Affiliate landing-page split testing: track sales, not just clicks

An affiliate landing-page split test compares what happens after visitors see each page—not merely which page gets more button clicks. In Tracked, you can put two eligible pages behind one campaign route, set their traffic weights, and inspect the page, offer click, and returned conversion in the campaign journey.

The setup below is a manual A/B test. Tracked supports weighted routing; it does not automatically declare a winner or change the weights for you.

For an advertorial or prelander specifically, the prelander tracking guide shows how to connect its CTA and downstream sale before comparing page variations.

For Facebook or Instagram traffic, start with the landing-page pre-spend checklist. The Facebook affiliate campaign guide covers the offer, ad promise, and conversion evidence around this page test.

Choose one page change and one primary outcome

Start with a question you can answer. For example: does a shorter explanation before the offer button produce more approved sales per eligible visit than the current page?

Create page A as the control and page B with that one change. Keep the offer, audience, ad creative, and callback setup as stable as practical. If you change the page and the offer together, a difference in results will not identify which change helped.

Define the outcome before traffic starts. A paid sale, approved lead, or another verified event is stronger evidence than a button click when that is what the campaign earns from. Record the test start time and the conversion-delay window; clicks from the last few hours may not have had time to convert.

If you operate the store rather than an affiliate offer, follow the ecommerce tracking guide to connect page exposure to authorized checkout evidence. Store-bound clicks do not prove purchases, and affiliate commissions are not the merchant’s order total.

Put both pages behind one campaign route

Import or edit the pages as separate hosted folders, preview each one, and publish the versions you intend to test. The landing-page hosting guide covers the HTML or ZIP import and custom-domain checks.

Configure the campaign route so both published page destinations are eligible, then assign explicit weights. A 50/50 split is a clear starting point when both pages may receive the same traffic. The weights are relative selection probabilities, not a guarantee that every ten visits will divide five and five.

Use the same tracked offer action on each page when the offer is meant to stay constant. For example:

<a href="/click/1">See the offer</a>

That path creates a tracked offer click on the campaign route. It is different from the page visit. Keep the offer destination and its click-ID mapping consistent across A and B. If you need to test offer rotation instead, use the separate weighted offer routing guide.

Test the entire path before buying traffic

Run a fresh visit through each page. A page rendering correctly is only the first check.

Check	Page A	Page B
Correct published page appears	Confirm	Confirm
Main button uses the intended /click/N path	Confirm	Confirm
Outgoing offer receives the real Tracked click ID	Confirm	Confirm
Test postback attaches to that offer click	Confirm	Confirm
Journey shows page, offer, and event together	Confirm	Confirm

An advertiser or network must return the ID from the tracked offer click. A page impression ID, an ad-platform click ID, and a Tracked offer-click ID are not interchangeable. The postback URL guide shows the outbound and callback fields.

If B has a broken button or missing callback, stop the test and repair the path. Do not interpret its lower conversion count as evidence about the page copy.

Read the result without hiding the denominator

Compare the same outcome over the same eligible traffic and time window. Keep visits, offer clicks, verified conversions, payout, and spend visible as separate measurements.

For example, a page may produce more offer clicks but fewer approved conversions. Another may produce fewer conversions but higher payout per conversion. Neither result is captured by button click-through rate alone. When spend is available, compare profit using the campaign's spend allocation method; do not treat missing spend as zero.

Avoid calling a winner after a handful of events. Wait for the planned window and account for delayed callbacks. If traffic mix, offer terms, or page versions changed during the test, note the change before interpreting the result. Tracked's reports show recorded outcomes; they do not supply a universal statistical threshold or prove causation.

Change the page or weights deliberately

If one page has a useful, repeatable advantage, record the evidence and then adjust its weight. Keep a rollback path. If the result is unclear, keep the split, refine the hypothesis, or run a new test with one focused change.

Tracked keeps the page files, route, attributed journey, and report in the same workspace. That lets you inspect a result and work on the page files without losing sight of the campaign path that produced it. Start free — early access. Use your invite or request access, then begin with one controlled page comparison.

Frequently asked questions

Can Tracked split traffic between two landing pages?

Yes. Put eligible destinations behind a campaign route and configure their weights. Test both routes and the conversion callback before sending paid traffic.

Does Tracked automatically send more traffic to the winning page?

No. Current weighted routing follows weights you configure. Automatic winner selection is a separate capability in development.

Should I judge the test by page clicks or conversions?

Use the event that matches the business goal. Offer clicks help diagnose the page, but an approved lead or paid sale is usually the stronger outcome when that is how the campaign earns.

What if conversions arrive days after the click?

Keep the test window open long enough for those callbacks to arrive. Compare cohorts at a similar maturity, and do not treat recent visits with incomplete outcomes as losses.