A/B Testing for Landing Pages

Category: Marketing  ·  Min plan: Pro  ·  Slug: ab-testing

A/B testing lets you run two or more versions of your landing page on one single link, split the visitors between them automatically, and see which page actually brings more orders. It is built for merchants who pay for traffic: you put one link in your ad, and DZBuild tells you which page deserves that budget.

Video tutorial

Features

• One unique test link that rotates visitors between your landing pages
• Equal split by default, custom weights when you want them
• Visitors, orders, conversion, average order value, confirmation and delivery rate per page
• A clear winner verdict with real statistical confidence
• End the test and the same link sends everyone to the winner — no ad changes needed
• QR code for the test link

What it does

The add-on adds a new A/B Testing page to your dashboard sidebar at /dashboard/ab-tests.

There you pick between 2 and 5 of your published landing pages, give the test a name, and DZBuild generates a single short test link on your own store address:

https://yourstore.dzbuild.app/t/k7m3qptr

If your store runs on a verified custom domain, the test link uses that domain instead.

That link is the only one you share. Every visitor who opens it is sent to one of the pages in the test, in rotation, and everything they do afterwards — the order, the confirmation, the delivery — is credited back to the page they landed on.

Each page in a test gets a letter: A, B, C, D, E.

How the test link works

• Rotation. Visitors are handed out in turn, so with equal weights the split stays even across the whole test, not just on average.
• Weights. By default every page gets the same share. In the advanced options you can give a page a bigger share, from 1 to 100, if you want most of your traffic to keep going to the page you already trust.
• Sticky visitors. Once a visitor has been sent to a page, they keep getting that same page for 30 days. A customer who comes back tomorrow does not see a different version.
• Ad parameters are kept. Everything after the ? in your ad link — fbclid, utm_source and the rest — is passed through to the landing page, so your Pixel and campaign reporting keep working exactly as before. See Marketing → Pixels.
• Bots do not count. Link previews and crawlers (the WhatsApp or Facebook preview when you paste the link, for example) are not counted as visitors and do not consume a rotation slot.
• Speed is untouched. The split happens on the test link itself, before the page, so your landing pages load exactly as fast as they normally do.

tip

Only visitors who came through the test link can produce a counted order. If you also send traffic straight to one of the pages by its normal /landing/... address, that page will show more orders than the test can explain — the results page flags it with a warning so you know the comparison is skewed.

Benefits

• You stop guessing which page sells. Two headlines, two offers, two price displays — the numbers decide, not opinion.
• You never touch your ad again. The same link stays in your campaign from the first day of the test until long after you have picked a winner, so you keep your ad's learning, comments and social proof.
• You compare what really matters in Algeria. Beyond raw conversion, you can decide the winner on confirmation rate or delivery rate — the page that brings orders customers actually confirm and actually receive, not just the page that fills the most forms.
• You know when the result is real. DZBuild refuses to announce a winner until there is enough data, so you do not kill a good page after twenty visitors.
• Nothing is thrown away. Ending or deleting a test never deletes a landing page, and the winning page keeps living at its own address.

How to activate

1. Open Dashboard → Add-ons at /dashboard/addons.
2. Find A/B Testing under Marketing.
3. Click Activate — the add-on requires the Pro plan or higher.
4. A new A/B Testing entry appears in the dashboard sidebar, and the management page opens right away.

warning

You need at least two published landing pages before you can create a test. A landing page that is still a draft appears in the picker but cannot be selected, and neither can a page that is already used in another live test. See Landing Pages.

Using it

Create a test

1. Go to A/B Testing in the sidebar and click New test.
2. Give the test an internal name — your customers never see it.
3. Pick 2 to 5 landing pages. Each one shows its view count, and unavailable pages show why (draft, already in another test, or not available on your plan).
4. Choose what decides the winner:
   • Conversion to order — the page that turns the most visitors into orders. This is the default.
   • Average order value — the page that produces bigger orders.
   • Confirmation rate — the page whose orders customers actually confirm.
   • Delivery rate — the page whose orders actually get delivered.
5. Optionally open Advanced options to change the traffic split, or leave it equal.
6. Click Create test. The test starts running immediately and your test link is shown, with buttons to copy it, open it, or display its QR code.

A store can keep 5 tests live at a time (running or paused) and 50 tests in total.

Read the results

Open a test to reach its results page. At the top you get total visitors, orders and revenue for the whole test, plus a verdict card:

Verdict	What it means
Collecting visitors	Not enough data yet — no page is declared ahead
Results too close	Enough data, but no meaningful gap between the pages
Slightly ahead	One page leads with at least 80% confidence — keep the test running
Ahead	One page wins with at least 95% confidence — you can end the test

Below it, one row per page shows visitors, orders, conversion rate, average order value, revenue, confirmation rate, delivery rate and a 14-day visitors trend. The best-performing page carries a badge.

The page refreshes itself every 60 seconds, and there is a manual refresh button next to the table.

tip

A page with fewer than 30 visitors shows a dash instead of its rates. That is deliberate — one order out of five visitors is not a 20% conversion rate, it is noise.

Before any winner is announced, all of these must be true: the goal's minimum sample on the weakest page (100 visitors for the conversion goal, 15 orders for confirmation, 10 orders for average order value and delivery), 25 orders across the whole test, and 72 hours of runtime. The 3-day floor is there so a good Thursday does not get mistaken for a winning page.

Adjust a running test

• Edit — rename the test or change the traffic weights at any time.
• Pause — the link keeps working but sends every visitor to the first page, and nothing is counted while paused. Resume whenever you want and counting continues where it stopped.
• Remove a variant — stops sending new visitors to that page and redistributes its share to the others. Its recorded results stay in the report. A test must always keep at least two pages.

End the test and keep the link

Click End test, choose the page that everyone should get from now on — the leading page is marked Recommended — and confirm. From that moment:

• the test stops splitting traffic;
• the same test link keeps working and sends every visitor to the winning page;
• your ads, posts, QR codes and bio links need no change at all.

warning

Ending a test cannot be undone. If the results page still says the data is insufficient, ending now may crown the wrong page — the confirmation dialog warns you when that is the case.

Deleting a test removes it from your dashboard and stops the split for good. Your landing pages are never deleted.

FAQ

Do I have to change my ad when the test is over? No. That is the whole point of the test link. You end the test, pick the winner, and the link you already published starts sending everyone to that page.

Will the same customer see a different page every time? No. A visitor is assigned to one page and keeps it for 30 days, so a customer who comes back to finish an order finds the same page, the same offer and the same price.

Why does my test say there is not enough data yet? Because at least one of the three gates is not met: enough visitors or orders on the weakest page, 25 orders across the test, and 72 hours of runtime. Landing-page conversion in Algeria is usually only a few percent, so a fair test needs a few days of real traffic.

Can I put the same landing page in two tests at the same time? No. A page can only belong to one live test, so the numbers stay unambiguous. Finish or delete the first test and the page becomes selectable again.

Do orders that arrive after I end the test still count? Yes. A visitor who was assigned to a page before the end can still order afterwards, and that order is credited to the page they saw. See Orders.

Related

• Landing Pages — build the versions you are going to test
• Pixels — keep Meta and TikTok attribution working on tested pages
• Add-ons overview — the full add-on catalog and plan requirements