How do I A/B test a landing page in ElasticFunnels?

Open the page's events, add a Split Test node and connect it to the Funnel Entry Point, name the variants and set the traffic split, point the challenger at the new page with Load Another Page, and publish. Read the result in Split Tests.

What does split testing mean, and is it the same as A/B testing?

Split testing shows two or more versions of a page to comparable visitors at the same time and keeps the one that converts better. Most people use split testing and A/B testing for the same thing; a test with more than two variants is sometimes called A/B/n.

Does the page URL change during the test?

No. Visitors keep landing on the URL your ads point to. The split runs in the page's events, and Load Another Page serves the variant's content at that same URL, so your ads never change.

How long should an A/B test run?

Until each variant reaches the sample size you planned before launch, and for at least one full week so every weekday is in both arms. Divide the visitors the test needs by your daily traffic to know the length before you start. Do not stop the first time the p-value dips below 0.05.

How many visitors does an A/B test need?

It depends on your current conversion rate and the smallest lift you want to detect. The demo test in step 7, 3.30% against 4.05%, needs about 9,900 sessions per variant at 95% confidence and 80% power; the A/B test sample size calculator works it out for your numbers. Min Sample Size per Variant in the Split Test settings (1,000 on the demo test) is only the point below which results are flagged as low sample, not a target.

How does ElasticFunnels decide whether a result is significant?

It compares each variant's conversion rate with the control using a two-proportion z-test at the significance level you set (0.05 by default, or 95% confidence). Statistical Significance shows the p-value and the confidence interval for the difference, and marks a variant Sig. only when p is below the level and the interval does not include zero. When you run 3 or more variants, the winner check applies a Bonferroni correction, dividing the significance level by the number of challengers.

What counts as a conversion in the split test report?

A purchase. Conversion Rate is the number of sessions that bought divided by all sessions, per variant, and each order is credited to the variant the buyer was shown. Primary Metric in the Split Test settings (Conversion Rate by default) sets what a winner is picked on. Revenue, AOV and Net EPC are shown per variant too, but the significance test runs on conversion rate.

An A/A test splits traffic between two identical versions to check that the setup and tracking are sound: both arms should come out about the same. In ElasticFunnels, leave both variants without a Load Another Page node, so both serve the current page.

How many variants can one split test have?

Between two and five. Set Total Variants on the Split Test node and it adds one Traffic Distribution node per variant.

Does a returning visitor see the same variant?

Yes, for 7 days in the same browser. The first visit stores the variant in a cookie, and later visits get the same page even if you change the percentages. A different browser or device, or cleared cookies, counts as a new visitor. While the test is paused everyone sees the control, and once you mark a winner everyone sees the winner.

Can I pause a test without losing its results?

Yes. Pause test stops sending traffic to the variants. The test stays open and keeps its results; pausing does not end it.

How do I end the test and keep the winner?

Click Mark as Winner on the winning variant's Traffic Distribution node. The test is finalized with that variant as the winner.

Can I test one section instead of the whole page?

Yes. Use a Component Split Test node instead of Split Test to rotate a single section of the page, such as the headline block, while the rest of the page stays the same.

Can AI run an A/B test for me?

It can do the legwork. With the ef CLI, an agent such as Claude Code or Codex can read a test's significance result and your funnel stats and propose the next test for you to approve. The decision rule stays the same: the planned sample size and the significance level, and you pick the winner.

Does A/B testing a landing page hurt SEO?

Not when it follows Google's guidance: show Googlebot the same variants people see, never pick a variant by user agent, and end the test when you have an answer. A same-URL test keeps it simple, because there is only one URL to index.

How to A/B Test a Landing Page in ElasticFunnels (7 Steps)

What you'll build

A same-URL split test on a VSL (video sales letter) page: 70% of visitors keep seeing the current page (the control) and 30% see a long-form version, while the ads keep pointing at one unchanged link. You set it up in the page's events, publish it, and read the result on purchases, with statistical significance computed for you.

New to A/B testing? Read the full guide first: split testing landing pages under one URL covers why one URL matters for paid traffic, sample sizes and how to judge a test. Unfamiliar term? The glossary explains it in a sentence or two.

You need two versions of the page: the one that runs today and the variant. In this tutorial they are the demo brand's Vitality VSL — control and Vitality VSL — long-form.

1Open the page's events

In Pages, search for the page your ads already point to and click its lightning icon (Page Events). The canvas opens with a Funnel Entry Point: every visit to this page starts there.

The Pages list is filtered to Vitality VSL, then Page Events opens on the control page with its Funnel Entry Point — Search, then the lightning icon on the row. The test lives on the page the traffic already reaches.

2Add a Split Test node

Click Add Node and pick Split Test. It arrives with two Traffic Distribution nodes, one per variant, at 50/50. The name fields are empty for now; you fill them in step 4.

Add Node opens the node picker; choosing Split Test adds the node with two Traffic Distribution nodes at 50 percent each — Split Test takes 2 to 5 variants (Total Variants). Start with two variants unless you have enough traffic to support more.

