Excel rounds numbers with the ROUND, ROUNDUP, ROUNDDOWN, MROUND, INT, and TRUNC functions, each following a specific tie-breaking and direction rule that you choose at the formula level. The most common formula is =ROUND(number, num_digits), where num_digits controls precision: a positive value counts decimal places, zero rounds to the nearest whole number, and a negative value rounds to the left of the decimal point (tens, hundreds, thousands). For example, =ROUND(2.5, 0) returns 3 because Excel uses half-away-from-zero, and =ROUND(-2.5, 0) returns -3 by the same rule. ROUNDUP and ROUNDDOWN ignore whether the discarded portion is more or less than half; they always move in the named direction, so =ROUNDUP(-1.21, 0) returns -2 while =ROUNDDOWN(-1.21, 0) returns -1. MROUND rounds to a chosen step (such as the nearest 0.05), INT returns the integer floor toward negative infinity, and TRUNC drops the decimal portion without rounding. Picking the wrong one is the usual source of off-by-one errors in financial and measurement worksheets.
Excel's Core Rounding Functions at a Glance
Excel ships with a small family of rounding functions. They differ in three ways: which digits they discard, how they handle an exact halfway case, and which direction they move when the discarded portion is nonzero. The table below summarizes what each function does for typical positive and negative inputs at zero decimal places.
| Function | Behavior | =func(2.5, 0) | =func(-2.5, 0) | =func(-1.21, 0) |
|---|---|---|---|---|
| ROUND | Half away from zero | 3 | -3 | -1 |
| ROUNDUP | Always away from zero | 3 | -3 | -2 |
| ROUNDDOWN | Always toward zero | 2 | -2 | -1 |
| MROUND | Nearest multiple of step | 3 (step 1) | -3 (step 1) | -1 (step 1) |
| INT | Floor toward −∞ | 2 | -3 | -2 |
| TRUNC | Drop fractional digits | 2 | -2 | -1 |
Note the contrast between ROUNDDOWN and INT in the column for -1.21: ROUNDDOWN returns -1 because it always moves toward zero, while INT returns -2 because it always moves toward negative infinity. The same split appears between TRUNC and INT on any negative value with a nonzero discarded portion. When a workbook must match a published policy, the difference between "drop digits" and "floor" usually decides the answer.
How to Round Numbers in Excel
- Type the source value in a cell, for example 123.4567 in cell A1.
- Click the cell where the rounded result should appear, then type =ROUND(A1, 2) to round to two decimal places, or replace 2 with any other count of decimal places you need.
- Use =ROUNDUP(A1, 2) to force the result away from zero, or =ROUNDDOWN(A1, 2) to force it toward zero, regardless of how close the discarded digits are to halfway.
- Use a negative second argument to round to the left of the decimal point: =ROUND(A1, -2) rounds 123.4567 to 100, and =ROUND(A1, -1) rounds it to 120.
- For cash-style rounding in fixed increments, use =MROUND(A1, 0.05) to round to the nearest nickel; combine with =CEILING(A1, 0.05) or =FLOOR(A1, 0.05) when you need ceiling or floor toward a step.
- Press Enter to confirm, then copy the formula down the column. If the result looks wrong, double-check whether you wanted nearest, ceiling, floor, or truncation, because Excel does not label which mode the formula picked.
The 1.005 Problem: Why Excel Sometimes Disagrees with Itself
The classic Excel rounding surprise is =ROUND(1.005, 2) returning 1.00 instead of the expected 1.01. The rounding rule is not at fault; the value stored in the cell is. Excel stores numbers as 64-bit binary floating-point values, and the decimal string 1.005 cannot be represented exactly in binary, so the cell actually holds something like 1.00499999999999989. By the time Excel applies ROUND, the half is already gone, and the function correctly rounds the smaller approximation down.
The same trap fires on other seemingly innocent values: =ROUND(0.145, 2) can return 0.14, and =ROUND(2.675, 2) can return 2.67. Each time, the displayed number is not the number Excel is actually rounding. Workarounds include the formula =ROUND(A1 + 1E-10, 2) to nudge values just above the halfway mark, or switching to ROUND applied to the text form =ROUND(VALUE(TEXT(A1,"0.00")), 2). Neither workaround is guaranteed across every input, which is why an independent check on the original decimal is the safer habit when the answer must audit cleanly.
