percentage guide
Combine Percentages With Different Group Sizes
Calculate an overall rate from the original counts, see why averaging two percentages fails, and identify cases where pooling is inappropriate.
Published October 9, 2026 by SOLVEOZA Editorial
Quick answer
If 18 of 90 items pass one check and 1 of 10 pass another, the combined rate is 19 of 100, or 19%. Averaging the two rates, 20% and 10%, gives 15% and incorrectly gives the small group equal weight.
Write the question before choosing the average
An overall item-level pass rate asks how many items passed out of all items checked. An average group-level rate gives each group equal influence. Those are different questions, even when both results are called an average.
Use fictional quality-check batches here: batch A contains 90 items and batch B contains 10. Each item has one pass/fail outcome and appears in exactly one batch. These conditions make a pooled item-level rate meaningful.
Add counts, then divide
Recover the numerator and denominator for each group. Add all successes and all eligible observations separately. Divide the first total by the second. Multiplying a displayed rate by its denominator only recovers an exact count if the rate was not rounded.
| Batch | Passed | Checked | Rate |
|---|---|---|---|
| A | 18 | 90 | 20% |
| B | 1 | 10 | 10% |
| Combined | 19 | 100 | 19% |
Check the weighted expression
The equivalent calculation is (20 × 90 + 10 × 10) ÷ 100 = 19 when rates are expressed as percentages. Do not multiply the final 19 by 100 again. In decimal form, (0.20 × 90 + 0.10 × 10) ÷ 100 = 0.19, which is 19%.
Percentage Calculator applies a percentage to a base; it does not pool a dataset. Enter 20 and 90 to verify 18, then 10 and 10 to verify 1. Add those counts outside the tool and divide by 100. Do not enter 90 and 10 as two percentages to combine.
Know when a simple average works
With equal-sized groups, an unweighted average of exact rates agrees with the pooled rate. For example, 20 of 100 and 10 of 100 give 30 of 200 = 15%, also the average of 20% and 10%.
If group sizes are unavailable, the overall item rate is usually not recoverable. Report the group rates with their missing denominators rather than inventing weights. A zero-sized group supplies no observations; its percentage should not be inserted as a measured 0% rate.
Avoid pooling incompatible observations
Before adding denominators, check that the groups use the same outcome definition and time window. Do not add monthly unique-user counts if the same users can recur and then label the result unique people. Do not pool clicks per visit with purchases per customer.
Keep a small audit record containing each group label, passed count, eligible count and exclusions. If only rounded percentages remain, label any reconstructed combined rate an estimate and state the rounding. A mathematically correct weighted average cannot repair incompatible measurement definitions.
Methodology
- Compute the overall rate independently from total passed and total checked.
- Compare equal and unequal denominator examples.
Limitations
- Not a statistical inference, sampling correction or significance test.
- Overlapping populations and incompatible metrics require a different definition.
- Rounded source rates can prevent exact reconstruction of counts.
Self-calculated material
Examples use deterministic SOLVEOZA calculations. Original hypothetical batch counts and arithmetic, not measured business performance.
FAQ
Is a simple mean always wrong?
No. It answers an equal-weight group question, and agrees with a pooled rate when denominators are equal and rates exact.
What if I only have percentages?
You need the denominators to determine an item-level combined rate unless equal group sizes are established.
Can overlapping groups be pooled?
Not as a unique-item rate without accounting for duplication.
Can a zero-sized group have a measured zero rate?
No. Its denominator is zero, so no observed rate is defined.