UGC Handbook / Lesson 13 of 14 / Chapter IV: Measure & scale

Amplify winners: recut, don't repost

Coconote had organic videos doing 30 million views. The founders assumed the obvious: if a video converts that well for free, put ad money behind it and it will convert just as well.

It didn't. On the Sub Club podcast, co-founder Zack Hargett named it as the thing that surprised him: "we always thought you could just pour fuel on the fire."

This lesson covers why the raw repost fizzles, what to do to a winner before you pay to push it, and the order that keeps paid spend from burning your winners.

Pouring fuel on the fire (and why it fizzles)

Coconote took raw, unedited footage from the same content team that posted the organic hits and had a clip agency chop it into something "specific to performance marketing."

That recut cohort outperformed their more organic approach as ad creative. Zack's reaction: "I found that really interesting because I always assume you could just pour fuel on the fire."

Two caveats, because the source gives them. Even with recuts, Coconote never reached first-purchase profitability on ads, the bar co-founder Brett Bauman named, and they stopped running paid. Brett also said neither founder gave paid more than half their time.

So the claim is narrower than "recuts win": the version of a video that wins organically and the version that wins as an ad are different objects.

Why the raw repost underperforms. An organic hit earned its views as a post, in a feed where nobody was asked to click. A boosted repost lands as an ad, carrying a click target it was never built to serve. Drew Levan of Sideshift, which runs UGC campaigns for consumer apps, described the organic ask on Superwall as stop, search the app, download, buy, then come back and comment, "a level of buying you don't get from paid."

This handbook's recommendation, not a Coconote finding: keep the hook that proved itself and rebuild everything after it for a viewer who can tap one button.

Don't go paid until you have five winning formats

Dan Kwan, who built Conch AI on viral content, closed his Superwall interview with a rule: "don't touch paid until you've cracked at least five winning formats. You're likely not going to see high ROI on your time." Five formats, not five videos.

Ryan Thorp, whose team the Superwall host introduced as a portfolio of 30 apps with over 50 million downloads, described the same sequencing. Post every video organically on your own channels first. If it picks up traction, boost it with ad spend, and if it really takes off, run it as a Spark Ad from the creator's own account.

Asked how much traction a video needs before boosting, Ryan's answer was "even if it's got like sort of 10K views."

Why the order matters. Ryan's own note: in quieter periods, post everything organically first and only boost what works, instead of paying $10 to $30 a day per creative to find out. For a team with no winners yet, organic is the cheaper test.

The Spark Ads arbitrage. Spark Ads put TikTok money behind the creator's original post from their own account (Meta's equivalent is partnership ads). Drew Levan's rule of thumb at Sideshift: any video above a 10% engagement rate should get a little Spark money, and at 7.5 to 8% engagement he would already test $20 behind it. With high engagement they "generally see way below dollar CPMs on Spark" (CPM is the cost per thousand impressions).

Engagement rate means (likes + comments + saves + shares) ÷ views.

His rules: start at $50 to $100 a day, run a week at a time, and kill the Spark when engagement drops below 5%. And never spark preemptively: "If organic is ripping, hold off as long as possible. Give it like a day or two."

One more caution from him: if you pay view bonuses, agree upfront that sparked views don't count.

Real example. Wasabi, a subscription app growing on TikTok, used Airbridge to read cost per install at the video level and to see which creatives carried users through to the paywall, not just to the install. Optimizing toward that read doubled paywall conversion and lifted monthly revenue run rate 200%. Read the Wasabi case study.

Budget split and rotation once you are paying

A RevenueCat blog post on creative fatigue explains that the algorithm shifts 90% or more of the budget toward just 2 to 3 creatives, and when those get overexposed you see rising frequency, falling click-through rate (CTR), and rising costs. The author warns that pausing your top winners to force spend onto weaker creatives "does not work." It usually makes cost per action worse.

The 70/30 structure. The author's best-case setup is one campaign per country with two ad groups (Meta calls them ad sets, TikTok calls them ad groups). A business-as-usual (BAU) ad group with 70 to 80% of the budget holds the proven winners. A testing ad group with 20 to 30% adds 2 to 3 new creatives every 2 to 3 days, so fresh winners keep flowing into BAU before fatigue hits.

How fast a boosted creative declares itself. A second RevenueCat post on creative testing puts numbers on it: winning creatives typically receive 80 to 95% of daily spend, and a creative that gets less than 50% of the budget after two days should be treated as a likely loser or false positive. Rotate creatives that don't spend every 2 to 3 days.

When a winner fatigues. If you can't produce new assets fast enough, the fatigue post's trick: pause one of the top 2 to 3 assets, let the others absorb spend for about a week, then pause those. The author says this works for up to 2 weeks, once. The real fix is a fresh rotation at least once a week.

Do this before you boost anything

1. Count your winning formats. A format wins when its videos hit 7.5%+ engagement and at least one Start Trial. Fewer than five formats? Back to the batch (Lesson 01).
2. Let it climb. Wait a day or two while organic views are still rising, per Drew Levan.
3. Spark or partnership first. $20 to $100 a day behind the creator's original post for one week. Kill below 5% engagement.
4. Recut in parallel. Keep the opening identical (our recommendation), rebuild the rest for a tap (checklist below).
5. Structure the account. BAU 70 to 80%, testing 20 to 30%. Feed 2 to 3 recuts into testing every 2 to 3 days.
6. Read the two-day verdict. Under 50% of spend after two days is a likely loser. Spend with no Start Trial gets paused too (our rule). Upload the next recut.