What does conversion rate optimization mean?
Conversion rate optimization means increasing the proportion of visitors to an online store who buy, by finding where the store loses them and testing changes that might keep them, with the result read from a controlled experiment rather than from a before-and-after glance. The word optimization implies a method: a hypothesis, a variant, a control, a sample size, a reading. Changing a button colour because a blog post said so is not CRO, whatever the button colour does.
The definition is narrow on purpose. Stores that treat CRO as a stream of redesigns cannot tell which change did what, and end up with a site that is different but not measurably better.
Where does a store funnel leak?
A store funnel leaks at three main steps, each with its own typical causes, and a drop-off map shows which leak carries the most revenue. The causes below are the ones a Conversion FTE looks for evidence of first; none is assumed without a signal from the store's own data.
- Product page: shoppers arrive and leave without adding to cart; typical signals are missing information, weak images, slow load, no visible trust or delivery answer.
- Cart: items are added and the shopper leaves; typical signals are shipping cost revealed late, no clear path to checkout, an unexpected total.
- Checkout: the shopper starts and does not finish; typical signals are forced account creation, form errors, missing payment options, a surprise at the last step.
Why is revenue per visitor the honest metric?
Revenue per visitor is the honest headline metric because conversion rate alone can be raised in ways that lose money. A blanket discount raises the rate and lowers the order value; a free-shipping threshold set too low raises the rate and eats margin; pushing the cheapest product raises the rate and lowers revenue. Revenue per visitor, total revenue over the test period divided by visitors, captures both the rate and the value, and it is the number a Conversion FTE reports first.
Conversion rate and checkout completion are reported beside it as diagnostics, because they say where the change acted. The headline is the revenue.
What makes an experiment valid?
An experiment is valid when visitors are assigned to control and variant at random, when the sample size needed to detect the smallest effect worth caring about was calculated from the baseline before the test started, when the result is read at that sample size and not before, and when the metric was chosen in advance. Break any of these and the result is a story, not a finding. Reading a test daily and stopping when it looks good is the most common break, and it produces false winners at a rate most stores would not accept if they saw it.
Validity also needs a stop rule for the other direction: a test that reaches the sample size with no detectable difference is a loss, reported as such, and the hypothesis is retired or revised.
What can be tested, and what should not be?
Almost anything on the path to purchase can be tested: product-page layout and copy, image order, the placement of delivery and returns information, cart summaries, shipping-threshold messaging, checkout field order, trust signals, and page speed changes. What should not be tested without a person's decision is anything that changes what the customer is charged or agrees to: prices, discounts, payment options, legal and policy copy. On ftes.ai those sit on the Conversion FTE's always-gated list.
The distinction is not about risk to the test; it is about accountability. A price test is a pricing decision, and pricing decisions belong to the owner.
What does a Conversion FTE change about CRO?
A Conversion FTE changes CRO from an occasional project into an owned function: the drop-off map refreshed daily, hypotheses ranked by revenue at stake and evidence, every experiment proposed with its sample-size plan and stop rule, launched only on approval, read only at sample size, and reported with losses named. The owner approves each test and reviews the numbers monthly.
What it does not change is who owns the offer and the price, and it does not pretend a store without a testing tool can run a valid test. It says so, and proposes the tool.