This average percentage calculator exists because averaging percentages is one of the most common quiet errors in everyday reporting. Adding up several percentages and dividing by how many there are only gives the right answer in one specific situation: when every percentage came from a group of exactly the same size. The rest of the time the correct answer is a weighted average, and the gap between the two can be large enough to reverse a conclusion.
Arb Digital sees this constantly in marketing reporting, where a monthly conversion rate is calculated by averaging four weekly conversion rates, or an account-level click-through rate is calculated by averaging the rates of thirty campaigns. Both are wrong whenever the weeks or campaigns differ in size, which is always. The tool above does it correctly, shows the naive answer alongside for comparison, and reports the difference in percentage points so you can see how much the shortcut cost.
What This Average Percentage Calculator Does
You enter up to five percentages together with the group size each one was measured on. The tool reconstructs the underlying count for each row, adds those counts, adds the group sizes, and divides. That gives the pooled rate: the percentage you would have got if you had combined all the raw data first and calculated a single percentage from the total.
It also computes the plain arithmetic mean of the percentages, ignoring group sizes, and reports the difference. That second figure is not there for decoration. Seeing the size of the error on your own numbers is far more convincing than being told the shortcut is wrong, and it tells you whether the distinction matters in your particular case — sometimes it does not, and the tool will show you that too.
Our weighted average calculator performs the same underlying arithmetic in general form, for any values and any weights. This page is the specialised version: it assumes the values are percentages, it assumes the correct weights are the denominators those percentages came from, and it frames the whole result around the mistake that specialisation is designed to prevent. If you are averaging things that are not percentages, or your weights are importance scores rather than sample sizes, use the general tool.
How to Use It
- Enter each percentage in the left-hand box of its row, as a number rather than a fraction — 45 rather than 0.45.
- Enter the group size that percentage was calculated from in the right-hand box. This is the denominator: people, sessions, items, marks available.
- Leave unused rows at zero. Any row with a group size of zero is skipped entirely.
- Read the weighted figure in the headline. That is the answer you should quote.
- Check the error figure in the grid. If it is a fraction of a percentage point the distinction is academic; if it is several points, the plain mean would have misled you.
The Formula and How It Is Calculated
The weighted average of percentages is (∑ pᵢ × nᵢ) ÷ ∑ nᵢ, where pᵢ is each percentage and nᵢ is its group size. Written that way it looks like a formula to memorise. It is easier to understand as reconstruction: pᵢ × nᵢ ÷ 100 recovers the count behind each percentage, and dividing the total count by the total group size gives the overall rate. You are undoing the percentages, combining the raw data, and taking the percentage once at the end — which is what you should have done in the first place.
The defaults show why it matters. Group one is 90 percent of 10, which is 9. Group two is 50 percent of 100, which is 50. Group three is 80 percent of 40, which is 32. The counts total 91 and the group sizes total 150, so the pooled rate is 91 ÷ 150 = 60.67 percent. The plain mean of 90, 50 and 80 is 73.33 percent. The shortcut overstates the result by 12.67 percentage points, and it does so because it gives a group of 10 exactly the same say as a group of 100.
Notice the direction of the error. The plain mean always pulls towards whichever groups are smallest, because it treats every group as equally important regardless of how little data stands behind it. When the small groups happen to have extreme percentages — which is exactly what small groups tend to have — the distortion is at its worst. The NIST/SEMATECH e-Handbook section on measures of location sets out why the choice of average has to follow the structure of the data rather than convenience.
When the Plain Mean Is Actually Correct
There are two cases where averaging the percentages directly is right, and it is worth knowing them so you are not weighting things that should not be weighted.
The first is equal group sizes. Five test sections each marked out of 20, five surveys each of 500 people, twelve months each measured on the same fixed denominator — in all of these the weights are identical, the weighting cancels out, and the plain mean is the weighted mean. The calculator demonstrates this: set every group size to the same number and watch the two figures converge exactly.
The second is when the percentages are not rates at all but scores of equal standing. If five judges each award a percentage and you want the average judge's opinion, the judges are the unit of analysis and each carries equal weight by design. The distinction is what the question is about. If you are asking "what proportion of all the items passed", weight by group size. If you are asking "what was the average of these five figures, treated as five equally important observations", do not.
The Marketing Version of This Mistake
Averaging conversion rates across campaigns is the most expensive form of this error, because the result usually flatters a poor account. A campaign with 20 sessions and one conversion has a 5 percent rate; a campaign with 20,000 sessions and 200 conversions has a 1 percent rate. The plain mean is 3 percent. The real account-level rate is 201 conversions from 20,020 sessions, which is 1.004 percent. A report quoting 3 percent is out by a factor of three, and the error is invisible unless someone checks.
