Xavier, CEO and co-founder of Ramdam, breaks down how subscription apps can scale creator ads on TikTok and Meta, why volume beats perfection in UGC testing, and where AI-generated video actually makes sense (and where it doesn't).
Xavier spent five years at Match Group working on AI teams after his dating app was acquired. He then launched an app studio and discovered firsthand how painful it was to find winning ad creatives: months of testing 50 different videos just to find one that cut his cost per install by 5x. That frustration became Ramdam, a platform that helps consumer apps produce creator ads at scale. The company now works with Tinder, PhotoRoom, Flo, and other category leaders, delivering over 10,000 creatives per month.
What you'll learn:
Xavier spent five years at Match Group working on AI teams after his dating app was acquired. He then launched an app studio and discovered firsthand how painful it was to find winning ad creatives: months of testing 50 different videos just to find one that cut his cost per install by 5x. That frustration became Ramdam, a platform that helps consumer apps produce creator ads at scale. The company now works with Tinder, PhotoRoom, Flo, and other category leaders, delivering over 10,000 creatives per month.
What you'll learn:
- Why a 5% success rate on ads is completely normal (and how to structure campaigns around it)
- How to start a UGC test: 20-40 creators, 4-5 concepts, $20-50K minimum spend
- Why US English ads often perform in non-English speaking markets
- How winning apps keep one narrative from ad to paywall
- Why TikTok carousel ads are massively underrated for dating apps
- How to structure "test" vs "scale" campaigns to measure both CPI and ROAS
- When AI-generated video makes sense: hard-to-source personas, scaling winning concepts
- Why the ad your team wants to reject might get 350 million views
- How Ramdam uses AI to match briefs with creators and QA videos before delivery
- Why "happy accidents" from real creators still outperform AI-perfect execution
Key Takeaways:
Volume always wins over perfection. 50 different creators who don't perfectly match your persona will beat 5 who do. You can't predict which ad will work. Even Xavier, after thousands of campaigns, has no idea which ad will succeed when he sees it. The only strategy that works is testing at scale and following the data.
Winning ads have a 2-3 week lifespan. Ad fatigue is real. If you're scaling on TikTok or Meta, you need to refuel with new creatives every month. The biggest spenders are producing 1,000+ creatives per month to stay ahead of fatigue.
Start broad, then replicate winners. Early briefs should leave room for "happy accidents" where creators interpret the concept in their own style. Once you find a winner, run replicate campaigns: same hook, same narrative structure, but new faces and fresh energy.
The ad-to-paywall story must be consistent. Winners keep one promise throughout the entire journey. If the ad says "sleep better in 7 minutes," that same message should appear on the store page, onboarding, and paywall. Breaks in this narrative kill conversion.
AI video is a complement, not a replacement. AI-generated creators work for hard-to-source personas (high-income demographics, pregnant women, complex scenes). But they can't produce the weird, human moments that go viral. Find winning concepts with humans, then scale variations with AI.
TikTok and Meta behave differently. TikTok rewards short (around 10 seconds), trend-driven content with trending sounds. Meta prefers structured narratives, product demos, 15-30 second videos. Carousels perform well on both, especially for storytelling.
Creator diversity expands reach. Meta and TikTok treat ads with the same creator as nearly identical. Using many different faces helps you reach new audiences. This is why Ramdam assigns one creator per video across their 50K creator network.
One ad can change everything. This business follows power law dynamics, similar to the music industry. Most ads do nothing. A small percentage capture all the budget. One viral hit can transform an app's trajectory overnight.
Links & Resources:
- Ramdam: ramdam.io
- Xavier on LinkedIn: https://www.linkedin.com/in/xavier-de-baillenx/
- Email: xavier@ramdam.io (mention Botsi Podcast for personalized demo)
- I also found the TikTok SwipeWipe video: tiktok.com/@vdanielle22/video/7298313654594800942
Timestamps:
00:00 Intro/Teaser
03:00 Xavier's background: Universal Music to Match Group to Ramdam
05:00 UGC formats explained: Classic, Trends, Carousels
09:30 Ad lifespan and creative fatigue
11:30 Why volume and experimentation beat perfection
15:30 Starting a UGC test: creators, concepts, budget
19:00 Creator diversity and platform algorithms
23:00 Balancing authenticity with replication
26:00 TikTok vs Meta: what works on each
30:00 Connecting ad performance to product funnels
36:00 Structuring test vs scale campaigns
38:00 How Ramdam uses AI for creator matching and QA
43:00 AI-generated video: use cases and limitations
49:30 Marketing fundamentals: clarity and authenticity
51:30 Counterintuitive learnings from UGC