How to Run an A/B Test on Your WordPress Website Homepage

Using Google Tag Manager and Google Analytics 4

No plugins required. Full control over tracking and results.

A/B testing is one of the most reliable ways to improve website performance. Rather than guessing which version of a page will work better, you can split your traffic between two versions and let the data decide.

This guide walks through a complete split test setup using Google Tag Manager and GA4. It does not require any WordPress plugins and gives you full control over variant assignment, tracking, and results.

What Is an A/B Test?

An A/B test (also called a split test) sends visitors to two different versions of a page and measures which one generates more conversions — whether that means form submissions, enquiries, bookings, or sales.

For this example we will test two versions of a homepage:

/ (Version A — existing homepage)
/new (Version B — new landing page)

Visitors arriving at the homepage are randomly assigned to one version and remain on that version across all future visits until the test ends or their browser data is cleared.

Important

localStorage vs sessions This setup uses localStorage to store the variant assignment. Unlike session storage, this persists across browser sessions — meaning a returning visitor will always see the same version they were originally assigned. This is intentional and produces more reliable results.

Step 1: Create the Two Landing Pages

Before setting up tracking, you need both versions of your page ready in WordPress.

Your URL structure should be:

/ Version A — your current homepage
/new Version B — the page you want to test

Version B should test one meaningful change. Common things to test include:

A different headline or subheading
A stronger or repositioned call-to-action
A simplified layout with fewer distractions
Additional social proof or testimonials
A new hero image or video

Best practice

Change only one major element at a time. If you change multiple things and Version B wins, you won’t know what caused the improvement. Isolating one variable gives you a clear, actionable result.

Step 2: Set Up the Split Test Script in GTM

In Google Tag Manager, create a new Custom HTML tag and paste in the following script.

This version includes variant tracking via the dataLayer, which is required for reporting in GA4.

The Script

<script>
(function() {
  try {
    var variant = localStorage.getItem('homepage_test');

    if (!variant) {
      variant = Math.random() < 0.5 ? 'A' : 'B';
      localStorage.setItem('homepage_test', variant);
    }

    // Push variant to dataLayer for GA4 tracking
    window.dataLayer = window.dataLayer || [];
    window.dataLayer.push({
      event: 'ab_test_assigned',
      ab_variant: variant
    });

    // Redirect B visitors on the homepage only
    if (variant === 'B' && window.location.pathname === '/') {
      window.location.replace('/new/');
    }
  } catch(e) {}
})();
</script>

What this script does

Checks whether the visitor has already been assigned a variant
If not, randomly assigns them to A or B (50/50) and stores the result in localStorage
Pushes an ab_test_assigned event to the dataLayer so GA4 can record the variant
Redirects B visitors from / to /new/ using location.replace (prevents back-button loops)
Wrapped in try/catch to handle rare cases where localStorage is unavailable (some private browsers)

GTM tag settings

Tag type: Custom HTML
Trigger: All Pages
Fire on: DOM Ready — not Page View or Window Loaded

Why DOM Ready matters

If the tag fires on Window Loaded, visitors assigned to Variant B will briefly see the original homepage before being redirected. Firing on DOM Ready executes the script earlier and minimises any visible flash.

Prevent /new/ from triggering the redirect

Make sure this tag only fires on the homepage, not on /new/ itself. In GTM, set the trigger to:

Trigger type: Page View — DOM Ready
This trigger fires on: Some Page Views
Condition: Page Path equals /

This prevents visitors who land directly on /new/ from being caught by the redirect logic.

Step 3: Set Up GA4 Tracking

The dataLayer push in the script fires an event called ab_test_assigned with a parameter ab_variant containing either A or B. You need to capture this in GA4.

In Google Tag Manager

Create a Data Layer Variable:

Variable type: Data Layer Variable
Data Layer Variable Name: ab_variant
Name it: DLV – ab_variant

Create a GA4 Event tag:

Tag type: Google Analytics: GA4 Event
Event name: ab_test_assigned
Add event parameter — Parameter name: ab_variant, Value: {{DLV – ab_variant}}
Trigger: Custom Event — Event name: ab_test_assigned

In Google Analytics 4

Go to Admin > Custom Definitions > Custom Dimensions
Create new dimension: Name: AB Variant, Scope: Event, Parameter: ab_variant

Track a meaningful conversion event

Set up a conversion event based on a real business outcome. Common examples:

Visit to /thank-you (form submission confirmation)
Click on a phone number or email link
Booking confirmation page

In GA4, mark your chosen event as a conversion under Admin > Conversions.

Why this step is essential

Without the dataLayer push and the GA4 custom dimension, you have no way to segment your conversion data by variant. The redirect alone tells you nothing — you need GA4 to record which version each visitor saw.

Step 4: Run the Test Long Enough

One of the most common mistakes is stopping a test too early. Random variation in traffic and conversions can make one version look better in the first few days even when there is no real difference.

Minimum thresholds

At least 2 to 4 weeks of data
A minimum of 100 conversions per variant where possible
Check a statistical significance calculator before declaring a winner

A free tool such as abtestguide.com/calc lets you enter your visitor and conversion numbers to confirm whether results are statistically significant.

Step 5: Declare a Winner and End the Test

Once you have enough data, compare performance in GA4 by filtering on the ab_variant dimension.

What to compare

Conversion rate per variant
Engagement rate and average session duration
Bounce rate
Revenue or lead value per visitor if applicable

If Version B wins

Make /new/ your permanent homepage
Set up a 301 redirect from /new/ to / once promoted

After the test ends — important cleanup steps

Remove or pause the GTM split test tag
Delete or archive the old homepage variant
Clear localStorage if running a follow-up test with a different key name

Do not leave the split test tag running after the test ends.

If the tag keeps firing, visitors will continue being randomly assigned and redirected even when the test is over.

Always pause or delete the tag once you have a result.

SEO Considerations

Running a redirect-based split test carries some SEO risk if not handled carefully.

Google may crawl /new/ and index it as a separate page, creating duplicate content
Repeated 302 redirects from / to /new/ may affect how Google perceives your homepage

Recommended precautions

Add a canonical tag to /new/ pointing to / so Google treats it as a variant, not a separate page
Keep the test duration reasonable — extended tests of several months increase risk
Monitor Google Search Console for any coverage issues during the test

For most short-duration tests on low-traffic sites, the SEO impact is minimal. For high-traffic homepages or longer tests, consider server-side splitting instead.

Best Practices Summary

Principle	Guidance
Test one change at a time	Changing multiple elements makes it impossible to identify what caused any improvement.
Use meaningful conversions	Track actual business outcomes — leads, bookings, purchases — not just page views.
50/50 traffic split	Equal splits produce the clearest results and reach significance fastest.
Run for at least 2–4 weeks	Short tests are often misleading due to day-of-week traffic variation.
Check statistical significance	Use a calculator to confirm results before making decisions.
Clean up after the test	Remove the GTM tag and redirect logic as soon as the test is concluded.
Monitor SEO during the test	Watch Search Console for unexpected indexing of the variant URL.

What to Test Next

Once you are comfortable with the process, you can use the same setup to test any page on your site:

Pricing pages
Service landing pages
Lead generation pages
Campaign pages from paid advertising

Over time, consistently testing and improving your pages turns your website into a system that progressively improves its conversion rate rather than staying static.