Landing page split testing
Variants rotate on the server, so Meta, Google and TikTok never see a new URL. Winners are picked on sales, not clicks.
Test the page, a block or the whole path
Every test runs on the same data as your checkout, so each variant is scored on real orders. See all features.
Split test without resetting ad learning
A new URL means a new ad review and a fresh learning phase. Here the variant is chosen before the page is sent, so the ad points at one address for the life of the test.
- Weight traffic per variant; new visitors rotate in exact proportion
- Returning visitors keep their variant
- Pause a test without losing its results
- Mark the winner in one click
Decide on revenue, not click-through
A hook that gets more clicks and fewer sales is a loser that looks like a winner. Pick the metric that pays, and let the statistics tell you when it is real.
- Primary metric
- Revenue, profit, EPC, AOV, CPA or conversion rate
- Method
- Two-proportion z-test (frequentist)
- Before launch
- Sample size calculator
- While it runs
- Fixed sample size, no early peeking
- Many variants
- Bonferroni correction
- Results
- Significance, EPC and revenue impact
What media buyers test first
VSL against text
Same traffic, two formats, one URL.
Presell angles
Two advertorial stories in front of the same offer.
Checkout paths
Route each arm to a different checkout or upsell flow.
Build it, test it, sell it, track it
Every variant tied to revenue
Revenue, EPC and profit per variant, per source.
Page Builder & AIMake the variant with AI
Describe the change or clone a page, then test it.
Developer PlatformScript tests from code
ef pages events pushes test graphs from git.
Test the checkout too
Route each arm to a different checkout or offer.
Split testing FAQ
Can I split test without changing my page URL?
Yes. Every variant runs under one URL. ElasticFunnels picks the variant on the server before the page is sent, so the browser only ever sees one address. Links, tracking and postbacks stay on that URL no matter how many variants you run.
Will testing send my Facebook or Google ads back into review?
No. The URL never changes, so the ad platform sees no new destination to review and no reason to restart learning. Campaigns keep optimizing while the page underneath them changes.
How is the winner of a split test decided?
You decide, with Mark as Winner, unless you turn on auto-winner (it is off by default). You choose the primary metric: conversion rate, revenue, net revenue, AOV, EPC, profit or CPA. Significance is tested on conversion rate with a two-proportion z-test: each variant gets a p-value and a confidence interval, with a Bonferroni correction when you run 3 or more variants. Plan the sample size before launch and judge the test when each variant reaches it.
What is the difference between server-side and client-side split testing?
Client-side testing loads one page, then swaps content in the browser, which causes flicker and can break attribution. Server-side testing picks the variant before the page is sent, so there is no flicker, no redirect and nothing for a tracker to misread.
What can I test besides headlines?
Whole pages, individual components such as a hero or pricing block, and whole funnel paths. Put a condition in front of the test to run it only for one campaign, source or query parameter.
Does a visitor always see the same variant?
Yes. Assignment is sticky for 7 days in the same browser, so a returning visitor gets the variant they saw first, even if you change the weights. You can weight traffic per variant and mark a winner in one click.