A/B tests

You will build one ad link that splits visitors between funnel variants, and see which of them leads to payment more often.

Before you start

The Pro or Business plan. On the free plan the section is closed entirely. What paid plans open — Plan and limits.

A free domain. A test takes up a whole domain. Eligible: an “Active” project domain — bought or your own — with none of these:

• a funnel, quiz or paywall, even unpublished. In “Domains” its “Usage” column shows that name or “Held by a route”; free it with “Unbind” (Domains) or by deleting the funnel;
• another A/B test, including a draft or a paused test.

The “Usage” column does not show tests. If a domain is empty there but missing from the choice, a test holds it: the “A/B tests” list shows addresses only for published tests, so for drafts and paused ones open ⋮ → “Domain”. Delete the extra test or change its domain.

No free domain — connect another subdomain of your own in “Domains”, say ab.yoursite.com: how — Publish a funnel.

Published funnels that you compare. A draft will not appear in the choice.

Step 1. Create the test

Open “A/B tests” and click “Create an A/B test”. On the “Details” step, enter a name and pick a domain. With no free domain, the window says “No active domains” — even when active ones exist but are taken; what to check is in “Before you start”.

“Next” saves the test as a draft, and it takes the domain at once. Changed your mind — delete the draft in the test list, or the domain stays taken.

Step 2. Describe the variants

Go to the “Directions” step. Each row is one variant that traffic is split between.

• “Add a funnel” — a whole published funnel. Start here: you compare what is already built and working.
• “Add a pair” — a separate quiz and a separate paywall.
• “Add a URL” — an external address, for example your old landing page.

If the project has “On your own” selected, the first two buttons are replaced by a single “Add a quiz”, and “Add a URL” stays. A test built before the project switched to “On your own” loses its pair and whole-funnel directions in the edit window: they are hidden, and the first save deletes them with their statistics.

Give each row a weight in percent — its share of traffic. “Equalize” splits it evenly. The weights must add up to exactly 100, or the test will not publish; the current sum is shown next to the buttons.

Optional: each row can narrow its countries and operating system. A visitor who matches no direction goes to the app store.

So the test also works as one link for all your ads: run one campaign worldwide and split people in the rows — for example, iPhone to one funnel, Android to another; give no row to countries where you do not sell, and those visitors go to the app store.

Step 3. Publish and give the address to your ads

In the test row, click ⋮ → “Publish”. The address appears in the “Public URL” column — copy it and put it into your ad account. For Meta and TikTok to see payments from this address, connect them — Ad networks and pixels.

Add campaign tags to the address as usual: https://ab.yoursite.com/?utm_source=facebook&utm_campaign=june. Subster remembers the tags for the visitor, so a tag reaches even a payment a week later. The funnel language is set the same way, with the lang parameter, for example &lang=es — Translate a funnel.

The breakdown by tags is in “Analytics”, not here: it shows utm_source, utm_medium and utm_campaign. utm_term and utm_content are only in the “Leads” export, so put campaign differences into the first three.

Step 4. Read the result

Click “+” to the left of the test name — a breakdown by directions for the last 30 days opens.

Below the table, a sentence says whether the difference can be trusted. Three kinds: a leader is named with a confidence in percent · “Too little data so far” and how many more visitors each direction needs · “The directions perform the same”. Until a leader is named, it is too early to declare a winner — even if the gap in the share of payers looks convincing.

A click on a row opens “Analytics” filtered by this test: there you set any period and group by UTM source.

Changes on the fly

Directions and weights can change at any time: ⋮ → “Edit”, then “Save”. Two consequences that the interface does not spell out:

• a visitor who once landed in a direction stays in it forever — new shares work only on new people;
• a deleted direction disappears from the report together with all its statistics. To stop a variant, do not delete the row: set its weight to 0 and hand those percent to the others.

Next

• Funnel analytics — at which question visitors leave and which ads pay off.
• Emails to customers — how to bring back those who reached payment and did not pay.
• Store — who pays you and how much they bring.
• Ad networks and pixels — so that Meta and TikTok see sign-ups and payments from this address and learn from them.