Choosing a test type

Creating an experiment starts with one decision: what kind of test this is. The choice shapes every later step, so it comes first — and it is permanent once the draft exists. Changing your mind means starting a new draft, which is deliberate: variants, verification, and measurement all depend on the type, and converting between them silently would invalidate all three.

Content test

Change something on a page and test it against the original, on the same URL. A headline, a hero, a section, or a whole page generated by Studio — the variant is applied in the visitor's browser, so there is no theme change and no second URL. This is the default kind of experiment and works on any platform the Runtime runs on. The traffic can be split by Proa or by an external tool that marks the variant with a URL parameter — see Traffic split and targeting.

Use it when the question is "does this change to the page sell more?"

Template test

Shopify only. Instead of changing elements on a page, each variant renders a different theme template — the control keeps the theme's current template, each variant renders an alternate template you created in the theme editor. The page's content (the product, the collection) stays the same; the whole layout around it is what's being tested.

Use it when the change is too large to express as element edits: a redesigned product page, a different collection layout. Template tests have their own setup and verification flow — read Template tests on Shopify before creating one.

Split URL test

Each variant sends the visitor to a different page of the same store. The control stays on the origin page; a visitor assigned to a variant is taken to that variant's destination URL, once, and measured from there. The destination is a full page you build however you want — a rebuilt landing page, an alternative offer page — not an overlay on the original.

Use it when the alternative already exists as its own page, or when the change is a whole different page rather than edits to the current one. Split URL tests have their own preparation and checks — read Split URL tests before creating one.

Campaign page

For paid traffic. You send ads to a URL carrying a campaign parameter, and each parameter value gets its own experiment — grouped so that a visitor only ever enters one of them. The wizard shows you the exact entry URL for each value as you type, so what you paste into the ad platform is never a guess.

Use it when different ad audiences deserve different tests, and the tests must not contaminate each other.

Choosing

If you are unsure, it is almost always a content test. Template tests exist for theme-level redesigns, split URL tests for testing one page against another, and campaign tests for ad-driven landing flows. The wizard shows a concrete example under each option, and nothing goes live at creation time either way: the wizard only produces a draft, and publishing is a separate, reversible step.