Point. Change. Ship.
Then wait for the maths.
A visual editor that runs against your real pages, and statistics honest enough to say "not yet". Most tools will happily call a winner off four conversions. This one refuses to.
Significant · +36.6% uplift
Six ways to change
a page you did not build.
Open your own site in the editor, click an element, and record a change against its selector. No branch, no build, no release window.
Text
Replaces only the text nodes, so any child markup inside the element survives untouched.
HTML
Replaces the inner markup entirely — sanitised before it is ever stored.
Attribute
Sets a single attribute: href, src, alt, title, target, rel, class, aria-label, placeholder or value.
Hide element
Removes it from view without deleting anything from your codebase.
Insert before
Injects new markup immediately above the element.
Insert after
Injects new markup immediately below the element.
Markup is sanitised before it is stored
Injected HTML is filtered to a safe tag list and a safe attribute list. Everything else is dropped — which takes out the whole on* family without having to enumerate it. Nobody on your team can accidentally plant script on a production site through this editor.
a
strong
em
span
p
h1–h6
ul
ol
li
img
picture
button
br
mark
Draft it. Run it.
Call it.
Draft
Build variants and goals
Running
Traffic is being split
Paused
Held without losing data
Completed
A winner is recorded
Where it runs
URL rules decide which pages take part — is exactly, contains, starts with, or a wildcard pattern.
How much traffic
Allocate a percentage of visitors to the experiment, and weight each variant against the control.
Sticky assignment
A visitor keeps the variant they were given, so nobody sees the page change under them mid-visit.
Measured on its
own terms.
A/B goals are deliberately self-contained: measurement never depends on the User Journey module being switched on. Mark one as primary and the rest are tracked alongside it.
Page pattern
CSS selector
Event name
Guard rails,
not vanity.
Results run through a two-proportion z-test at two-sided 95% confidence — and refuse to report anything until the sample can carry it.
per variant, minimum, before a result is reported at all
per variant, minimum, or the estimate is still noise
two-sided, before anything is called significant
Until then the experiment reads collecting, with a plain note explaining what is still missing — "needs 100 visitors per variant before a result means anything."
Stop losing the visitors
you already paid for.
More leads from the traffic you already have, and follow-up that sends itself. Free to start — no card, no monthly fee.