Trimmed Mean Calculator
A trimmed mean is an average taken after dropping the most extreme values from both ends of a dataset, so one wild outlier cannot drag the whole result with it. Paste your values, pick a trim percentage, and this calculator sorts the data, removes the same count from the low end and the high end, and reports the original mean beside the trimmed mean, plus exactly which values were removed and which remain. The worked example is pre-filled: ten values with a 20% trim, where the outlier 95 lifts the plain mean to 32.100000 but the trimmed mean is 26.750000.
Trimming convention: the percentage is the total share removed, split equally across both tails, with the per-tail count rounded down. 10% on 20 values removes 1 from each end (2 in total), not 2 from each end. This matches Excel’s TRIMMEAN function. If a tool you used before removed the percentage from each end separately, enter double the percentage here.
Quick Answer
With the pre-filled values 12, 15, 18, 22, 25, 28, 31, 35, 40, 95 and a 20% trim, the calculator removes 12 and 95 and reports a trimmed mean of 26.750000, against an original mean of 32.100000. The single outlier 95 pulls the plain mean up by more than 5 points; trimming drops it, and the low value 12 with it, so the average describes the bulk of the data instead of its extremes. Paste your own values above to get the same four outputs: original mean, trimmed mean, values removed and remaining values.
Key takeaways
- A trimmed mean is a different statistic from the plain mean: it averages only the middle of the sorted data after dropping both tails.
- The trim percentage here is the TOTAL share removed, split equally across both tails. 20% on 10 values removes 1 from each end, not 2.
- The per-tail count is rounded DOWN, so a small dataset at a small percentage may trim nothing at all. 10% on 7 values removes zero values, and the calculator says so plainly with “None”.
- Values are sorted before trimming, and the removed list shows the low-end values first, then the high-end values, whatever order you typed them in.
- Quick Answer
- How to use the trimmed mean calculator
- The trimming rule, stated exactly
- Worked example: ten values, 20% trim, one big outlier
- Trimmed mean vs plain mean vs median: which one answers your question
- Input traps: commas, spaces, and numbers the tool reads loosely
- What competing calculators miss
- Frequently asked questions
- Add This Calculator to Your Website
- References
How to use the trimmed mean calculator
Paste two things and click Calculate. The results stay hidden until you calculate, so nothing is shown for the pre-filled example until you press the button.
- Enter your data values, separated by commas: the box is pre-filled with 12, 15, 18, 22, 25, 28, 31, 35, 40, 95. Decimals and negatives are fine, and spaces around a value are fine. You need at least 2 values.
- Enter the trim percentage (0 to 50): this is the total share of values to remove, not the share per end. 20 is pre-filled.
- Click Calculate: the original mean, trimmed mean, values removed and remaining values appear, all means to six decimal places.
- Copy or share the result with the buttons in the results box when you want the four outputs together.
Invalid input is rejected with a named message: blank data gives “Please enter data values.”, a trim outside 0 to 50 gives “Please enter a valid trim percentage between 0 and 50.”, a single value gives “Please enter at least 2 data values.”, and a non-numeric entry gives “Error: Invalid number:” followed by the offending entry. Reset restores the worked example and hides the results.
The trimming rule, stated exactly
Trimming has three steps, and the only step where tools disagree is the second one, so it is stated here in full. First, sort the values from smallest to largest. Second, compute how many to drop from each end: trim count = floor(trim percentage ÷ 100 × number of values ÷ 2). Third, average what is left: trimmed mean = sum of remaining values ÷ number of remaining values.
trimmed mean = (sum of remaining values) ÷ (number of remaining values)
0% trim = the plain mean (nothing removed) 50% trim = the middle portion averaged (at most a quarter of the values from each end, so at least half the values always remain)
The floor in that formula is doing real work. It keeps the trim symmetric, one whole value from each end at a time, and it is the same convention Excel’s TRIMMEAN function uses: Excel rounds the total number excluded down to an even count, which is the same as rounding the per-end count down. So =TRIMMEAN(data, 0.20) on the worked example gives 26.75, exactly what this calculator reports. The practical consequence is the floor trap: on small datasets a small percentage trims nothing. 10% of 7 values is 0.7 values in total, which floors to 0 per end, so nothing is removed and the trimmed mean equals the original mean. That is correct behavior, not a bug, and the calculator reports “None” for values removed so you can see it happened.
