How much traffic does each variant need before the split is reliable?

Watch deposit counts, not click counts — until a variant has at least a few dozen deposit conversions, it's too early to call a winner. On a small GEO that can take a couple of days.

Can I split-test push copy, not just pre-landers?

Yes — APEX lets you test push message variants the same way as creatives; the base segment splits sticky, so each subscriber consistently gets the same copy variant.

When should I kill the losing variant?

Once both variants have a comparable sample size and the ROI gap holds steady for several days in a row, shift all traffic to the winner and turn the weaker variant off.

What if both variants perform almost the same?

That means the element you tested isn't a major lever for this funnel — stop testing it and move on to the next variable, like the domain or the pre-lander, instead of the creative.

How to Split-Test PWA Funnels

The same funnel can perform completely differently depending on the pre-lander, domain, or push copy — and there's no way to guess the winner by eye. A split test divides traffic between variants and shows which one actually drives deposits. In APEX, this is a built-in feature with sticky per-visitor logic.

What to Split in a Funnel

You can split traffic on any single funnel element — the creative, the pre-lander (spin-the-wheel or quiz), the push copy, the landing domain, or the offer itself. Change one variable per test, otherwise you won't know what actually moved the needle. New buyers get the most signal from testing pre-landers and creatives first, since they usually drive the biggest swings in opt-in and deposit rates.

Judge the Winner by Deposits, Not Clicks

CTR and push opt-ins only measure top-of-funnel interest, not whether the traffic actually pays. The real winner is decided by deposits and ROI, which arrive later via postbacks from the ad network or affiliate program. Don't call a split after the first few hours — wait until each variant has accumulated enough deposit conversions to mean something.

Sticky Logic: Why a Visitor Can't Bounce Between Variants

If the same user sees variant A on one visit and variant B on the next, your data turns to noise — the path to deposit gets blurred across variants. APEX splits are sticky per visitor: the system locks in which variant a given user saw the first time and always serves that same variant afterward. This matters even more for push — a subscriber should keep seeing the funnel they originally subscribed to.

Setting Up a Split Test in APEX

In the dashboard, create your funnel and add a second (or third) variant — a creative, pre-lander, or domain. Set the traffic split, 50/50 to start is standard, or 70/30 if one variant is already a proven performer. Connect S2S postbacks from your affiliate network or Keitaro so APEX tracks deposits per variant, not just clicks and opt-ins. Then leave it alone and watch the per-variant stats until you have a meaningful sample.

Common Split-Testing Mistakes

The biggest mistake is running 4-5 variants on a small budget — traffic gets spread too thin for any variant to reach significance. The second is killing a test after a couple of hours based on early clicks, before deposits have had time to come in. The third is changing multiple variables at once, which makes it impossible to tell what worked. Stick to 2-3 variants and change one element per test.

FAQ

How much traffic does each variant need before the split is reliable?: Watch deposit counts, not click counts — until a variant has at least a few dozen deposit conversions, it's too early to call a winner. On a small GEO that can take a couple of days.
Can I split-test push copy, not just pre-landers?: Yes — APEX lets you test push message variants the same way as creatives; the base segment splits sticky, so each subscriber consistently gets the same copy variant.
When should I kill the losing variant?: Once both variants have a comparable sample size and the ROI gap holds steady for several days in a row, shift all traffic to the winner and turn the weaker variant off.
What if both variants perform almost the same?: That means the element you tested isn't a major lever for this funnel — stop testing it and move on to the next variable, like the domain or the pre-lander, instead of the creative.

Build your PWA in APEX in minutes

Cloaking, anti-bot, push, split tests and your own domains — in one service.

Get started free