TRIMMEAN in Google Sheets: Explicit Tail Removal and Reproducible Means

• Introduced: first introduction date is not established by the reviewed official history; no desktop release number is inferred.
• Availability: documented in the rolling Google Sheets service; these examples use the web worksheet workflow.
• Checked: October 14, 2026; official TRIMMEAN reference. Documentation and independent models, with no native Sheets execution or wider client certification.

=TRIMMEAN(data,exclude_proportion)

Both arguments are required. The proportion describes excluded observations, not a numerical cutoff on the measurement scale. A value of 0.2 represents 20%, not the number 20.

The complete Google body does not state how a requested noninteger or odd total removal count is rounded between tails. These examples deliberately use ten numeric observations and proportions 0,0.2,0.4,0.6, producing exact even removal counts.

The independent symmetric-tail model removes the same number of ordered observations from each end and averages the remainder. It is stated explicitly rather than importing Excel’s rounding rule into a Google guide.

Enter 3,4,5,6,8,10,15,30,60,200 in A2:A11:

=TRIMMEAN(A2:A11,0.2)

Expected result under the stated model: 17.25. Twenty percent of ten is two observations in total, so remove one low value, 3, and one high value, 200.

The retained values are 4,5,6,8,10,15,30,60, totaling 138. Dividing by eight gives 17.25. The source rows themselves need not be deleted; keep the full original dataset available for inspection.

The untrimmed total is 341, giving mean 34.1. The median is 9, halfway between 8 and 10. These are different summaries, not conflicting displays of the same statistic.

=TRIMMEAN(A2:A11,0.4)

Four observations are excluded in total: 3 and 4 at the low end, 60 and 200 at the high end. The remainder is 5,6,8,10,15,30, totaling 74.

Expected result: 37/3, approximately 12.333333. It is not a 40% reduction of the original mean. The proportion selects a count of observations to exclude, after which the remaining values are averaged.

=TRIMMEAN(A2:A11,0.6)

Remove three values from each end. The retained values 6,8,10,15 total 39, so expected mean is 9.75.

This still differs from median 9. A trimmed mean is not automatically a median. The fraction should be part of a documented reporting policy rather than adjusted until the result looks desirable.

At proportion zero, no observations are excluded in the mathematical model and expected output is 34.1. Proportion one is outside Google’s expressly documented domain, even though another application’s reference might phrase its upper boundary differently.

Use a second dataset 2,2,2,4,6,8,10,12,100,100. A 20% symmetric trim removes one 2 and one 100.

The retained observations are 2,2,4,6,8,10,12,100, totaling 144. Expected trimmed mean is 18.

Other copies of the extreme values remain. A filter deleting every value equal to the minimum or maximum would implement a different policy and remove a different number of records.

Reverse the original ten values. The sorted low and high tails are unchanged, so the 20% model still gives 17.25. A row’s worksheet position is not its numerical position in the tail calculation.

Add 5 to every original value and the expected trimmed mean becomes 22.25. Double each value instead and it becomes 34.5. Positive affine transformations preserve the same selected positions and transform their retained average consistently.

Noninteger and odd requested exclusion counts remain a native-rule hold here. The guide does not claim that Google’s implementation rounds or truncates them in any particular way. Exact error displays for invalid proportions, empty data and mixed-type inputs also remain unverified.

Trimming does not prove that discarded values were mistakes. Preserve original observations and report both the requested fraction and actual modeled retained count. AVERAGE and MEDIAN provide useful comparisons, while answering different statistical questions.

• Google TRIMMEAN reference.
• Editorial testing standards.

Examples use fictional data, English function names and comma separators. Expected outputs are independent mathematical calculations or explicitly documented rules, not native Sheets captures. Check the spreadsheet locale when reproducing punctuation-sensitive formulas.