Split URL testing sends a share of the visitors who land on one page — the origin — to a different URL, then compares how the two pages perform. Reach for it when the two experiences live at different addresses: a campaign landing page, a hand-curated collection, or a page built in a page builder — rather than a restyle of the same page.
If you're restyling the same page, use a Theme or Template experiment instead — see Which experiment should I run?
How it works
Origin page — you choose the page the test runs on. Visitors enter the test only when they land on this page.
Assignment — each visitor is assigned to an arm the first time they reach the origin page. The assignment is sticky: returning visitors always get the same experience, so your data stays clean.
Control — control visitors stay on the origin page, unchanged.
Variant — variant visitors are redirected to the variant URL before the page renders. The redirect replaces the current history entry, so the back button never loops them through it.
Anti-flicker — the page is briefly hidden until the decision is made, so no one sees a flash before the redirect.
Origin URL and variant URL
A Split URL experiment has two arms, each identified by a URL:
Arm | URL you set | What the visitor sees |
Control | Origin URL — the page the test runs on (e.g. /pages/black-friday) | Stays on the origin page |
Variant | Variant URL — an absolute https:// address | Redirected to the variant page |
The origin is matched on the page's canonical path, exactly — no wildcards. A product reachable at both /products/shoe and /collections/summer/products/shoe is matched by its canonical /products/shoe either way.
How it differs from Theme and Template
Theme keeps the same URL and swaps your whole theme.
Template keeps the same URL and swaps one page type's layout within your theme.
Split URL sends the visitor to a different URL.
Reading your results
Cooee shows the same per-arm metrics as the other experiment types:
Metric | What it measures |
Impressions | Unique visitors entered into the arm in a real session |
Conversion Rate | % who completed a purchase |
Revenue per Visitor | Total revenue ÷ total visitors |
Add to Cart Rate | % who added to cart |
Bounce Rate | % who left without engaging |
Session Duration | Average time on site |
Results update daily. Run for at least 2 full business cycles (typically 14 days) before deciding — see Ending a split URL experiment.