PERCENTILE.EXC in Excel: Exclusive Percentiles and Small-Sample Limits

• Introduced: the Microsoft function catalogue marks PERCENTILE.EXC as an Excel 2010 addition; an exact first build or preview date is not established here.
• 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, Excel Web App.
• Checked: October 14, 2026; official PERCENTILE.EXC reference. Documentation and independently modeled examples; no native Excel execution or broader platform certification.

=PERCENTILE.EXC(array, k)

Both arguments are required. For the numeric controls below, the one-based sorted position is (n+1)*k. An integer selects an observation; a fractional position between observations is interpolated.

The minimum and maximum occupy positions 1/(n+1) and n/(n+1) on this probability scale. They are not assigned percentile probabilities zero and one. This is why an exclusive percentile is not obtained merely by deleting the first and last observations.

Enter 10, 20, 40, 80 and 160 in A2:A6.

=PERCENTILE.EXC(A2:A6,0.25)
=PERCENTILE.EXC(A2:A6,0.75)

Expected results are 15 and 120. With five observations, the positions are 6*0.25=1.5 and 6*0.75=4.5. These are halfway between 10 and 20, and halfway between 80 and 160.

The median request =PERCENTILE.EXC(A2:A6,0.5) returns 40, at position 3. For comparison, PERCENTILE.INC gives quarter points 20 and 80 on the same data. Neither convention changes the source records.

=PERCENTILE.EXC(A2:A6,0.1)
=PERCENTILE.EXC(A2:A6,0.9)

Both requests have expected documented error #NUM!: their positions are 0.6 and 5.4, outside the five observed positions. There is no bracketing pair for the interpolation.

Microsoft’s illustrated reference similarly rejects a 1st percentile for the nine-value dataset 1 through 9. Do not replace this limitation with an invented extrapolated value below the minimum.

Enter the integers 1 through 19 in C2:C20. The count is 19, so the plotting-position multiplier is 20.

The successful 10th-percentile request now has an observed position. Adding observations is not a license to invent records merely to make a formula succeed; it changes the dataset and must reflect real collected data.

For A2:A6, place the exclusive upper quartile in E2 and the inclusive upper quartile in E3:

=PERCENTILE.EXC(A2:A6,0.75)
=PERCENTILE.INC(A2:A6,0.75)

The outputs are 120 and 80. With a rule at or above the cutoff, COUNTIF returns 1 and 2 respectively. A report that swaps functions without updating its method can change which records are selected.

Keep the original numeric output and the method label in separate cells. Broad error suppression with a zero fallback would turn an unsupported percentile into a seemingly valid zero threshold.

Microsoft documents #NUM! for an empty array, k<=0, k>=1, or a requested percentile that cannot be interpolated; a nonnumeric k gives #VALUE!. These are documentation-based expectations, not executed Excel error captures.

The examples avoid exact floating-point boundary comparisons at 1/(n+1) and n/(n+1), mixed cell types and single-observation cases. The algebraic positions explain the convention without claiming every numerical edge is native-tested.

See QUARTILE.EXC for the quarter-point interface and MEDIAN when the middle value alone is needed.

• Microsoft PERCENTILE.EXC reference.
• Editorial testing standards.

Expected results are independent statistical/arithmetic models or explicitly documented rules, not captured native output. Examples use English function names and comma separators. Datasets and reporting policies are fictional.