is academic. And the mean is the right tool when you genuinely care about totals: “average revenue per store × number of stores = total revenue” works precisely because the mean is built from the total. Forecasting, capacity planning, budgeting — mean territory. Describing a typical customer, employee, or ticket — median territory.

One more tool worth recognizing: percentiles. Engineering dashboards rarely show average response time; they show p95 — the time 95% of requests beat — because the mean hides the slow disasters users actually complain about. If you’ve seen p50/p95/p99 on a dashboard and nodded politely, that’s all it is: the median and its stricter cousins.

Common misunderstanding

“The average customer spends $180, so let’s design for the $180 customer.” If your customers are actually two clusters — hobbyists at $40 and pros at $900 — then the average customer doesn’t exist, and a product designed for them serves no one. When two groups hide inside one number, the mean lands in the empty space between them. Always ask what the distribution looks like before designing for its center.

Spot it: mean or median?

Try this at work

Next time someone quotes an average in a meeting, ask one question: “What’s the median?” If the two are close, carry on — the average is honest. If they’re far apart, you’ve just found a tail of outliers, and the follow-up (“what’s dragging it?”) is usually where the actual insight lives. One question, ten seconds, and you’ll never be the person planning around a $48k deal that doesn’t exist.

The bottom line

The mean adds and divides; the median lines up and picks the middle. On skewed data — salaries, deals, wait times — the mean drifts toward the outliers and describes nobody. “What’s the median?” is the cheapest smart question in corporate life.

Why it matters

Almost every business number you’ll ever hear is a summary — and the summary hides as much as it shows.

Averages drive real decisions: pricing, hiring bands, staffing levels, SLA promises. Get “typical” wrong and every downstream decision inherits the error. The rest of this course builds on this move — asking what’s behind the summary number — for percentages (next module), charts, correlations, surveys, and A/B tests. Same skill, six disguises.

Quick check

1. Nine employees earn about $50k and the CEO earns $1M. The mean salary is $145k. What's the best description of a "typical" salary?

2. When is the mean the right tool?

3. Mean resolution time is 9 hours but the median is 2 hours. What does the gap tell you?

Answers explained

B is correct — salaries are skewed data, and the single $1M outlier drags the mean to a value nobody actually earns; the median stays with the real cluster. (If you picked A: including everyone is exactly how the mean gets distorted — fairness to values isn’t accuracy about people. If you picked C: splitting the difference invents a third number that describes nobody either.)
C is correct — the mean is built from the total, so it’s the honest tool for totals, forecasts, and roughly symmetric data. (If you picked A: skew is precisely when the mean misleads. If you picked B: “always use the median” throws away the mean’s real job — anything built on totals.)
A is correct — a mean far above the median is the signature of a long tail: a few extreme values pulling the average up. (If you picked B: the gap is a property of skewed data, not an error. If you picked C: the median tells you most tickets close around 2 hours, not 9.)