A/B test your GoHighLevel funnels like an operator (not a hobbyist)
Split tests fail when the mechanics are wrong: no domain connected, ads bypass the split URL, cookies confuse your testing, or tracking fires on the wrong step. This page is the power-user setup so your experiments actually change revenue.
Start in 12 minutes (the setup sequence that keeps data clean)
If you do these steps in order, you avoid 90% of split-test "mystery" issues.
Pick one variable
Headline OR hero image OR CTA placement. Not all three.
Confirm domain is connected
Split testing requires a domain added to the funnel step.
Create variation
Duplicate control / use existing / create from blank.
Name variant URL paths
Edit dedicated URL paths (gear icon). Keep them readable.
Set split ratio
Start 50/50 unless you have a reason.
Lock tracking + launch
One conversion event, one confirmation step, then run traffic.
Do not edit the control while the test is running
If you change the control mid-test, your results become meaningless. Make changes on the variation and keep the control stable.
What clean tracking looks like during a therapy practice’s GHL split test.
GHL split test setup (what to do inside HighLevel)
Use HighLevel’s built-in split testing first. It’s lighter, faster, and keeps routing simple.
The click-path (clean + repeatable)
- Sites → Funnels (or Websites) → open your funnel
- Select the step you want to test
- Click Create Variation → choose method
- Adjust the traffic slider and Start Split Test
- Use the Stats tab to view performance
Behind the scenes, traffic is routed to each variant’s dedicated URL path based on your split settings.
How to QA properly (cookie-aware)
- Open the split-test step URL once → you’ll get assigned a variant
- To see the other variant, use incognito/another browser/device
- Don’t "refresh 20 times" expecting it to flip
Cookie-based assignment is why teammate QA should use multiple devices/browsers.
URL + ad link rules (the #1 reason split-test stats look "wrong")
Split tests can be configured perfectly, then ruined by the wrong link in ads, email, or buttons.
| Do this | Avoid this | Why it matters |
|---|---|---|
| Send ads to the split-test step URL | Using direct Control/Variation URLs in ads | Direct URLs bypass split logic and skew ratios/stats. |
| Keep the ad domain and funnel domain aligned | Sending traffic to a different domain | Split testing requires the correct domain context for consistent behavior. |
| If ad policy rejects "random routing," test on step 2 | Split testing the ad landing step when policy blocks it | Workaround: educational bridge step first, split test the next step. |
Operator move: name variant paths for analytics
Use clean paths like /book-a and /book-b. Your analytics can read variant identity from the URL with zero hacks, and debugging becomes 10x faster when someone asks "what did I see?"
Tracking-safe measurement (GA4/GTM) that doesn’t corrupt conversions
Your experiment is only as good as the conversion event. The clean rule: one "conversion" fires once, on the right step, and includes variant identity.
What to measure (pick one primary)
- Lead funnels: form submit / survey submit
- Booking funnels: appointment booked (confirmation step)
- Checkout funnels: purchase complete (thank-you step)
Don’t use "button clicks" as your primary win. It’s a proxy metric that often picks the wrong winner.
Common failure: double conversions
If your conversion fires on both the opt-in and thank-you steps (or global + step code), you’ll declare fake winners. Fix tracking before you spend on traffic.
Variant labeling snippet (URL-path based)
This reads the current path and sets a variant label you can push into GTM/GA4. It assumes you name the dedicated paths clearly (ex: /book-a vs /book-b).
(function(){
// Example path rules (edit to match your dedicated URL paths)
var p = (location.pathname || '').toLowerCase();
var variant = 'control';
if (p.indexOf('book-a') !== -1 || p.indexOf('var-a') !== -1) variant = 'A';
if (p.indexOf('book-b') !== -1 || p.indexOf('var-b') !== -1) variant = 'B';
// Make it available for GTM/GA4
window.atjExperiment = window.atjExperiment || {};
window.atjExperiment.variant = variant;
// Optional dataLayer push (GTM)
window.dataLayer = window.dataLayer || [];
window.dataLayer.push({
event: 'atj_experiment_view',
atj_variant: variant,
atj_path: location.pathname
});
})();
External scripts: when to use them (and when they quietly wreck your tests)
Best practice: use HighLevel split testing first. External experiment scripts add weight (hurts mobile), increase failure modes, and can conflict with cookie routing.
Use HighLevel split testing when…
- You’re testing page content/layout/CTA
- You want minimal script overhead
- You’re running ads and need reliability
- You want stats inside HighLevel
Consider external tools only when…
- You need multi-armed bandits / complex targeting
- You need cross-domain experimentation
- You have strict governance + engineering QA
Do NOT layer systems
Running HighLevel split testing and an external A/B script at the same time is how you get: variant conflicts, skewed ratios, broken tracking, and "impossible" results.
Declaring winners (and avoiding fake wins)
A winner is only real if it improves the primary conversion metric on enough traffic - and you didn’t change the rules midstream.
How to read results inside HighLevel
- Open the step → click Stats
- Compare Control vs Variation conversion rates
- Confirm the split ratio is roughly what you set
If the ratio is wildly off, it’s often because a direct variant URL is being used somewhere.
Anti-fake-win checklist
- Control wasn’t edited mid-test
- Same traffic source to the split-test step URL
- Conversion fires once on the correct step
- Mobile experience didn’t regress (speed matters)
Conversions double-counted across variants. GA4 shows 2x actual submissions. Can’t tell which variant won. $1,200 in ad spend wasted.
One conversion per session. Variant parameter in every event. Clear winner declared in 14 days. Ad spend ROI: 4.2x.
The difference between broken and clean tracking during a split test.
QA checklist before you spend money on traffic
Treat this like a release checklist. If any item fails, don’t launch.
Domain verified
Split test requires a domain added to the funnel step.
Variant paths named
Readable paths like /book-a and /book-b (helps analytics + debugging).
Ad links correct
Ads point to the split-test step URL (not direct control/variation URLs).
Cookie test done
Incognito/device checks confirm both variants show.
Conversion fires once
No duplicate fires from global code + step code.
Ad policy fallback ready
If rejected, split-test on step 2 and send ads to step 1.
Want us to run your experiment sprint (and harden tracking)?
Send the funnel URL(s), the step you want to test, and the conversion event you care about. We’ll implement split testing correctly, name variant paths for analytics, prevent double conversions, and launch with clean measurement so the "winner" is real.
FAQ
Fast answers for split-testing in HighLevel.
