A/B testing funnel steps

An A/B test puts a second page behind one funnel step. Half of the visitors see each page, at the same address, until there is enough evidence to say which earns more. The product will not call a winner on a lucky opening run, and it never switches pages by itself.

Setting one up

On the Tests tab of the funnel editor, pick a step and a page to test it against. Write it as an ordinary page — the alternative is a normal WordPress page with the step’s block or shortcode on it. Start the test. From then on each visitor is assigned to one page and keeps it.

What counts as a conversion

That depends on the step. For a landing page, a conversion is a visitor reaching the checkout. For the checkout, it is an order being paid. For an upsell or downsell, it is the offer being accepted. The Tests tab says which applies.

When the product is allowed an opinion

Two floors, both hard:

• 100 conversions on each page. A lucky opening run crosses any threshold; a hundred does not.
• A full week. Weekday and weekend traffic convert differently, and the opening days are not the whole picture.

Until both are met the tab says Not enough data to call this yet, and why. After them it says one of three things: B wins, with the probability it is better and what switching would cost you if that is wrong; No difference worth acting on, keep the one you have; or At this traffic it will not separate any time soon.

The arithmetic is Bayesian — the probability that one page’s true rate is higher than the other’s, and the expected loss from choosing it — because stopping on a probability threshold alone is how you ship noise.

Deciding

Use the winning page makes it the step’s page. Stop the test keeps the page the step already had. Nothing is applied automatically — the decision is yours; the product only says when it is allowed to have an opinion.

What it is not

A visual editor, a multivariate engine, or a traffic allocator. One step, two pages, one honest answer.