A/B Page Testing
Point a stream at more than one destination page and Cloaking X rotates real, passed visitors across them automatically — with a built-in report showing which one actually converts better.
What it is#
Every stream can hold one or more destination URLs. Add a second one (or more, depending on your plan) and any visitor who passes the check is sent to one of them instead of always the same page — the start of an A/B (or A/B/C…) test with no separate tool to set up.
This is the same destination-pages field documented field-by-field in the Stream Settings Reference — this page focuses on how to actually run a test with it and where to find the results.
Setting up a test#
- Open a stream (new or existing) and go to its Destination Pages step.
- Click + Add URL to add a second variant — and a third, fourth, and so on, up to your plan's limit (see below).
- Pick a rotation mode for how visitors are split across them.
- Save the stream and send traffic as usual — there's no separate "start test" step. Rotation begins the moment two or more pages are configured.
Choosing how pages rotate#
Three rotation modes decide how a passed visitor picks a page:
- Random (default) — the fairest starting point for a fresh test: every visitor has an equal, independent chance at each page.
- Sequential — cycles through your pages in order, one visitor at a time, for an even, deterministic split.
- Timer — every visitor sees the same page for a fixed window, then the whole stream advances together — useful if you'd rather compare pages over full stretches of time than visitor-by-visitor.
Full field detail, including the exact rotation-interval setting for Timer mode, is in the Stream Settings Reference.
OS-specific pages aren't part of the test#
If a stream also has an OS-specific override configured (Apple, Windows, or Android), a visitor on that OS always goes straight to that page — checked before rotation is ever consulted, so that traffic never enters the test at all.
That's deliberate: an OS override routes one specific device population somewhere on purpose, so folding its numbers into a randomized test would compare "this page" against "this operating system's users", not against your other destination pages. Reports keeps that page's own numbers visible — it's just kept out of the winner comparison so you don't draw the wrong conclusion from it.
Reading your results#
Open Reports, select the stream, and scroll to Variant Performance — it appears automatically once a stream has two or more comparable page variants. It shows Clicks, Conversions, and Conv. Rate per page, and marks the current best performer Winner once it has enough traffic for the comparison to mean something — a handful of clicks either way isn't enough to call it.
How many pages you can test#
The Free plan includes a single destination page, so there's no second page to test against until you upgrade. Every paid plan raises that limit:
up to 5 pages
up to 20 pages
up to 100 pages
unlimited pages