3Connect the entry point to the split

Drag the two variant nodes apart so the graph reads left to right. Then drag from the Funnel Entry Point's output pin (the connector on its right edge) to the Split Test's input. From now on every visit goes through the split.

The two Traffic Distribution nodes are dragged apart, then a wire is drawn from the Funnel Entry Point to the Split Test — A node that is not connected to the entry point never runs.

Faster way: drag from a node's output pin and release on empty canvas. The node picker opens, and the node you pick is created already connected to that output. Step 5 uses it.

4Name the test and set the traffic split

Give the test and each variant a name you will recognise in the report, such as A — control and B — long-form. Then set % of traffic, the traffic split: type 70 on the control and the other variant takes the remaining 30. New visitors are assigned in a fixed rotation that follows those percentages, not a random draw, so the split holds from the first visits.

Sending most traffic to the page that already sells limits the cost if the variant is worse, but it is slower. The 30% arm collects visitors at 3/5 the pace of a 50/50 arm, so reaching the same number per variant takes 1.67 times as long. The bigger control arm makes up part of that: for the same statistical power, a 70/30 test needs about 1.2 times the total traffic of a 50/50 test.

The split test is named VSL page, control vs long-form; the variants are named A, control and B, long-form; typing 70 on A sets B to 30 — The percentages always add up to 100: change one and the others rebalance.

5Point variant B at the new page

Drag from variant B's output pin and release on the canvas. In the picker, choose Load Another Page: it is created already connected. Select the long-form page. Variant A has nothing after it, so its visitors keep seeing the current page.

A wire is dragged from variant B, Load Another Page is picked and arrives connected, and V1 Vitality VSL long-form is selected — Load Another Page replaces the content at the same URL. The visitor never sees a redirect.

6Publish

Click Publish. The Split Test node now shows the test as live, with Pause test and View results, and each variant gets a Mark as Winner button.

What counts as a conversion: a purchase. The report divides the sessions that bought by all sessions for each variant, and each order is credited to the variant the buyer was shown. The metric a winner is picked on is Primary Metric in the Split Test node's Advanced Settings (Winner Selection tab), Conversion Rate by default.

Publish saves the page events; the Split Test node shows Pause test, Live and View results, and each variant shows Mark as Winner — Pause test stops sending traffic to the variants but keeps the test and its results.

7Read the result

Open Split Tests and click Results on your test, or View results on the node. A new test needs traffic first, so the clip shows the same two pages after a 30-day run on the demo brand, split 65/35 (the report shows the traffic each arm actually got, 66.5/33.5):

A (control): 5,493 sessions, 181 purchases, a conversion rate of 3.30%
B (long-form): 2,764 sessions, 112 purchases, 4.05%

Conversion Comparison shows B at +23.0%, and Statistical Significance marks it Not Sig.: the p-value is 0.079, and the confidence interval for the difference runs from −0.12 to +1.63 points, so it includes zero.

B is ahead, but a test this size can't confirm a lift like that. Detecting 3.30% → 4.05% at 95% confidence and 80% power takes about 9,900 sessions per variant (check it in the sample size calculator). At this page's 275 or so sessions a day, that is about 10 weeks at 50/50. So plan the sample before launch, run the test to it, or test a bigger change that needs fewer visitors. The split testing guide explains how to plan it.

Min Sample Size per Variant (Advanced Settings, 1,000 on this test) is not that number. It is only the point below which the report flags a result "interpret with caution". Set it to the sample your test actually needs.

Split test results: B long-form converts at 4.05 percent against 3.30 percent for the control, plus 22.97 percent, marked not significant at p 0.0793 — When each variant has reached the planned sample and the variant is marked Sig., click Mark as Winner on it to end the test.

Common mistakes

Leaving the entry point unconnected. The split never runs, and every visitor sees the control.
Changing the ad at the same time. Change one thing per test, or you cannot tell what moved the result.
Stopping at the first good-looking p-value. Decide the sample size before launch with the sample size calculator and enter it as Min Sample Size per Variant in the Split Test node's Advanced Settings (ElasticFunnels flags results until each variant reaches it). Run at least one full business cycle, and judge the result at the planned sample size, not the first time p dips below 0.05.
Judging on clicks. Clicks to the checkout don't pay. Read buyers (the Customers column) and Net EPC per variant: the Results page shows both in the variant table, and Net EPC again in the EPC Comparison card.
Building the variant at a new URL for the ads. That edits the ads and resets their learning. Keep one URL and let Load Another Page swap the content.

Keep going

Why same-URL testing matters for paid traffic, and how many visitors a test needs: Split testing landing pages under one URL. Everything the split test engine does, including section-level tests: landing page split testing. Letting an AI agent read the results and propose the next test: AI A/B testing. Plan the sample before launch: A/B test sample size calculator. Terms explained: the glossary. All tutorials: the Academy.

A/B test a landing page in 7 steps