We measure CRO success on three levels, and our primary metric is revenue per visitor — not conversion rate in isolation. Conversion rate can rise while average order value falls, leaving total revenue flat. RPV closes that gap.
At the test level, we report uplift, statistical confidence, and incremental revenue. Every test runs for full business cycles so weekday, payday, and campaign patterns don't distort the result, and we define the minimum detectable effect and required sample size before the test goes live — not after we've seen the numbers.
At the program level, we track testing velocity, win rate, average uplift per winning test, and cumulative annualized incremental revenue across 6–12 months. A single test can be luck; a year of program output cannot. This is also the number that answers the only question a CFO actually asks: annualized incremental revenue against program cost.
Alongside both, we track guardrail metrics — average order value, return and cancellation rates, support ticket volume, and page performance — so a winning test never wins by shifting a cost somewhere else in the business.
We also count losing tests as results. A variation that underperforms is a launch decision you didn't make and revenue you didn't lose, validated in three weeks instead of discovered in a quarter. Every test is reported with its hypothesis, its data, and the decision it informs, whether it won or not.
None of this works without a locked baseline. Before any test runs, we validate the analytics setup and establish the pre-program benchmark — we don't claim improvement on numbers we couldn't measure in the first place.