Reels-Creatives: What Meta recommends – and how you can test it sensibly
A Reels video is not automatically a good Reels ad. Meta gives concrete design rules for the format. This post translates them into a production and testing list and shows what conclusions the data actually allows.
Metas format specifications for Reels
Meta recommends 9:16 vertical videos, audio, and key elements within the Safe Zone for Reels ads. The Safe Zone takes into account Reels interface controls that may obscure text or calls to action. This particularly applies to subtitles, logos, product demonstrations and the call-to-action.
The meta page also describes the choice between Advantage+ placements and manual placements at the ad group level. There is a placement-specific adjustment for existing creatives. A landscape-format advertising film therefore doesn't just have to be cropped automatically: check in preview mode whether the promise and the next step remain understandable on the actual mobile screen.
Whether a native 9:16 version performs better in your own account than an existing format can only be checked with a suitable comparison. Format recommendations alone do not prove lower costs for a particular offering.
- Produce or export the video in 9:16 format and control every fade in the Reels preview.
- Keep the value proposition, subheadings, and CTA out of surface areas.
- Check sound, subtitles and intelligibility even if users have the sound disabled.
- If you have multiple placements, check the preview for each placement instead of just the original file.
The entry must appeal to the right person
The first few seconds decide whether someone will continue watching. This is a NOSA working hypothesis for the creative test, not a meta universal rule of seconds. A clearly stated problem can qualify for high-priced offers better than an unspecific attention stimulus. For example, formulate who the offer is intended for and what specific problem is being solved.
Test different openings: a direct problem question, a short demonstration, a clear comparison of before and after, or a well-founded thesis. Keep the offer and landing page the same at first so that you don't mix the effect of the entry with a different funnel. Assumptions about “the algorithm” do not replace this cross-check.
Read video numbers without confusing them with sales
For a video test, you look at delivery, video starts, play points achieved, link clicks, leads and - if properly connected - qualified leads and sales. Strong playback is a signal of attention. Whether the video appeals to the right people only becomes clear later in the process.
If a lot of people drop out early, check the first statement and image. If many people continue to see but hardly click, check the offer and CTA. If there are a lot of clicks but inappropriate contacts are generated, examine the ad-form match and qualification questions. This interpretation is a NOSA diagnostic aid; a single measurement does not prove a cause.
Compare playback rates only for similar video lengths and with the same definitions. A 15-second video and a 60-second video have different requirements for 100 percent completion. Even a good ThruPlay result is no substitute for sales data.
A practical A/B test
Meta recommends A/B testing to compare Reels creatives and rankings. To make a useful creative comparison, determine in advance what is different. For example, only change the introduction, while the target group, offer, form and evaluation period remain the same. If the budget or the form is changed at the same time, it is difficult to attribute a difference to the video.
A useful testing protocol records hypothesis, creative versions, campaign, ad group, time period, spend and result definition. For high-priced services, the result definition should also take sales quality into account in addition to the meta lead. Small samples are a signal for further review, not a safe decision for all future campaigns.
- Write down a concrete hypothesis, such as: “A product demo leads to more qualified conversations than the problem question.”
- Create two versions with the same offering and length; vary the start specifically.
- Document spend, reach, play points, link clicks, leads, and qualified leads over the same time period.
- Make decisions based on the previously determined target size; record discrepancies and data gaps in the protocol.
What the meta source proves – and what it doesn’t
The meta pages linked below cover formatting guidelines, placement options and creative practices. Our suggestions for qualification, test protocol and interpretation for high-priced offers are NOSA classification. The post does not report any new algorithm change and does not promise any specific cost reduction.