Practitioner pattern · Thomas Petit (Thomas Petit client that requires 2x immediate return on paid, Thomas Petit marketplace client with lopsided categories, VC-backed startup where Thomas first ran the whole budget)
Test very different ad ideas first, then make variations of whichever gets traction
A rare winner can be 100x and change a company's course
Workedpaid-socialexperimentation
What they did
Don't test 20 variants of one concept. Test radically different concepts, even unpolished: person vs animation, provocative vs bland, strong colours, lots of text, long vs short, open on the app vs on a person. Put ~20 ads in and let the platform pick: the ad getting most spend after 100-200 impressions is the winner (right ~99% of the time). Where a concept gets traction, iterate: reorder scenes, lengthen, swap the person, change one element at a time. Keep off-brand winners running but try to realign them with brand messaging. Don't trust intuition about which ad will win.
What happened
Real winners are game changing (10x-100x); one off-brand ad brought twice the money.
In their words
Sub Club by RevenueCat · What it Takes to Succeed with Paid User Acquisition — Thomas Petit, App Growth Consultant“One of the most common mistake in creative is people are testing 20 variants of the same concept. Like they're testing over and over the same thing with small variants. You have to test radically different things.”
“Wherever we get a bit of traction, this is where we're going to add layers of let's do variants of this.”
“Like the creative winner is the one that's getting all the spend and”
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
- Only run ads that return at least 2x the spend right awayHit 2x on Apple Search Ads, but only up to a limited scale
- Buy buyers where sellers pile up, and sellers where buyers waitLooked expensive, but kept the other side from leaving
- Buy 500 new users a day early to test onboarding, paywall and ads
- Spend on ads if they pay back by day 366, counting first renewals
- Don't start Meta, TikTok or Google ads below $10-20K a month per platformBelow that, campaigns can't learn; $20 a day is money lost
- Before buying ads, check your goal, cash you can lose, and revenue per installUnder $1 per US install on a subscription app, paid won't work
- With only $5-10K, test messages with small Google search and Apple Search Ads buysRarely profitable, but shows which keywords and messages make people tick
- Start with each ad channel's free tracking; add an MMP once you'll run several
- Watch how total installs, trials and subscribers shift as spend changes, per countryCatches bad trials and losing segments that ad dashboards miss
- Get ads from in-house designers, agencies, many creators and even the CEO's selfie videosCreator-made content has been winning on paid social for 1-2 years
- Hook viewers with the topic they care about, then make clear it's an appToo much clickbait crushes install rate and the platform stops delivering
- Monkey Taps: Run a screenshot of one of your app's notifications as the adBest ad of all time for the app; didn't work when Thomas copied it elsewhere
- Blinkist: Hook ads on a famous CEO's reading habit, then pitch the appAmong the app's most successful ads, per Thomas