What Is Rounding Error? Definition, Formula, Examples, and How to Calculate It

A number can change by only 0.004 when it is rounded, yet that small difference can matter when calculations are repeated many times. For example, rounding 3.456 to 3.46 creates a difference of 0.004. That difference is called rounding error, and understanding it helps you know how much accuracy is lost when an exact value is replaced with an approximation.

• Original value: 3.456
• Rounded value: 3.46
• Signed difference: 0.004
• Absolute rounding error: 0.004

Rounding error is the mathematical difference between an exact value and its rounded, approximate value. Rounding replaces an exact value with an approximate value, which may differ from the original. When evaluating this error, the signed difference shows the direction of the change, while the absolute rounding error expresses the magnitude of the difference as a positive number. If the original number is already exactly at the target rounding increment, the error can be zero (for example, rounding 7.0 to 7).

To find out exactly how much the number changed, you calculate the signed difference: Signed Difference = Rounded Value − Original Value

Often, we only care about the size of the error, regardless of whether the rounded number is larger or smaller than the original. In these cases, we use absolute rounding error: Absolute Rounding Error = |Rounded Value − Original Value|

The absolute error is expressed as a positive quantity because it represents the pure distance between the exact number and the rounded number on a number line.

To quantify rounding error accurately, mathematicians use a specific set of formulas.

1. Signed Difference

• Formula: Signed Difference = Rounded Value − Original Value
• Explanation: This tells you the exact mathematical shift. A positive result means the number was rounded up; a negative result means the number was rounded down.

2. Absolute Error

• Formula: Absolute Error = |Rounded Value − Original Value|
• Explanation: This strips away the positive/negative sign, giving you the total magnitude of the error.

3. Relative Error

• Formula: Relative Error = |Rounded Value − Original Value| / |Original Value|
• Explanation: This compares the size of the absolute error to the size of the original number. It shows how significant the error is relative to the starting value.

4. Percentage Error

• Formula: Percentage Error = Relative Error × 100
• Explanation: This simply converts the relative error into a more readable percentage format.

Worked Example

Imagine an original value of 150.7 that is rounded to the nearest whole number, giving 151.

• Signed Difference: 151 - 150.7 = 0.3
• Absolute Error: 0.3
• Relative Error: \approx 0.00199
• Percentage Error: \approx 0.199%

The same absolute error can represent very different relative errors depending on the size of the original value.

Example 1: Small Original Number

• Original: 1.2
• Rounded: 1.0
• Absolute Error: 0.2
• Relative Error: 0.2 / 1.2 \approx 0.1667
• Percentage Error: \approx 16.67%

Example 2: Large Original Number

• Original: 1,000.2
• Rounded: 1,000.0
• Absolute Error: 0.2
• Relative Error: 0.2 / 1000.2 \approx 0.0002
• Percentage Error: \approx 0.02%

Here are calculations covering various rounding scenarios using standard Round Half Up rules:

1. Decimal Rounding (Nearest Tenth)

• Original number: 4.23
• Target: Nearest tenth
• Rounded result: 4.2
• Signed difference: 4.2 - 4.23 = -0.03
• Absolute error: 0.03

2. Whole-Number Rounding

• Original number: 18.6
• Target: Nearest whole number
• Rounded result: 19
• Signed difference: 19 - 18.6 = 0.4
• Absolute error: 0.4

3. Nearest Ten

• Original number: 344
• Target: Nearest ten
• Rounded result: 340
• Signed difference: 340 - 344 = -4
• Absolute error: 4

4. Nearest Hundred

• Original number: 8,475
• Target: Nearest hundred
• Rounded result: 8,500
• Signed difference: 8,500 - 8,475 = 25
• Absolute error: 25

5. Very Small Decimals

• Original number: 0.0076
• Target: Nearest thousandth
• Rounded result: 0.008
• Signed difference: 0.008 - 0.0076 = 0.0004
• Absolute error: 0.0004

6. Currency Values

• Original number: $14.994
• Target: Nearest cent (hundredth)
• Rounded result: $14.99
• Signed difference: 14.99 - 14.994 = -0.004
• Absolute error: 0.004

When determining the signed error, the math relies on the relationship between the rounded value and the original value. Because the formula is Rounded − Original, the sign corresponds directly to the direction of movement on the number line.

• Positive signed difference: Occurs when you round to a greater value. For example, rounding 4.6 to 5 gives a signed difference of 5 - 4.6 = 0.4.
• Negative signed difference: Occurs when you round to a smaller value. For example, rounding 4.3 to 4 gives a signed difference of 4 - 4.3 = -0.3.
• Zero difference: Occurs when the original number is already at the target increment. Rounding 7.0 to the nearest whole number results in 7, giving a difference of 7 - 7 = 0.

Remember, while the signed difference can be positive or negative, the absolute error is always evaluated as a positive magnitude.

Calculating rounding error with negative numbers uses the exact same formulas, but it is important to pay attention to mathematical signs.

Let’s say we round -2.7 to the nearest whole number.

• Original number: -2.7
• Rounded result: -3 (because -3 is the nearest whole number)

Calculate the signed difference: Rounded Value − Original Value $-3 - (-2.7) = -3 + 2.7 = -0.3

The signed difference is -0.3$. Even though the magnitude (absolute value) of the number increased from 2.7 to 3 (moving away from zero), on a mathematical number line, the rounded result is a smaller value. The value moved downward (to the left) from -2.7 to -3.

The absolute error is |-0.3| = 0.3.

The precision of your rounding target directly dictates the possible size of your error. Finer rounding (retaining more decimal places) inherently restricts the maximum possible absolute rounding error.

