A/B Testing Without Tools

You do not need Optimizely or VWO for every A/B test. For simple copy or layout tests — changing a headline, swapping a CTA colour, reordering sections — GTM can handle assignment, variant application, and result tracking entirely on its own. This recipe shows you how.

Anti-flicker snippet

Add this snippet to your <head> before the GTM snippet. It hides the page until the variant is applied:

<!-- Anti-flicker: hide body until GTM has applied the variant -->
<style id="anti-flicker">.async-hide { opacity: 0 !important }</style>
<script>
  (function(a,s,y,n,c,h,i,d,e){
    s.className+=' '+y;
    h.start=1*new Date;
    h.end=i=function(){s.className=s.className.replace(RegExp(' ?'+y),'')};
    (a[n]=a[n]||[]).hide=h;
    setTimeout(function(){i();h.end=null},c);
    a.dataLayer=a.dataLayer||[];
  })(window,document.documentElement,'async-hide','dataLayer',3000,{});
</script>

This hides the page for up to 3 seconds. Your variant tag must call dataLayer.hide.end() when it finishes to reveal the page.

Create a Custom HTML tag in GTM that runs on the target page and assigns the user to a variant. Fire it with your highest tag priority.

(function() {
  var EXPERIMENT_ID = 'headline-test-v1';
  var VARIANTS = ['control', 'variant_a'];
  var TRAFFIC_PERCENT = 100; // Percentage of users in the experiment

  // Check existing assignment
  var cookieName = 'exp_' + EXPERIMENT_ID;
  var existingVariant = null;

  document.cookie.split(';').forEach(function(cookie) {
    var parts = cookie.trim().split('=');
    if (parts[0] === cookieName) existingVariant = parts[1];
  });

  // Not in experiment (traffic split)
  if (!existingVariant && Math.random() * 100 > TRAFFIC_PERCENT) {
    if (window.dataLayer && window.dataLayer.hide) window.dataLayer.hide.end();
    return;
  }

  // Assign new variant
  if (!existingVariant) {
    existingVariant = VARIANTS[Math.floor(Math.random() * VARIANTS.length)];
    var expires = new Date(Date.now() + 30 * 86400000).toUTCString(); // 30 days
    document.cookie = cookieName + '=' + existingVariant + '; expires=' + expires + '; path=/; SameSite=Lax';
  }

  // Store assignment for GTM to read
  window.dataLayer = window.dataLayer || [];
  window.dataLayer.push({
    experiment_id: EXPERIMENT_ID,
    experiment_variant: existingVariant
  });

  // Apply the variant
  applyVariant(existingVariant);

  // Reveal page
  if (window.dataLayer.hide) window.dataLayer.hide.end();

  function applyVariant(variant) {
    if (variant === 'control') return; // No change for control

    if (variant === 'variant_a') {
      // Example: Change headline text
      var headline = document.querySelector('h1.main-headline');
      if (headline) {
        headline.textContent = 'Your new compelling headline here';
      }

      // Example: Change CTA button colour
      var cta = document.querySelector('.hero-cta');
      if (cta) {
        cta.style.backgroundColor = '#2ecc71';
        cta.style.borderColor = '#27ae60';
      }
    }
  }
})();

Step 2 — Track the variant assignment

dataLayer.push() experiment_impression

Fires after variant assignment to record which variant the user saw.

// Add at the end of applyVariant(), or push separately after assignment:
window.dataLayer.push({
  event: 'experiment_impression',
  experiment_id: EXPERIMENT_ID,
  experiment_variant: existingVariant
});

Create a Custom Event Trigger
- Trigger type: Custom Event
- Event name: experiment_impression
Create Data Layer Variables
- DLV - experiment_id → experiment_id
- DLV - experiment_variant → experiment_variant
Create a GA4 Event Tag
- Event name: experiment_impression
- Parameters:
  - experiment_id → {{DLV - experiment_id}}
  - experiment_variant → {{DLV - experiment_variant}}
- Trigger: the Custom Event trigger
Register custom dimensions in GA4

Go to GA4 → Admin → Custom Definitions → Custom Dimensions:
- experiment_id: Event scope
- experiment_variant: Event scope
Set as a User Property for persistent attribution

In your GA4 tag → User Properties:
- active_experiment → {{DLV - experiment_id}}_{{DLV - experiment_variant}}
This allows filtering any GA4 report by experiment variant, not just the impression event.
Test in Preview Mode

Visit the experiment page. The variant assignment tag should fire, followed by the experiment_impression GA4 event. Refresh the page — the same variant should be assigned (cookie persists). Open in a private window — may get a different variant.

Reading results

In GA4, create an Exploration:

Dimensions: User > active_experiment
Metrics: conversions, sessions, engagement rate
Filter: active_experiment contains headline-test-v1
Compare: headline-test-v1_control vs headline-test-v1_variant_a

For statistical significance, export to BigQuery or Google Sheets and run a chi-squared test on the conversion counts.

Test it

Open GTM Preview and visit the experiment page
The assignment tag should fire first
The experiment_impression event should follow
Verify experiment_variant is either control or variant_a
If variant_a, verify the DOM changes have been applied (headline text, button colour)
Refresh the page — verify the same variant persists (cookie)
Open in a private window — may get a different variant (new cookie assignment)

Common gotchas

FOOC despite anti-flicker. If the variant tag fires after the page renders (e.g., Window Loaded trigger), users briefly see the control before the variant applies. Use DOM Ready or a Page View trigger (not Window Loaded) for the variant assignment tag. Set tag priority to a high number (e.g., 99) to ensure it fires before other tags.

50/50 randomisation is not guaranteed. Math.random() is not truly 50/50 across sessions — with small sample sizes, one variant may get 55-45. This is normal. Over 1000+ sessions, the split approaches 50/50. Do not manually adjust the cookie logic to force even splits.

Variant A users see control on second visit before cookie is set. The cookie is set synchronously in the same event loop as the variant application. There should be no gap between assignment and application. If you see users switching variants between visits, check that the cookie path and SameSite settings are correct for your domain.

A/B Test Tracking Integrating data from professional A/B testing tools with GA4.

User-ID Implementation Consistent user identification for cross-device test attribution.