Why You Can't Just Average Percentages: The Weighted Average Trap

Two departments grow 10% and 30% respectively — what’s the company’s overall growth rate? The intuitive answer, 20% (the simple average), is wrong almost every time this scenario comes up in the real world, and understanding exactly why reveals a genuinely important, widely-made mistake in how percentage changes get combined.

The trap, worked through in full

Consider a company with two divisions: Division A has $1,000,000 in revenue and grows 10% to $1,100,000. Division B has $100,000 in revenue and grows 30% to $130,000. The simple average of 10% and 30% is 20% — but that’s not the company’s actual combined growth rate.

The actual combined figure: total old revenue is $1,100,000; total new revenue is $1,230,000. Combined growth = ($1,230,000 − $1,100,000) ÷ $1,100,000 = 11.8%, dramatically lower than the naive 20% average — because Division A, which grew more slowly, represents a much larger share of the total base than Division B, which grew faster but from a much smaller starting point.

Why this happens: weight, not just percentage, matters

A simple average treats both percentages as equally important, regardless of the size of what they’re each a percentage of. But a combined total is genuinely dominated by whichever component contributes more dollars (or units, or whatever’s being measured) to that total — Division A’s $1,000,000 base contributes far more to the combined total than Division B’s $100,000 base, so the combined growth rate is pulled much closer to Division A’s 10% than to Division B’s 30%, exactly reflecting each division’s actual weight in the total.

The correct method: weighted average

The correct approach never actually requires “weighting” as a separate explicit step — it falls out naturally from calculating the combined change the same way you’d calculate any single percentage change: total change in the numerator, total original base in the denominator. This is exactly the standard percentage-change formula applied to combined totals rather than to a single figure, which is why it’s easy to get right once you resist the temptation to average the already-computed percentages directly.

When simple averaging IS valid

Scenario	Simple averaging valid?
Two percentages from equal-sized bases	Yes — with equal weight, simple average matches weighted average
Two percentages from meaningfully different bases	No — must weight by base size
Averaging percentage rates across many equal-sized time periods	Often reasonable, if periods are genuinely comparable in size
Combining growth rates across unequal business units, regions, or samples	No — almost always needs proper weighting

It’s worth noting simple averaging isn’t always wrong — if the two underlying bases happen to be exactly equal, a simple average and a properly weighted average produce the identical result, since equal weighting is what a simple average implicitly assumes. The mistake specifically appears when bases differ, which is the overwhelmingly common real-world case for anything involving business divisions, demographic subgroups, or unevenly-sized samples.

A second real-world example: survey response rates

Imagine combining survey results from two groups: Group 1 has 900 respondents with a 40% approval rate; Group 2 has 100 respondents with an 80% approval rate. The naive average of 40% and 80% is 60% — but the actual combined approval rate is (900×0.40 + 100×0.80) ÷ 1,000 = (360 + 80) ÷ 1,000 = 44%, much closer to Group 1’s rate because Group 1 represents 90% of the total respondents. Reporting “60% combined approval” here would be a meaningfully misleading overstatement of the true combined figure.

Getting the combined figure right

Calculating a correct combined or overall percentage change means going back to the actual underlying totals — total old value and total new value — rather than averaging already-computed individual percentages. The Percentage Change Calculator applies the correct old-value-denominator formula once you’ve summed the actual totals correctly; the Ratio Percentage Calculator is useful for the related task of expressing a part-to-whole relationship (like a weighted contribution) as a percentage directly.

FAQ

Is averaging percentages ever mathematically valid? Yes, specifically when the underlying bases being combined are equal in size — in that special case, a simple average and a properly weighted average produce the same result.

What’s the correct formula for combining multiple percentage changes from different bases? Sum the actual before-and-after totals across all components, then apply the standard percentage-change formula (new total minus old total, divided by old total) to those combined totals.

Why is this called a “weighted average” problem? Because the mathematically correct combined figure implicitly weights each component’s percentage by its share of the total base — larger-base components pull the combined figure closer to their own rate.

Does this mistake only apply to business growth rates? No — it applies to any scenario combining percentages from differently-sized underlying groups, including survey results, test score improvements across classes of different sizes, or defect rates across production batches of different volumes.

How can I quickly sanity-check whether a reported “average percentage” might be misleading? Ask whether the underlying groups being averaged have meaningfully different sizes — if they do, a simple average is very likely misleading, and the actual combined totals should be calculated directly instead.

Is this related to Simpson’s paradox? Yes, closely — Simpson’s paradox describes cases where a trend present in separate groups reverses or disappears when the groups are combined, and naive percentage averaging without weighting is one common way that kind of misleading aggregate figure gets produced. This is a distinct issue from stacking successive percentage changes on a single value — both are common sources of percentage-math errors, but they arise from different situations.