Consider the original number 5.4444:

• Nearest whole number: Rounds to 5. Absolute error = |5 - 5.4444| = 0.4444
• Nearest tenth (1 decimal place): Rounds to 5.4. Absolute error = |5.4 - 5.4444| = 0.0444
• Nearest hundredth (2 decimal places): Rounds to 5.44. Absolute error = |5.44 - 5.4444| = 0.0044
• Nearest thousandth (3 decimal places): Rounds to 5.444. Absolute error = |5.444 - 5.4444| = 0.0004

As you can see, rounding to finer increments reduces the maximum possible absolute error for nearest rounding.

For nearest-value rounding, the absolute error is at most half of the rounding increment. The exact result at the midpoint depends on the tie-breaking method.

• Nearest whole number: maximum magnitude is 0.5
• Nearest tenth: maximum magnitude is 0.05
• Nearest hundredth: maximum magnitude is 0.005

Half Up and Half Even can differ in how they handle exact midpoint values, but the distance from the midpoint to either neighboring target is still half the increment. Ceiling and Floor are directional methods and do not follow the same half-increment error bound; their maximum absolute error approaches the full increment size.

Rounding a single number creates an isolated error. However, if you round multiple intermediate numbers before adding them together, the individual errors can compound, and rounding intermediate values can change the final result.

Example: Imagine adding three numbers: 1.4 + 2.4 + 3.4

• Exact total = 7.2
• Rounded exact total to nearest whole number = 7

If the individual numbers are rounded first:

• 1.4 \rightarrow 1
• 2.4 \rightarrow 2
• 3.4 \rightarrow 3 Then: 1 + 2 + 3 = 6

By rounding the intermediate numbers prematurely, we pushed all three values downward. The accumulating negative signed differences (-0.4 each) resulted in a final answer (6) that is the result of the prematurely rounded calculation, rather than the true rounded total (7).

In finance and accounting, managing precision is important because rounding errors in prices, tax rates, or interest calculations can accumulate over time.

If a store calculates a $7.5% sales tax on a 10.25 item, the exact tax is 0.76875. Because currency generally uses two decimal places, this might be rounded to 0.77. Over thousands of transactions, these fractions of a cent can add up.

Different financial systems can define when and how values are rounded based on specific regulations or business rules. For example, a system may require line items to be rounded before being summed, while another system might maintain full precision until the final invoice total. Using the wrong method can cause unexpected discrepancies in accounting totals.

In science and engineering, numbers often represent physical measurements. When calculations are performed on these measurements, significant figures are used to ensure that the final output doesn’t falsely imply more precision than the physical instruments provided.

It is important to clearly separate rounding error from significant figures and measurement uncertainty.

• Measurement uncertainty is the physical limitation of the measuring tool.
• Significant figures are a set of rules used to track that uncertainty during calculations.
• Rounding error is the purely mathematical difference created when you discard digits (often to comply with significant figures rules). They are distinct concepts that work together to maintain realistic precision.

Rounding and truncation both shorten numbers, but they work differently and produce different errors. Rounding evaluates the dropped digits to seek the nearest value. Truncation ignores the dropped digits entirely, simply chopping off the number.

Truncation toward zero behaves differently for negative numbers. For example, truncating -2.8 to a whole number yields -2. Meanwhile, nearest rounding yields -3.

To minimize the negative impacts of rounding error, consider these practical recommendations:

• Keep sufficient precision: Maintain unrounded, high-precision numbers during all intermediate calculations.
• Round only when necessary: Apply rounding exclusively at the final step of the calculation or when generating the final report.
• Avoid double rounding: Never round a value that has already been rounded (e.g., do not round 3.449 to 3.45, and then to 3.5).
• Use appropriate methods: If statistical bias is a concern, use Round Half to Even (Banker’s Rounding) instead of standard Half Up.
• Follow standards: Follow the precision required for the specific engineering, scientific, or financial application.
• Verify tools: Be aware of the default rounding behavior of your programming language or spreadsheet software, as they do not all behave identically.

You can easily compute exact rounding errors using our interactive calculator. Simply enter your original number, choose your target precision (such as nearest tenth or nearest whole number), and select the rounding method. The calculator will instantly display the rounded result alongside the absolute, relative, and percentage errors.

Here is an example of the step-by-step calculation it performs for an exact value of 6.786 rounded to 7:

Step 1. Identify Exact and Rounded Values

• Exact Value = 6.786
• Rounded Value = 7

Step 2. Calculate Absolute Error

• Absolute Error = |Exact − Rounded|
• |6.786 - 7| = 0.214

Step 3. Calculate Relative Error

• Relative Error = Absolute Error / |Exact|
• 0.214 / |6.786| \approx 0.03154

Step 4. Calculate Percentage Error

• Percentage Error = Relative Error × 100
• 0.03154 \times 100 \approx 3.154%

You can also use the main Rounding Calculator to explore how different methods change a number’s final value.

When analyzing rounding errors, avoid these frequent pitfalls:

• Confusing signed difference with absolute error: The signed difference can be negative, but absolute error is always a positive magnitude.
• Rounding intermediate results too early: This causes small errors to compound into a drastically inaccurate final result.
• Ignoring negative numbers: The mathematical sign matters when calculating the signed difference (Rounded − Original).
• Confusing rounding error with measurement uncertainty: Measurement error comes from physical tools; rounding error is purely mathematical.
• Assuming all methods produce identical error bounds: While nearest-value rounding restricts error to half the increment, Floor and Ceiling methods do not.
• Using insufficient precision: Not keeping enough decimal places during computations can destroy the accuracy of the result.
• Using the wrong rounded value: Misapplying the rounding rules to select the wrong increment.