A/B testing landing pages - split ad sets or not?

trendspotter

I'm deep in the weeds of A/B testing two landing pages for a client's ads, and I need to see which one actually drives form submissions. The goal is to keep the creative and caption identical - just swap the URL. But here's where I keep hitting a wall: do I duplicate the ad set so I have two separate sets in the same campaign, or just duplicate the individual ads within one ad set?

I used to cram everything into one ad set and the algorithm would pick a favourite in like two hours. Split them and let it run clean. That's what I'm leaning towards now - separate ad sets, same audience, same budget split, no overlap. That way the test feels more controlled, and I'm not letting Meta's whims skew the results based on which page loads first or has a slightly better load speed.

Someone in the thread suggested keeping both ads in the same ad set, only changing the URL, and making sure load speed and tracking are identical - otherwise the difference could be from tech, not design. That makes sense too. But my gut says the algorithm will optimise toward one ad too quickly, drowning out the other before you see any real conversion data.

What's the main difference between the two pages? One is ultra-clean, lots of white space, minimal copy - very editorial, almost like a luxury magazine spread. The other is more direct, bold headlines, trust signals near the top, a stronger call-to-action button colour. Both feel on-brand, but one leans into sophistication, the other into urgency. I want to see which aesthetic actually converts better for this audience.

So - separate ad sets in a CBO campaign to ensure equal budget distribution? Or keep them in one ad set and risk the algorithm picking a winner too fast?

storyweaver

Oh the classic "let's throw different audiences and budgets at the problem" approach - nothing like a completely useless A/B test to burn through your ad spend. Keep both ads in the same ad set and only change the landing page URL. That's how you actually isolate what you're testing. Oh and for the love of sanity, make sure both pages have similar load speeds and tracking set-up. I've seen a "winning" design get slaughtered just because one page loaded 0.5 seconds slower. So what's genuinely different between these two pages? Please tell me it's more than just font colour...

pixelperfect

The key is to duplicate the ad set within one CBO campaign - that way both landing pages get the same budget allocation, making the test clean and fair. I've done this for my own clients and it avoids the headache of one variant eating up more spend.

bigcoffee

Separate ad sets. Non-negotiable.

The algorithm has a nasty habit of picking a winner within the first couple of hours if you bundle everything into one set. That biases the whole test - you're not testing the landing page, you're testing Meta's early preference.

Here's the structure that's held up across dozens of tests for me:

Same audience targeting (exact same parameters)
Identical budget split (50/50 or 33/33/33 if more variants)
Zero audience overlap (use exclusion lists if needed)
Minimum 3-5 days of data before drawing conclusions

Also, watch your cost-per-result threshold. If one variant's CPA is >20% higher than the other inside the first 48 hours, I've found it rarely recovers. Kill it early and reallocate budget.

Let them run clean without touching anything - no audience changes, no bid adjustments. The cleaner the data, the more confident you can be that the difference is the page, not the setup.