PERCENTILE Function (Gnumeric)

A percentile is a data-dependent threshold. A 90th-percentile value of 86 minutes does not say that any particular delivery took 86 minutes; it means that 86 is the interpolated threshold for the 90% position according to the selected convention. Use it to investigate unusually slow deliveries, higher-than-typical invoices or customer-service response times. For skewed samples, calculate percentiles as a summary rather than treating them as promises about future observations.

Exact syntax and parameters

=PERCENTILE(array,k)

• array (required): Numeric sample or worksheet range, for example B2:B9. The official documentation calls it data points. For this guide, enter numeric cells only; the manual’s PERCENTILE entry does not specify every text, blank or error coercion case.
• k (required): Percentile as a proportion from 0 through 1, inclusive. 0 selects the minimum, 0.5 the median under this convention, and 1 the maximum. The value 90 is not the 90th percentile: it is outside the allowed interval. Enter 0.9 for 90%.

Both arguments are required; no default k is documented. Official Gnumeric says an empty dataset or k<0 / k>1 gives #NUM!. Gnumeric Statistics manual: PERCENTILE. Its public function catalogue marks PERCENTILE Complete / Basic, a project classification rather than proof this guide was natively tested. Gnumeric catalogue.

Create the following worksheet. The unit is minutes; keep the values numeric, not text with an m suffix.

On a narrow screen, scroll this table sideways to see every column.

Enter these formulas in D2:D8, outside the input range. Display one or two decimal places if you want consistent presentation.

On a narrow screen, scroll this table sideways to see every column.

*Expected outputs were calculated independently using exact decimal arithmetic for Hyndman–Fan method 7, the method named in Gnumeric’s official PERCENTILE definition; these are not captured application results.

Understand the interpolation rather than guessing

The sorted eight-item sample is 5, 10, 20, 25, 31, 40, 80, 100. For Gnumeric’s inclusive method, a percentile at proportion k has one-based index h = 1 + (n−1) × k. For 90%, h = 1 + 7 × 0.9 = 7.3, between the 7th (80) and 8th (100) observations. Interpolating yields 80 + 0.3 × (100−80) = 86. For 25%, h = 2.75, so 10 + 0.75 × (20−10) = 17.5. This is why selecting the nearest listed number can be wrong.

Make k user-selectable. Enter 0.9 as a numeric value in D10, then use =PERCENTILE($B$2:$B$9,D10) in E10. Expected output: 86. With D10=0.6, expected output: 32.8. Absolute range anchors preserve the sample if you copy the formula; they do not freeze the contents of those cells.

Flag slow-processing jobs. Enter the following in C2 and copy to C9:

=IF(B2>PERCENTILE($B$2:$B$9,0.9),"Above P90 threshold","At or below P90")

With the displayed sample, only Parcel H (100) is expected to read Above P90 threshold; Parcel G (80) is at or below the computed cutoff 86. This is a sample-specific result, not a promise that exactly 10% of a small set will be marked, particularly when there are ties.

Compare incompatible percentile definitions deliberately. Gnumeric documents PERCENTILE.EXC separately, using Hyndman–Fan method 6 (N+1), not method 7. For the same eight numbers, the expected =PERCENTILE.EXC(B2:B9,0.25) is 12.5, while inclusive PERCENTILE(...,0.25) is 17.5. Do not call the .EXC function an alias or silently substitute it. Official PERCENTILE and PERCENTILE.EXC definitions.

Locale, version and application differences

Examples reproduce the Gnumeric manual’s comma-separated argument notation and period decimal notation. Gnumeric supports localization of numeric entry/display; check your installation’s formula separators and decimal conventions instead of pasting a localized formula blindly. Official Gnumeric data types; localization. The checked documentation and release scope are 1.12.62; an older distribution package or plugin arrangement may differ. Gnumeric’s public catalogue says the implementation is complete, but that says nothing about mixed-input coercion or this guide’s validation.

The inclusive calculation corresponds to Excel’s PERCENTILE (legacy) and PERCENTILE.INC semantics for the numeric-only examples here. Microsoft documents interpolation and the 0..1 interval for PERCENTILE.INC; that is documentation-level compatibility, not a Gnumeric-to-Excel file round-trip test. Microsoft PERCENTILE.INC. Gnumeric’s PERCENTILE.EXC is separate and uses a different percentile convention.

QUARTILE · MEDIAN · MIN · MAX · LARGE · SMALL

• Official Gnumeric Statistics reference — PERCENTILE and PERCENTILE.EXC — definitions, arguments, errors, method classifications; accessed 2026-10-12.
• Official Gnumeric function catalogue — implementation/testing labels; accessed 2026-10-12.
• Official Gnumeric 1.12.62 release announcement — release context; accessed 2026-10-12.
• Microsoft Excel PERCENTILE.INC documentation — cross-compatibility comparison; accessed 2026-10-12.

The syntax and behaviour are based on the Gnumeric references cited above. The sample calculations were independently checked; they are expected results, not outputs observed in Gnumeric. Test locale-sensitive text conversion and imported files in your own installation. View the example-check record.

Official PERCENTILE reference · Editorial standards