How to split test the destination of a Meta ad (not just the hook)

If anything else differs, you're testing two things at once and can't tell which one moved the number.

1. Pick an ad set that already sells. It needs enough purchases a week to read a difference. A brand-new ad set adds noise from Meta's learning phase.

2. Duplicate it. In Ads Manager, duplicate the ad set with its ads. In the copy, change only the website URL to your Product Expert's ad link, pasted whole with Meta's URL parameters ({{campaign.id}}, {{ad.id}} and the rest) left in place. Don't run it through a link shortener: Meta fills those parameters in on each real click.

3. Keep your pixel as the dataset. Conversion location stays Website, the dataset stays your brand's own Meta pixel, and the conversion event stays Purchase. Ads Manager may warn that it can't find your pixel on the website you entered. That's expected, because the chat runs on getclickless.ai. The Product Expert passes Meta's click id to your store with the shopper, so your pixel records the purchase on your store.

4. Create the A/B test. In Ads Manager, open Experiments, choose A/B test, and pick the two ad sets. Meta splits the audience at random so each person sees ads from only one version (Meta Business Help Center, About A/B testing). Set cost per purchase as the key metric.

5. Run it 7 days and leave it alone. Seven days matches Meta's default 7-day click window. Changing the budget mid-test resets Meta's learning and muddies the read.

6. Read both columns, counted the same way. Compare cost per purchase on each side by Meta's count, on the same attribution setting, and by the orders confirmed in your store. Once you connect your Meta ad account (read only), Clickless's report shows Meta's numbers beside the confirmed orders for each ad.

7. Decide on Meta's confidence, not the first two days. Meta's A/B report shows how likely you'd get the same result if you ran the test again. If that's low, call it a tie and run it again or longer.

8. Then test inside the destination. Once a Product Expert holds its own, test what it does: the ad's first line, its deal, and how the chat opens (chat, buttons or a quiz first), split 50/50 inside the Product Expert.

Meta's default attribution setting for website purchases counts a purchase within 7 days of a click on the ad's link, and also some purchases within a day of a view or an engagement such as a like or comment (Jon Loomer's summary of Meta's March 2026 changes, Mar. 10, 2026; check the setting in your own account). Both sides of your test use the same setting, so Meta's comparison is fair.

Clickless counts differently on purpose: a sale within 7 days of the ad click by default (28 or 60 days if you choose), and never a view without a click. A sale after the window is shown in the report but not billed. That's why the report puts Meta's count and the confirmed orders side by side, ad by ad.

More than a handful on each side. With a few purchases a side, one order can flip the result. If your ad set can't produce a reasonable number of purchases in a week, run the test longer or on a bigger ad set rather than calling it early.

For scale, here is one brand's read. It wasn't a 50/50 test, so treat it as an example of what to compare, not a benchmark: in a two-week test inside one Meta campaign, the ad sets that opened its Product Expert ran at $100 per purchase on 24 purchases, against $203 on 118 purchases for the campaign's other ad sets (one brand affiliated with Clickless; Meta-reported, Meta's default attribution; campaign report window Aug 23 to Oct 24, 2026). The full numbers and caveats are on the results page.

One brand affiliated with Clickless, with a $400 product: a two-week test inside the same Meta campaign (report window Aug 23–Oct 24, 2026). Meta-reported, Meta's default attribution; 24 vs 118 purchases. Not a randomized test. Results vary.

A Meta A/B test answers "landing page or Product Expert?" The next questions are about the Product Expert itself, and each one is a setting, per ad:

The report calls a side ahead only once enough people have seen both. Every ad gets its own line in the per-ad report, with chats, visits to your site, orders and revenue.

• Changing the creative too. Then you're testing creative, not destination.

• Comparing different weeks. A landing page's September against a chat's October isn't a test. Run both sides on the same dates.

• Mixing attribution settings. Compare Meta's number with Meta's number on the same setting, and confirmed orders with confirmed orders.

• Comparing add-to-carts. The Product Expert's handoff places the item in the cart, so its add-to-carts aren't the same act as a product page's. Judge on purchases.

• Editing the URL parameters. If a real click shows {{ad.id}} unfilled, the parameters were changed when the link was pasted. Copy the link again and paste it whole. (Clicks from Meta's ad preview show it too; Clickless leaves those out of the report.)

• Judging on day two. Late purchases from early clicks land through day 7.

How long should a destination test run on Meta?

At least 7 days, the length of Meta's default click window, so purchases from the last clicks have time to land. Don't change the budget during the test.

Can I test a landing page against a chat in Meta's A/B test?

Yes. Duplicate the ad set, change only the website URL to the chat's ad link, and pick both ad sets in Meta's A/B test. Meta splits the audience so each person sees only one version.

Will my Meta pixel still see purchases if the ad opens a chat?

Yes, on your store. The chat runs on getclickless.ai, so Ads Manager may warn that it can't find your pixel on that site. The Product Expert passes Meta's click id to your store with the shopper, and your pixel records the purchase there.

Which number should I compare?

Cost per purchase on both sides, read the same way: Meta's purchases on the same attribution setting, and the orders confirmed in your store. Add-to-carts aren't comparable, because the Product Expert's handoff places the item in the cart.

What if the test is close?

Treat it as a tie and keep testing. A handful of purchases can swing either way, so wait for Meta's confidence reading before moving budget.

Media buyers can plan a test and get a partner link at getclickless.ai/media-buyers. Bring a brand and earn a share of Clickless's fee on it, out of Clickless's share; the brand's bill doesn't change. Brands pay no Clickless fee on sales in their Product Expert's first 7 days live, so a one-week test costs only the ad spend.

Related: Run an ad to your agent (docs) · Ads to a chat vs a landing page · Clickless results · Glossary: 50/50 test, 7-day click · AI product quiz for Shopify

Meta is a trademark of Meta Platforms, Inc. Clickless isn't affiliated with Meta.