Five Rounding Modes and Which One You Actually Need
Excel exposes nearest, away-from-zero, toward-zero, and floor behavior through its built-ins, but it does not name them or document their tie rules in one place. Outside Excel, the same decimal can round five different ways depending on the policy of the software reading it. The table below lists the modes and their behavior, using the definitions published by the Unicode Consortium and the ICU library, which treat the discarded digits as text rather than as a binary approximation.
| Mode | Direction | Tie behavior | 1.235 → 2 places | -1.235 → 2 places | -1.21 → 0 places |
|---|---|---|---|---|---|
| Half up | Nearest | Away from zero | 1.24 | -1.24 | -1 |
| Half even (banker's) | Nearest | Last retained digit is even | 1.24 | -1.24 | -1 |
| Floor | Toward −∞ | Not applicable | 1.23 | -1.24 | -2 |
| Ceiling | Toward +∞ | Not applicable | 1.24 | -1.23 | -1 |
| Truncate | Toward zero | Not applicable | 1.23 | -1.23 | -1 |
Half even is often called banker's rounding because it removes the persistent upward bias that half up produces when many independently rounded ties are summed. For example, with half up the inputs 2.5, 3.5, and 4.5 round to 3, 4, and 5; adding those rounded values gives 3 + 4 + 5 = 12. With half even the same inputs round to 2, 4, and 4; adding those gives 2 + 4 + 4 = 10, a half-step smaller overall. Excel's ROUND behaves like half up, not half even, so a workbook that needs half even has to add a custom formula or call an external function. When a finance, tax, or reporting rule specifies "round half to even" or "round half away from zero," the choice changes the answer on every exact tie, and there is no Excel function that reads the rule off the formula bar.
Checking an Excel Result Against an Exact Calculator
Once you know which mode a workbook should follow, you can verify any rounded cell by running the original decimal through a tool that performs the rounding directly on the digit string. The Rounding Calculator accepts a plain signed decimal, lets you pick zero through twenty decimal places, and offers all five modes from the table above. Because it inspects the discarded digits as text instead of converting the value through binary floating point first, it agrees with the spelled-out rule on cases where Excel often does not.
Concretely, take the input 1.235 at two decimal places. The retained portion before the cut is 1.23; the discarded digit is 5, an exact tie. With half up selected, the calculator inspects the tie and increments the last retained digit, giving 1.24. Switching the mode to half even gives 1.24 as well, because 4 is already even. Switching to floor gives 1.23, ceiling gives 1.24, and truncate gives 1.23. The same input on -1.235 shows the modes diverging on negatives: half up and half even return -1.24, ceiling returns -1.23, floor returns -1.24, and truncate returns -1.23. The published Unicode and ICU rounding-mode definitions agree on each of these directions, so the calculator's behavior matches what an audited spreadsheet policy expects.
Rounding Best Practices for Spreadsheets
Three habits prevent most rounding disputes before they start. First, keep full precision during intermediate calculations and round only at the stage required by your method; rounding a running total on every row drifts further from a final answer rounded once. Second, paste decimals into the calculator verbatim, without thousand separators or scientific notation, so the digits the tool sees are the digits you intend; the input box rejects commas and exponents on purpose, since 1,000.5 and 1e3 would otherwise be silently reinterpreted. Third, when a workbook disagrees with another application, check the mode and the input representation on each side of a tie: a different stored value, a different tie rule, or a different stage where rounding occurs can each explain a one-unit gap, and only an exact comparison pinpoints which one.
If a result needs to display trailing zeroes to communicate precision (so 1000.00 reads differently from 1000), format the cell with a number format such as 0.00 after rounding, or use the calculator's output, which retains the requested precision verbatim. Excel's display format only changes appearance; the underlying cell still holds whatever value ROUND produced. For external reporting, that distinction is the difference between a number that audits cleanly and one that fails a precision check.
Related reading: How to Convert Time to Decimal in Excel: Faster Method.
Related reading: Calculate Time Duration in an Excel Sheet.