The decision starts with the relationship between the values
Use percentage change when one value is a genuine original or reference baseline and the other is a later, revised, or compared value. Its signed formula is (new − original) ÷ original × 100. Use percentage difference when the values are peer measurements and neither deserves baseline status. A common formula is |A − B| ÷ ((A + B) ÷ 2) × 100 for positive quantities.
The formulas are not competing ways to calculate the same statistic. They encode different questions. Change asks how far a value moved relative to its starting point and preserves direction. Difference asks how far apart two peers are relative to their average and is symmetric. Before reaching for a calculator, write one sentence: “A is the baseline for B,” or “A and B are equal-status measurements.” That decision determines the denominator.
Directional change for a time series
Input: annual electricity use moves from an original 4,000 kWh to a new 4,600 kWh. Formula: percentage change = (new − original) ÷ original × 100. Calculation: (4,600 − 4,000) ÷ 4,000 × 100 = 600 ÷ 4,000 × 100 = 15%. Result: use increased by 15%. Interpretation: the 600 kWh rise equals 15% of the earlier year's consumption.
Reversing the years gives (4,000 − 4,600) ÷ 4,600 × 100 ≈ −13.04%, not −15%. That is expected because the baseline changes from 4,000 to 4,600. Time order supplies a defensible original value, and the sign communicates increase or decrease. State the period and verify that definitions and measurement coverage are comparable across both observations.
Symmetric difference for peer measurements
Input: two laboratories measure the same sample as 100 mg and 110 mg, with neither designated as the reference method. Formula: percentage difference = |A − B| ÷ ((A + B) ÷ 2) × 100. Calculation: |100 − 110| ÷ 105 × 100 = 10 ÷ 105 × 100 ≈ 9.52%. Result: the readings differ by approximately 9.52% relative to their mean.
Interpretation: swapping A and B leaves both the absolute gap and average unchanged, so the result remains 9.52%. That symmetry is the point. It does not identify which reading is correct, establish measurement uncertainty, or replace a laboratory's required relative-error or method-comparison procedure. If a certified reference value exists, the reference belongs in a baseline-based calculation instead of the peer-difference formula.
Same numbers, different questions
Input: compare 80 and 100. Formula: use change = (new − original) ÷ original × 100 for an ordered pair, or difference = |A − B| ÷ their average × 100 for peers. Calculation: from 80 to 100, change is 20 ÷ 80 × 100 = 25%; from 100 to 80, it is −20 ÷ 100 × 100 = −20%; as peers, difference is 20 ÷ 90 × 100 ≈ 22.22%. Result: the three measures are 25%, −20%, and 22.22% respectively.
Interpretation: all three answers are arithmetically correct because their questions differ. The 25% and −20% results have direction and depend on which endpoint is original. The 22.22% result is unsigned and unchanged when the peers are reordered. Reporting only “the percentage is 22%” hides the denominator choice. Name the measure, values, and baseline relationship so another reader can reproduce it.
Zero values require explicit handling
For percentage change, an original value of zero makes the standard formula undefined. A move from 0 to 12 has an absolute increase of 12 but no finite conventional percentage change. If the new value is zero and the original is positive, the change is −100%. Those cases are not interchangeable: only the latter has a nonzero denominator.
For percentage difference with nonnegative peers, one zero and one positive value produce 200% under the average-denominator formula: |0 − 12| ÷ 6 × 100 = 200%. Two zero values create 0 ÷ 0, which is undefined even though the raw difference is zero. Some fields define a special result for reporting; use it only when the applicable standard says so and disclose the convention.
Opposite signs can make the average denominator fail
The usual average-based percentage-difference formula is best suited to positive quantities on a ratio scale. If A and B have opposite signs, their arithmetic mean can be zero or close to zero. For −10 and 10, |−10 − 10| = 20 while the mean is 0, so the percentage difference is undefined. A denominator near zero can create an enormous value that obscures the actual comparison.
Signed temperatures on Celsius or Fahrenheit scales, net profit and loss, and deviations around zero often need absolute difference, a stated reference magnitude, or a domain-specific normalized metric. Do not repair the formula by silently taking absolute values in the denominator; that creates a different measure. Explain the scale, signs, and selected comparison method.
Difference, change, and percentage points
When both inputs are already percentages, direct subtraction answers a percentage-point question. A survey share moving from 40% to 46% changes by 6 percentage points; relative change is (46 − 40) ÷ 40 × 100 = 15%. Applying the peer percentage-difference formula would answer yet another question. Choose language that matches the intended denominator rather than using “percent difference” as a catch-all phrase.
In experiments, the governing method may specify relative error against a reference, relative standard deviation across repeated observations, or another statistic. In time series, revisions to population, price, or sampling definitions can make a mathematically correct change incomparable. Formula selection is necessary, but comparable inputs and an honest description of the baseline are equally important.
Using the calculator and checking the result
The Percentage Calculator directly evaluates the average-based percentage difference between two peer values. It does not directly calculate original-to-new percentage change; use the change formula above for that directional question. Confirm the selected mode and denominator shown by the tool. It also cannot decide whether observations have equal status, whether a laboratory standard requires another metric, or whether a time-series definition changed; those are contextual decisions.
Common errors are using absolute value for a directional change, treating an arbitrary first-listed peer as the baseline, dividing difference by only one peer, using the average formula across opposite signs, and calling percentage points “percent difference.” As a check, swap the values: percentage difference should remain the same, while percentage change should generally change direction and magnitude because the baseline changes.
Document the denominator before publishing
A reproducible comparison names the formula and why its denominator is appropriate. For a time series, identify the original period and any revisions. For peer measurements, say that neither observation is the reference and that their arithmetic mean is used. Include units and unrounded inputs, then state the rounding applied to the result. This short record prevents a later editor from treating change and difference as interchangeable labels.
Be cautious when software offers a generic “percent difference” field. Some products use that label for change from a baseline, relative error, absolute percentage error, or the symmetric average-based formula. Inspect the documented equation rather than trusting the button name. If a regulated, scientific, or contractual method defines its own denominator, that definition takes priority over a general educational convention.
Rounding should come after the denominator has been computed from the original observations. In the laboratory example, using the exact mean of 105 gives 9.5238%, which can be reported as 9.52%. Rounding each input or the mean prematurely can shift the last reported digits. Keep enough precision to reproduce the result and label approximations rather than implying that a rounded percentage is exact.