Worked example: ten values, 20% trim, one big outlier
The calculator is pre-filled with 12, 15, 18, 22, 25, 28, 31, 35, 40, 95 and a 20% trim. Sorting changes nothing here, since the values are already in order. The trim count is floor(20 ÷ 100 × 10 ÷ 2) = floor(1) = 1, so one value leaves each end: 12 from the bottom and 95 from the top. The eight remaining values sum to 214, and 214 ÷ 8 = 26.75, reported as 26.750000. The original mean is 321 ÷ 10 = 32.1, reported as 32.100000. The gap between 32.100000 and 26.750000 is the whole point of the statistic: the outlier 95 was worth more than 5 points of average all by itself.
Every row below is a real run of this calculator. The second row is the one to study if you are coming from another trimmed-mean tool: a 10% trim on 20 values removes just 1 value from each end here, because 10% is the total, not the per-end share.
| Dataset and trim | Original mean | Trimmed mean | Values removed |
|---|---|---|---|
| 12, 15, 18, 22, 25, 28, 31, 35, 40, 95 at 20% | 32.100000 | 26.750000 | 12, 95 |
| 5 to 59 in steps of 3, plus 200, at 10% (20 values) | 40.400000 | 33.500000 | 5, 200 |
| 5, 7, 9, 10, 15, 20, 100 at 10% (7 values) | 23.714286 | 23.714286 | None (floor trap) |
| Worked example at 0% | 32.100000 | 32.100000 | None |
| Worked example at 50% | 32.100000 | 26.500000 | 12, 15, 40, 95 |
| −4, 2, 3, 5, 6, 7, 8, 9, 11, 50 at 20% | 9.700000 | 6.375000 | −4, 50 |
Trimmed mean vs plain mean vs median: which one answers your question
The three statistics sit on a spectrum from “use everything” to “use the middle only”, and picking the wrong one is the most common trimmed-mean mistake. The plain mean uses every value, so it is the most efficient summary when the data is clean, and the most misleading one when it is not. The median uses only the middle value (or the middle two), so no outlier can move it at all, but it also throws away almost everything you measured. The trimmed mean sits between: it keeps most of the data, discards only the tails, and stays stable when a few values go wild.
That middle position is why the trimmed mean shows up in judging panels, where the highest and lowest scores are dropped, and in Olympic-style scoring, and why it is the right tool for salary data with a few executive pay packets, or sensor readings with occasional spikes. It is not a substitute for looking at your data: if trimming changes the answer a lot, that gap is itself information, and the original mean beside the trimmed mean on this page exists so you can see it. For the plain mean, median and mode together, the mean median mode calculator covers all three at once. For a mean over grouped or weighted values, the mean of distribution calculator fits better, and the sample mean calculator handles the plain sample average on its own.
Input traps: commas, spaces, and numbers the tool reads loosely
Two input habits cause silent wrong answers, and both come from how the tool parses text. First, values must be separated by commas. Typing “1 2 3” with spaces only does not give three values: the tool splits on commas, finds one entry, reads it as 1, and quietly drops the rest. Spaces around a comma are fine (“1, 2, 3” works), but spaces are not separators. Second, the tool reads each entry with the same loose number parsing browsers use, so “12abc” reads as 12 without complaint, while a trailing comma creates an empty entry and triggers the invalid-number error. The safe habit is to paste clean comma-separated numbers and check the remaining-values output: if it shows fewer values than you typed, something in your input was not parsed the way you expected.
One more boundary to know: the trim percentage caps at 50, and even at 50 the dataset can never empty out, because the per-end count is at most a quarter of the values. A 50% trim on the worked example removes 2 from each end and averages the middle six, giving 26.500000. The tool also guards the empty case in code, but with the 0 to 50 validation in place that guard never fires.
What competing calculators miss
Most rival trimmed-mean tools compute the same average, so the differences are in the trimming convention, which many of them never state:
- Per-tail tools that never say so: several popular trimmed-mean pages describe a 10% trim as removing 10% from each end, so 10% on 20 values drops 2 from the bottom and 2 from the top, four values in total. That is a legitimate convention, but it is not this page’s convention and not Excel’s, and a user moving between the two gets a different answer from the same input with no warning. This page states its total-share convention in the calculator note, the formula section and the FAQ, so the difference is visible before you calculate.
