Sizzle AI · B2C · AI learning app
Put $1-$20 of ads behind every organic video; back the one getting clicks
One winner per 20-30 videos; the winner drove hundreds of purchases
Workedpaid-socialugc-creatorsorganic-short-videoexperimentation
What they did
Make many organic creator videos (multiple creators, multiple hook styles), then put the platform minimum behind each one rather than judging by organic views, since good videos sometimes flop on timing or the algorithm. TikTok: ~20 creatives at the $20 minimum each, can switch off after ~$5; Meta: 30-40 creatives at ~$1 each since it is easier to duplicate. Judge within hours on top-of-funnel clicks only: most get 0-2 clicks, one gets 5-20. Keep spending on that one to see whether it was a fluke, then track it down to purchases and make variations (change the first or last second). Mid-virality videos (top 25%) often convert better than the most viral, clickbaity ones. For a $5K starting budget he'd spread it over as many hypotheses as possible, aiming for 100 variations, and compress '500 videos' of learning into a week. Test the same creative across Google, YouTube, Meta and TikTok; one platform can give 2x results.
What happened
Took 20-30 organic videos to find one clearly better than the rest; the winning TikTok (79K likes) drove hundreds of purchases as an ad and became the base for scaled variants.
- Stage
- growth
- Effort
- Ongoing; 20-40 creatives per test round, platform-minimum budgets
In their words
The Superwall Podcast · Yev Marusenko - How I Grow Viral Apps to 1M+ Downloads (just copy me)“You take all of your organic pieces and just put a few dollars into them. If you're kind of on TikTok, there's like a minimum, all of that, but just put a couple of dollars into each one.”
“You'll have 20 pieces of content, and each of them get 0 or 1 click or 2 clicks, which is kind of bad. It's not a good sample size. But this is where the magic is. One of them will have like 5 or 10 clicks.”
“So, yes, this is actually the approach where 20 creatives, $20 each. Turn it off halfway if you want, but now you have 20 pieces of content. Within hours, you'll know which one is an outlier.”
“Yeah, I even go more extreme on Facebook. You could put lower amounts, and it's just easier to duplicate and scale everything. So, it's the same thing, but I try to have 30, 40.”
“So, instead, think of as many variations as possible. Like 100 if you can.”
Related topics
Get tactics like this every Monday
New tactics from the week's founder interviews, each linked to where it was said.
More from this episode
- Sizzle AI: Find the funnel step losing over half your users and have everyone fix it2-3x more users reaching the next step, vs ~10% gains from new features
- Sizzle AI: Hire creators via JoinBrands, SideShift, Whop and Discord to post on their own accountsA JoinBrands creator's TikTok hit 79K likes and became the top ad
- Sizzle AI: Open with a silent shocked face, demo the app, end on a goofy twist79K likes; this ending beat the creator's other endings
- Sizzle AI: Buy test traffic with your best ad and run 3-4-way paywall tests in daysCost per purchase fell from ~$500 to ~$300 within days
- Sizzle AI: Scale a winning video by testing 10 headlines, then new countries and devicesWinner scaled into hundreds of ads; one-word headline changes made a difference
- Sizzle AI: Switch ads off and on to measure real sales lift; compare every attribution sourceReal purchases usually about double what the ad platform reports
- Sizzle AI: Add columns showing each ad's step-by-step drop-off from install to paywallSurfaces ads and journeys with 10x higher purchase conversion
- Sizzle AI: Test clean, warm and social-proof sign-up screens; break ties with assisted conversionsSocial proof: more purchases in some groups, neutral overall on assisted conversions
- Run 10+ faceless Shorts channels with fake texts, cartoon clips and satisfying videos26M views in 28 days on one channel; over a billion total, but low revenue
- Sizzle AI: Use viral formats from unrelated niches in study-app videos
- Sizzle AI: Get first users from the founder's launch post and personal networkHelped get the first users; no numbers