The same structure produces misleading click-through rates, bounce rates, open rates and completion rates. The rule is simple and worth writing on a wall: never average a rate. Add the numerators, add the denominators, divide once. If you are combining rates across time periods or channels, our percentage calculator handles the individual conversions and the percentage change calculator compares two periods once you have the pooled figures right.
Simpson's Paradox: When Weighting Reverses the Answer
There is a stronger version of this problem in which combining groups does not merely shift the average but flips the comparison. Suppose method A scores 80 percent on a group of 10 and 40 percent on a group of 100, while method B scores 75 percent on a group of 100 and 35 percent on a group of 10. Method A wins on both groups individually. Pooled, A gets 48 out of 110 or about 44 percent, and B gets 78.5 out of 110 or about 71 percent. B wins overall despite losing every subgroup.
This is Simpson's paradox, and it happens whenever the group sizes are distributed very differently between the things being compared. It is not a trick of arithmetic; both answers are correct answers to different questions. The subgroup comparison answers "which method is better for people like these", and the pooled comparison answers "which method produced more successes across the population we actually treated". Reporting only one of them without saying which is how a genuinely accurate table becomes genuinely misleading. Penn State's STAT 500 applied statistics course covers the conditional-probability structure behind this in its lessons on summarising and comparing data.
Percentages of Percentages and Other Traps
Two related mistakes are worth separating from this one, because they look similar and have different fixes. The first is averaging percentage changes rather than percentage levels. A value that rises 50 percent then falls 50 percent has not returned to where it started; it is down 25 percent. Growth rates compound multiplicatively, so the correct average of a series of growth rates is the geometric mean, not the arithmetic one, and our geometric mean calculator handles that case.
The second is averaging rates expressed per unit, such as speed or cost per acquisition, where the correct combined figure is a harmonic mean rather than an arithmetic one. Driving 60 km/h out and 30 km/h back does not average 45 km/h over the trip, because you spend twice as long at the slower speed; the honest answer is 40 km/h, which the harmonic mean calculator produces. The common thread across all three traps is that the arithmetic mean assumes each value carries equal weight over the same base, and whenever it does not, a different average is required. Our mean median mode calculator is the right starting point when you simply want to describe a set of values rather than combine rates.
Arb Digital audits paid search and analytics reporting and rebuilds the numbers so account-level figures reflect the data underneath them.
See Our Google Ads Services Talk To Our TeamCommon Mistakes to Avoid
- Averaging rates instead of pooling them — add the numerators, add the denominators, and take the percentage once at the end.
- Using the wrong denominator as the weight — the weight must be the base the percentage was calculated from, not a revenue figure or an importance score.
- Including rows with a group size of zero — a percentage with no data behind it contributes nothing and should not drag the answer.
- Averaging growth rates arithmetically — compounding changes need a geometric mean, or the result will overstate the trend.
- Quoting a pooled figure without the subgroups — when group sizes are lopsided, the pooled number can disagree with every subgroup it contains.
Related Free Tools From Arb Digital
Handle any weighted mean with the weighted average calculator, work out a single percentage with the percentage calculator, compare two figures with the percentage difference calculator, average compounding growth with the geometric mean calculator, or summarise a data set with the mean median mode calculator. The full free online tools hub lists every statistics tool we publish.
Frequently Asked Questions
Only when every percentage was calculated from a group of exactly the same size. If the group sizes differ, that shortcut gives too much influence to the smallest groups and the answer will be wrong.
Multiply each percentage by its group size, add those products, and divide by the total group size. That is the same as adding all the raw counts, adding all the denominators, and taking the percentage once.
The denominator that each percentage was calculated from. If a rate was 40 percent of 250 sessions, the weight is 250. Using anything else, such as revenue or a subjective importance score, answers a different question.
Because campaigns differ enormously in traffic. A tiny campaign with a freak rate counts as much as a huge one in a plain mean, which can move the reported figure by a factor of several.
Whenever all the group sizes are equal. The weighting then cancels out exactly, and the plain mean and the weighted mean give identical results.
Not with this method. Growth rates compound, so the appropriate average is the geometric mean of the growth factors rather than the arithmetic mean of the percentages.
It is the situation where one option beats another in every subgroup yet loses once the subgroups are pooled, because the group sizes are distributed very differently between the two options.
This page explains a statistical calculation for educational purposes. The result depends entirely on using the correct denominators as weights, so confirm what each percentage was measured on before entering it.
