A/B Testing Results: Which Percentage Formula Should You Use?

Percent change uses the formula (new − old) / old × 100. Use it when you have a clear baseline (e.g., control, or a before value). Percentage difference uses |A − B| / average(A, B) × 100. Use it when you have two parallel values with no established baseline.

The key question: Is there a natural baseline? If yes—e.g., a stock price last month vs. this month—percent change fits. If no—e.g., two thermometers reading different temperatures at the same moment—percentage difference fits. For the full breakdown of when to use each formula, see our guide on Percent Change vs. Percentage Difference. Here we focus on how that applies to A/B tests.

When you have a Control, that is your baseline—the status quo. The control represents what you have today; the treatment is what you are testing against it. In that setup, percent change (often called relative lift) is the correct formula: (treatment − control) / control × 100. Many widely used A/B testing platforms use percent change for exactly this reason—they treat the control as the baseline.

Percentage difference is only appropriate when you are comparing Variant A vs. Variant B—two brand-new ideas where neither is the established baseline. For example, testing two new headline designs against each other, with no "current" version. In that case, there is no natural "old" value, so you use the symmetric formula: |A − B| / average(A, B) × 100.

With Control vs. Treatment, the control is the baseline by definition. You are measuring "how much did the treatment improve over the status quo?" Percent change answers that directly: (2.4 − 2.0) / 2.0 = 20% lift. Swapping control and treatment would change the result—but you should not swap them, because the control is the established baseline. The asymmetry is correct: you care about lift from the current state, not a symmetric comparison.

For Variant A vs. Variant B, neither is the status quo. You are asking "how different are these two options?"—not "how much did B improve over A?" In that case, percentage difference is correct. Always state which formula you used so stakeholders understand the framing.

Use percent change when you have a baseline: Control vs. Treatment (control is the status quo), or pre/post tests (before is the baseline). Use percentage difference when you have Variant A vs. Variant B with no established baseline—two new ideas compared as peers. Our Percentage Change Calculator includes both modes so you can run the numbers and see the difference.

Many well-known A/B platforms report percent change—often labeled as "relative lift"—when a control is present, because they treat the control as the baseline. If your test has a control, you are aligned with that convention. If you are testing two new variants with no control, use percentage difference instead.

State which formula you used. For Control vs. Treatment: "Treatment showed 20% lift over control (percent change, control as baseline)." For Variant A vs. Variant B: "Variant B showed an 18.2% relative difference vs. Variant A (percentage difference formula)." Include the absolute numbers (e.g., 2.0% vs. 2.4%) so readers can verify. A quick checklist: formula used, absolute rates, sample size, and whether the result is statistically significant.

A simple template for Control vs. Treatment: "Control: 2.0% conversion. Treatment: 2.4% conversion. Relative lift: 20% (percent change). Statistically significant at p < 0.05." That gives stakeholders everything they need to understand and replicate the analysis. Clarity beats brevity when the formula choice affects the number.

Mistake 1: Using percentage difference when you have a Control. Control is the baseline—use percent change (relative lift). Fix: Use percent change for Control vs. Treatment; reserve percentage difference for Variant A vs. Variant B.

Mistake 2: Using percent change for Variant A vs. Variant B when neither is the baseline. Fix: Use percentage difference when comparing two new ideas with no established status quo.

Mistake 3: Reporting only the percentage and hiding the small absolute difference. A 20% "lift" from 2% to 2.4% is a 0.4 percentage point increase. Fix: Always report the raw absolute rates (e.g., 2.0% and 2.4%) alongside the percentage point difference so stakeholders understand the real impact.

Mistake 4: Mixing formulas across reports or dashboards. One slide says "25% lift" (percent change), another says "22.2% difference" (percentage difference)—same test, different numbers, and no explanation. Fix: Standardize on one formula and document it. Note: many A/B platforms use percent change (relative lift) when a control exists, because they treat it as the baseline—so for Control vs. Treatment, you are aligned with common practice.

Control conversion 2.0%, Treatment 2.5%. For Control vs. Treatment, percent change is correct: (2.5 − 2.0) / 2.0 = 25% lift. If the same numbers were Variant A vs. Variant B (no baseline), percentage difference would apply: |2.5 − 2.0| / 2.25 ≈ 22.2%. Same data, different formulas depending on whether you have a baseline. Be explicit about which scenario applies.

The absolute gap is 0.5 percentage points. Whether that is "25%" or "22.2%" depends entirely on the denominator you choose. For more everyday applications of percent change, see our guide to real-world percent change; for a broader view of when math calculators matter, see everyday math calculators that matter.