What counts as a video ad here (and what does not)
A result counts as a video ad if the ad that produced the number was a video: in-feed, Reels, Shorts, TopView, in-stream or a video built from a catalog template.
Static images, carousels and playable ads are not video ads.
Where a static ad was tested next to a video, the result is tagged with both formats.
Many results here are tagged with a second format as well.
92 of the 219 also count as UGC, Spark or partnership ads, which means the video was made by a creator or ran from a creator's handle.
To see every format together, start from all results.
Results at a glance
| Measure | Count (of 219) |
|---|
| Won / mixed / lost | 218 / 1 / 0 |
| Vendor-reported | 167 |
| Split test or lift study | 30 split, 39 study |
| Before/after or campaign result | 65 before/after, 76 campaign |
We counted the last two rows from each result's metric tags.
ROAS (49) and CPA (48) are the most common metrics, followed by CTR (32) and ad recall (22).
Close to a quarter of the hub never reports a cost or sales number, and many of those are brand-lift studies.
The smallest examples are single ads from practitioners.
One new Meta video showed a Cost per purchase $8.43, ROAS 4.50 (@EcomRedneck), but that was a partial-day read.
A B-roll mashup cut from old footage ran at a CPA $58.61 on $107k spend (@alexgoughcooper).
On YouTube, @runtraffic reported AI-made video ads at Conv. value/cost 1.58 (+0.21), and an account that added Spend +$682k, ROAS 0.76 → 1.51. Both are platform-reported numbers with no control.
Head-to-head against other formats
Very few results here put video directly against another format.
The cleanest one: talabat's catalog video templates on TikTok cut CPA −12%, CTR +42% vs top static ad (Smartly, vendor-reported).
Genomelink reported 73% lower customer acquisition cost after moving from static and carousel ads to creator Spark Ads (inBeat, before/after).
Video did not always win. Triumph Arcade found playable ads beat video-only, at +85% Day-0 ROAS vs video-only (TikTok for Business, vendor-reported).
And 14 results ran static and video together, so they cannot say which format did the work.
For the full comparison, see static vs video ads.
Best results by platform
TikTok shows up in 125 results, Meta in 51, YouTube in 47 and Google's other networks in 13 (some results cover more than one).
The platforms differ in how much of their evidence is vendor-reported: 108 of the 125 TikTok results, 37 of 47 for YouTube, and only 21 of 51 for Meta.
Most of the Meta results come from agencies and practitioners, so they are mostly before/after and campaign numbers (45 of 51).
For more, see facebook video ad examples. On TikTok, Toyota Germany reported 38% lower CPA from Automotive Ads than from regular conversion campaigns.
See tiktok video ad examples for the rest.
YouTube is the most study-heavy platform (16 of 47 are lift or MMM studies), so its results lean toward ad recall and incrementality rather than ROAS.
By industry
Ecommerce has the most results (45), followed by apparel (26), food and beverage (25), beauty (21), apps (15), home (14) and finance (13).
Industry changes what gets measured. Many food and beverage results are lift studies (10 of 25), such as ad recall lifts for Magnum, Lay's and Somersby.
Apps report CPI and cost per sign-up. Finance results include YouTube-vs-TV comparisons and app-install lift studies.
If you sell direct to consumers and need a CPA or ROAS answer, the ecommerce, apparel, beauty and apps results are the most useful.
Many of the food and beverage results answer a different question.
What the winning ones had in common
- Native, low-cost production. In scandiweb's test for Baker Creek, a catalog-page video made in 10 minutes got 233 purchases at 1.32 ROAS vs 1 purchase at 0.29 for the client agency's video.
- Creators. 92 results use UGC, Spark or partnership ads. Admiral Media's KaufDA test reported CPI −18% for creator content against in-house ads.
- Volume. 46 results tag creative volume as a lever. Pilothouse's supplement client tested 1,000+ assets from 36 creators, and CAC fell 29%.
- Length and hooks are worth testing. TUI's 10-second template beat the 6-second one with 36% lower cost per purchase (Smartly, vendor-reported). 20 results tag hooks as a lever.
The one mixed result is a warning about what you optimise for.
In Wolt's YouTube Shorts test, delayed branding gave a 15% lower CPV, while early branding gave the 5.86% consideration lift.
The cheaper views did not come with the consideration lift, so decide which metric matters before you test.
How to make or source video ads (tools, creators, agencies)
The winners came from three sources: creator marketplaces (16 results credit Billo), agencies running creator programs (Flighted, Pilothouse, Admiral Media, inBeat), and in-house or template edits (Smartly's catalog templates).
Before you brief anyone, use swipe libraries to see which hooks and formats competitors keep running.
For quick variants, AI ad generators are an option.
Nine results here are tagged AI ads, and the @runtraffic YouTube account above is one of them. To find a specialist instead:
Caveats: vendor-reported results
Three in four results here are vendor-reported. TikTok for Business alone supplied 98 and Think with Google 37. Platforms publish success stories, not failures, and that is a big reason the hub shows zero losses.
Read how often vendor numbers overstate before you plan budget around a headline lift.
Video probably won't help when your offer or landing page is the bottleneck, when you can only afford one or two videos (most winners tested many), or when your goal is a sale and the evidence is a recall study.
To test it fairly, run video and your best static ad in the same ad set with the same budget.
Then judge on cost per purchase over at least a week, not on CTR or hook rate.