Inside 140+ A/B Tests: What the Data Actually Shows About CRO Win Rates

Most win rate statistics you encounter in CRO are aggregated across thousands of disconnected programs on platforms like Optimizely or VWO. That data is valuable, but it flattens the signal: a first-time experimenter running a button color test and a mature team running hypothesis-driven funnel tests both count equally.

This analysis is different. It draws from 142 A/B tests across two continuous, structured experimentation programs — one covering 45 tests across a multi-brand direct energy acquisition funnel, one covering 97 tests from a digitally-mature e-commerce operation. Both programs ran tests with documented hypotheses, pre-committed sample sizes, and consistent outcome classification.

The data is fully anonymized. No brand names, no raw conversion rates. What follows is aggregate findings, relative lifts, and directional patterns — the kind of numbers you can actually benchmark against.

Methodology

Dataset composition:

• Program A: 45 tests, energy sector, SaaS-adjacent acquisition funnel (CTA, form, layout, copy, pricing categories)
• Program B: 97 tests, e-commerce, transaction-focused (conversion rate primary metric in 57% of tests)
• Combined: 142 tests, all with documented outcomes (winner / loser / inconclusive)

What “winner” means: A test was classified as a winner when the primary metric reached statistical significance at ≥95% confidence AND the variant outperformed control on that metric. Tests that hit significance on a secondary metric only are classified as inconclusive.

Anonymization approach: All lifts are reported as relative percentages. Duration data is in days. No absolute conversion rates or revenue figures are included.

The Aggregate Picture