Average Calculator: How It Works
'Average' is three different statistics wearing one name. Choosing the wrong one is how a summary ends up technically accurate and completely misleading, and this page covers when each is the honest choice.
The three averages
| Measure | What it is | Best when |
|---|---|---|
| Mean | Sum ÷ count | Data is roughly symmetric with no extreme outliers |
| Median | The middle value when sorted | Data is skewed or contains outliers |
| Mode | The most frequent value | Data is categorical, or you want the typical case |
Why the choice matters
Nine employees earn 30,000 and the owner earns 570,000.
| Mean salary | 84,000 |
|---|---|
| Median salary | 30,000 |
| Mode | 30,000 |
Both are correct. One describes nobody in the company. This is exactly why income, house price and waiting time statistics are reported as medians — a small number of extreme values pulls the mean far away from any typical case.
A useful diagnostic: when mean and median differ substantially, the data is skewed, and the median is usually the more honest summary.
Weighted average
When values do not carry equal importance:
Weighted mean = Σ(value × weight) ÷ Σ(weight)
Course grades are the everyday example. Scoring 90 on a 20%-weighted assignment and 70 on an 80%-weighted exam gives (90 × 0.2 + 70 × 0.8) ÷ 1 = 74, not the simple average of 80.
The averaging-averages error
Averaging two averages without accounting for their sizes produces a wrong answer. Class A: 30 students averaging 70. Class B: 10 students averaging 90.
- Naive: (70 + 90) ÷ 2 = 80 — wrong
- Correct: (30 × 70 + 10 × 90) ÷ 40 = 3,000 ÷ 40 = 75
This appears constantly in dashboards that average per-region conversion rates, per-day response times, or per-store margins without weighting by volume. Whenever you are averaging numbers that are themselves averages, you need the weights.
Geometric mean for rates
For growth rates and returns, the arithmetic mean overstates. An investment gains 50% then loses 50%: the arithmetic mean is 0%, but 100 becomes 150 and then 75 — a 25% loss.
The geometric mean handles this: multiply the growth factors and take the nth root. Here √(1.5 × 0.5) = √0.75 = 0.866, so −13.4% per year, which correctly compounds to the −25% outcome. Any time you are averaging percentage changes over periods, the geometric mean is the correct one.
Distributions with no mode, or several
Data where no value repeats has no mode; data with two equally frequent peaks is bimodal. Bimodality is worth noticing rather than smoothing over — it usually means two distinct groups have been mixed together, and the right response is to separate them and describe each rather than average across both.