- Tools that round the trim count instead of flooring it: at least one rival rounds 20% of 6 values (1.2) down to 1 per end, which on this page removes nothing, since floor(0.6) is 0. Rounding up would break the symmetry the statistic depends on; flooring keeps both tails equal, and this page reports “None” when the floor is zero instead of pretending a trim happened.
- Tools that show only the trimmed mean: a single number with no original mean, no removed list and no remaining list leaves you unable to check whether the trim did what you expected. This page shows all four outputs, so the floor trap and any parsing surprise are visible in the result itself.
- Spreadsheet guidance without a calculator: the Excel TRIMMEAN documentation is the authoritative statement of the convention this page follows, and it is genuinely useful, but it is not a tool: you still need a spreadsheet open. This page gives the same convention with no spreadsheet.
- The gap this page fills: the Excel-matching total-share convention stated in plain language, the original mean beside the trimmed mean, the removed and remaining lists for checking, and honest handling of the small-dataset floor case.
Frequently asked questions
A trimmed mean is an average calculated after removing a specified percentage of the smallest and largest values from a dataset, then averaging what remains. It is more robust than the plain mean because extreme values and outliers cannot drag it around, while still using most of the data, unlike the median.
The percentage is the total share of values removed, split equally between the low end and the high end. The count removed from each end is the trim percentage divided by 100, times the number of values, divided by 2, rounded down. So 20% on 10 values removes 1 from each end, and 10% on 20 values removes 1 from each end, not 2. This matches Excel’s TRIMMEAN function. If a tool you used before removed the percentage from each end separately, enter double the percentage here.
Because the per-end count is rounded down, and on a small dataset a small percentage rounds to zero. 10% of 7 values is 0.7 values in total, which gives 0 per end, so nothing is removed and the trimmed mean equals the original mean. The calculator reports None for values removed so you can see it happened. Raise the percentage or add more values if you want an actual trim.
The plain mean averages every value, including outliers, while the trimmed mean averages only the middle of the sorted data after dropping both tails. On the worked example, the plain mean is 32.100000 and the 20% trimmed mean is 26.750000, because the outlier 95 is dropped. For the plain mean, median and mode together, use the mean median mode calculator.
Common choices are 5%, 10% or 20%, and the right one depends on how dirty your data is. A light trim like 5% or 10% tames occasional outliers without discarding much data, while 20% suits data with frequent extremes. At 0% nothing is removed and you get the plain mean, and the calculator caps the trim at 50%.
Yes for decimals and negatives, and partly for spaces. Values must be separated by commas, but spaces around a value are fine, so 1, 2, 3 works. Spaces alone do not separate values: typing 1 2 3 is read as a single value of 1, and the rest is silently dropped. Always use commas between values.
Blank data gives Please enter data values. A trim percentage outside 0 to 50 gives Please enter a valid trim percentage between 0 and 50. A single value gives Please enter at least 2 data values. A non-numeric entry gives Error: Invalid number, followed by the offending entry. A trailing comma creates an empty entry and triggers the invalid-number error, so check for a comma at the end of your list.
Yes. It runs entirely in your browser, needs no account, and stores nothing. Copy the result with “Copy my results” or send it to another app with “Share results”.
Add This Calculator to Your Website
Embed this calculator on your own site with the code below. Pick a size to preview it live, then copy the code.
Powered by Calculator Zilo
References
- Microsoft Learn, Excel WorksheetFunction. TrimMean method (mean of the interior of a dataset after excluding a percentage from the top and bottom tails, excluded count rounded down to an even multiple). Accessed October 24, 2026.
- National Institute of Standards and Technology, DATAPLOT Reference Manual. Trimmed Mean Standard Error (trimmed mean as compensation for tail influence, 50% trimmed mean as the interquartile mean, 90% trimmed mean as 5% truncation per tail). Accessed October 24, 2026.
- National Institute of Standards and Technology. Trimmed Weighted Means (Yen, 2008, Journal of the American Statistical Association: trimming weight tails for robustness while retaining precision). Accessed October 24, 2026.
- Scilab Documentation. trimmean: Trimmed mean of a vector or a matrix (discarding a percentage of the lowest and highest scores before averaging, median as the 100% trimmed mean, arithmetic mean as the 0% trimmed mean). Accessed October 24, 2026.
- ArXiv. Improving Sensitivity in A/B Tests: Integrating CUPED with Trimmed Mean Techniques (sample trimmed mean definition as the mean of the ordered sample after removing g values from each end). Accessed October 24, 2026.