RAND Function (Gnumeric)
Generate changing uniform draws with Gnumeric RAND, using its zero-argument syntax and distinguishing samples from reproducible constants.
Verification status: See the evidence and test scope described in this guide. How formula examples are checked
RAND produces one pseudorandom number in the interval from zero inclusive to one exclusive. It is useful for simulation inputs, randomized trial assignments, and a temporary random ordering of synthetic records. It does not return a percentage automatically, and it does not accept a seed argument.
Syntax and recalculation
RAND()
There are no arguments. The 1.12.57 registration marks the function volatile: recalculation can replace its previous value. Two separate RAND calls request separate draws, even when written in the same expression. Do not rely on a value remaining unchanged while editing the worksheet.
For a reusable random input, put =RAND() in A2 and reference A2 from dependent formulas. This ensures those formulas use the same cell value within that calculation state. To keep a permanent scenario, save a values-only copy separately from the live formulas, then verify the saved values before using them in an audit or comparison.
Worked example: a bounded duration
Enter =RAND() in A2. In B2 enter =10+5*A2. A2 is at least zero and less than one; B2 is consequently at least 10 and less than 15. The units of B2 are whichever units you assign to the constants, such as minutes.
There is no single expected decimal value for A2. For illustration only, if a saved draw is 0.4, B2 is 12. That arithmetic example is not a promised native draw. The guide’s native checks include RAND itself, a range predicate, and the equivalent direct formula =10+5*RAND(); observed values are run-specific. One retained run returned 0.2972627345749217 for RAND and 12.678362600692523 for the separately drawn duration; these are observations, not reproducible constants.
For a cell-level range check, =AND(A2>=0,A2<1) checks the stored A2 draw. Writing two RAND calls inside the predicate instead checks two different draws. Either can test the basic bounds, but only the referenced-cell version checks one shared value.
Advanced use: compare scenarios fairly
For a small simulation, keep the draw in one column, the transformed input in another, and each model’s output in separate columns that reference the same transformed input. When comparing two policies, this prevents accidental comparison of different random scenarios simply because each output formula generated its own draw.
Summarize many observations only after deciding how many trials are needed and how results will be saved. A single successful bounds check is not a test of distribution quality, independence, or statistical convergence. RAND is a worksheet simulation function; do not use it to create passwords, authentication tokens, or security secrets.
Errors and troubleshooting
=RAND(1) supplies an unsupported argument and returns #N/A in the tested version. Putting a number between the parentheses does not establish a reproducible seed. If the result appears to be 1 after rounding or percentage formatting, inspect more decimal places before concluding that the underlying value equals 1.
If successive reported outputs differ, check whether recalculation occurred. Preserve both the formula and a frozen result when documenting a demonstration. A screenshot or exported workbook with one draw demonstrates that calculation state, not the next draw or a guaranteed random sequence.
Version and verification
This guide uses the official 1.12.57 random-function registration and generator implementation. The first introduction release and any final supported release are not established here. The selected examples were recalculated by the actual Gnumeric 1.12.57 engine under C.UTF-8 on 17 October 2026; no GUI screenshot or statistical certification is claimed.