metrics
Conversion rate
Conversion rate is the share of people who complete a goal - buy, sign up, subscribe - out of everyone who could have. It's usually quoted end to end, but it's really the product of every step's survival rate along the way.
The demo
A thousand people land on the shop. Drag each step's survival rate and watch the final conversion rate move - usually far more than you'd expect, because the steps multiply rather than add.
What this demo shows (text version)
A four-step purchase funnel begins with 1,000 visitors. Three sliders set the survival rate at each step: visit to basket, basket to checkout, and checkout to purchase. The bars show how many people remain at each stage, and a headline figure shows the overall conversion rate - the share of the original 1,000 who reach the end.
The point is that the steps multiply. Keeping 80% at each of three steps leaves 0.8 × 0.8 × 0.8 = 51%, not 80%. So a small improvement to the weakest step moves the total far more than polishing a step that already performs well, and a single end-to-end rate hides which step is leaking. Break the rate into its funnel and fix the biggest drop first.
Nudging one slider, you watched the final rate swing far more than the step you touched - because the steps multiply, they don't add. A funnel that keeps 80% at each of four steps doesn't keep 80%; it keeps about 41%. That compounding is why fixing the worst step usually beats polishing the best one.
A headline conversion rate hides where the money leaks. "2.3%" tells you nothing actionable; the per-step funnel tells you the basket-to-checkout drop is where half your buyers vanish. Always break the rate into its steps - the aggregate is a symptom, the funnel is the diagnosis.
My rule: hunt for the biggest drop, not the lowest rate. A step that holds 95% has little left to give; the step bleeding 60% is where a week's work moves the whole number. And watch for a rate that climbs while revenue falls - it usually means you scared off the browsers, not that you converted more.