Always test similar formats

There are hidden gains to be had from tests that initially look pointless with LI Ads.

Two near-identical formats with the same message and same audience, surprisingly (or frustratingly), can give very different results.

Two of the tests we run a lot:

(1) 𝐂𝐚𝐫𝐨𝐮𝐬𝐞𝐥 𝐚𝐝𝐬 𝐯𝐬. 𝐃𝐨𝐜𝐮𝐦𝐞𝐧𝐭 𝐚𝐝𝐬

Carousel ads have been around since roughly 2017. We've always found it difficult to make them perform for 3 reasons:

→ 1. They appear small on mobile.

→ 2. They're fixed on mobile. You can't zoom in.

→ 3. They tend to have low CTR + engagement

Essentially, it's difficult to make them work for both branding and activation. But for a long time, they were the only option for multi-slide / storytelling ads.

Document Ads came along a few years later and became our go-to choice for storytelling creative.

With a Doc ad, the viewer can scroll through the slides without triggering a click charge. They can zoom in, properly read the content, and only click when they're ready to either 'Download', 'unlock the document' or visit the landing page.

Most importantly, they're also bigger and more interactive on mobile.

If clients want a carousel-type ad, we often test both formats side by side. In most cases and for most metrics, Doc ads win, but it's never certain.

If you're running anything with slides, it's still worth testing both.

But absolutely don't default to carousel.

(2) 𝐂𝐨𝐧𝐯𝐞𝐫𝐬𝐚𝐭𝐢𝐨𝐧 𝐚𝐝𝐬 𝐯𝐬. 𝐌𝐞𝐬𝐬𝐚𝐠𝐞 𝐚𝐝𝐬

Message ads are the evolution of Sponsored InMail. They require a subject line and have CTAs top and bottom. The multiple CTAs make them look more like an ads than convo ads.

Convo ads were built around a decision-tree model. The subject line is optional and the idea was that you create multiple CTAs and could send people down different journeys depending on their click choice.

Over time, we noticed that most people only ever take one click-max on an inbox ad - then the drop-off beyond that is steep, so we stopped building the decision trees.

So we build Convo ads with a single CTA, which makes them functionally very similar to Message ads.

Also, without a subject line, they look more natural in the inbox. Open rates tend to be higher, and they look more like an actual message.

We've seen Message ads outperform Convo ads and vice-versa many times.

There's no consistent pattern.

What we do know is that the differences are often big enough that it's ALWAYS worth testing. Some cases we've had:

→ Test 1: 55% lower CPL on Message vs Convo

→ Test 2: 22% lower CP-SQL on Convo vs Message

→ Test 3: 42% higher Open-Rate on Convo vs Message

It's an easy test to set up and can be well worth it.

We're big advocates of running both and letting the data decide.