TRIMMEAN in Excel: Symmetric Tail Counts and Fractional Trimming
Calculate Excel trimmed means by inspecting the actual removed observations, even-count rounding and the difference between trimming and deleting every tied extreme.
Verification status: See the evidence and test scope described in this guide. How formula examples are checked
Quick answer: TRIMMEAN(array,percent) removes a specified total fraction of observations symmetrically from both tails, then averages the remaining values. The fraction applies to the total removed count, not to each tail separately.
Availability and version history
- Introduced: first introduction version is not established by the reviewed official history; the current product list is not a launch record.
- Availability: official help lists Excel for Microsoft 365, Excel for Microsoft 365 for Mac, Excel 2024, Excel 2024 for Mac, Excel 2021, Excel 2021 for Mac, Excel 2019, Excel 2016.
- Checked: October 14, 2026; official TRIMMEAN reference. Documentation and independent models; no native Excel execution or wider platform certification.
Syntax and the even-count rule
=TRIMMEAN(array,percent)
Both arguments are required. Microsoft rounds the number of excluded observations down to the nearest multiple of two, preserving equal counts at the low and high ends.
For n numeric observations, the independent count model is 2*INT(n*percent/2). Half that count is removed from each end of a conceptual sorted list. The formula does not require deleting rows from the source range.
Microsoft specifies #NUM! when percent is below 0 or above 1. The exact outcome at percent=1 and other no-surviving-data cases is not separately established here; the worked controls retain observations.
Example 1: inspect exactly which values remain
Enter 1,2,3,4,5,6,8,20,30,100 in A2:A11:
=TRIMMEAN(A2:A11,0.2)
Expected result: 9.75. Ten times 0.2 is two, so remove one observation from each end: 1 and 100. The remaining values are 2,3,4,5,6,8,20,30, totaling 78. Dividing by eight gives 9.75.
The untrimmed arithmetic mean is 17.9, from 179/10. The median is 5.5, halfway between 5 and 6. These different summaries do not disagree because of rounding; they use different definitions.
Example 2: a requested fraction may trim fewer rows
=TRIMMEAN(A2:A11,0.3)
Expected result is still 9.75. Ten times 0.3 is three, but the excluded count rounds down to two. The formula removes one low and one high observation, not one from one end and two from the other.
At fraction 0.19, the requested count is 1.9 and rounds down to zero. Expected result is the untrimmed 17.9. The function can change in steps as the requested fraction crosses an even-count boundary.
Advanced example: increase the actual tail removal
=TRIMMEAN(A2:A11,0.4)
=TRIMMEAN(A2:A11,0.6)
At 0.4, remove two observations from each tail. The retained values 3,4,5,6,8,20 total 46, giving expected 23/3, approximately 7.666667.
At 0.6, remove three observations from each tail. The retained values 4,5,6,8 total 23, giving expected 5.75.
The last result still differs from median 5.5. A heavily trimmed mean is not automatically a median, and the chosen trimming fraction should be explained rather than selected merely to obtain a preferred answer.
Advanced example: ties are observations, not deletion rules
Use a separate ten-value dataset 1,1,1,2,3,4,5,6,100,100. With fraction 0.2, remove one low observation and one high observation.
The retained values are 1,1,2,3,4,5,6,100, whose total is 122. Expected trimmed mean: 15.25.
Two other copies of 1 and one copy of 100 remain. TRIMMEAN does not remove every row equal to an extreme value. Filtering out all minimum and maximum values would remove a different number of records and produce a different statistic.
Advanced checks: ordering and unit changes
Reverse the original ten values and the same conceptual tail selection gives expected 9.75 at 0.2. The sorted values determine the tails, not their worksheet row order.
Add 10 to every original observation. After the same symmetric count removal, expected trimmed mean becomes 19.75. Multiplying all original values by two gives 19.5. These controls help distinguish a unit or origin change from a changed trimming policy.
Data policy and unresolved boundaries
Trimming is a statistical summarization choice, not proof that removed values are errors. Keep the original data, the requested fraction, the actual count removed and the remaining count available for review.
The compact dedicated reference does not specify every text, Boolean, empty-cell or source-error combination. The examples use complete numeric values. AVERAGE and MEDIAN provide useful comparisons, but neither should be silently substituted for the selected trimmed-mean definition.
References and verification
Expected outputs are independent mathematical calculations or explicitly documented rules, not captured native Excel results. Examples use fictional data, English function names and comma separators. Native floating-point boundaries remain untested.