A campaign can bring 10,000 cheap installs and zero paying subscribers. I judge campaigns on blended ROAS and payback against real revenue, checked against both the ad platform and your MMP, because the two rarely agree.
This page is for a subscription app whose paid channel has stalled, or has never been built properly. It explains what makes subscription different from every other kind of app, what has to be fixed upstream of the campaign, and what happened on two accounts where that was done: Videa and EatBetter.
What is different about subscription
The purchase happens days after the install, usually behind a trial, and after ATT the ad platform sees only a slice of it. Trial starts come back. Renewals and cancellations do not. Left alone, Meta optimizes on half the data and finds you the users most likely to start a trial, which is a different population from the users most likely to pay. That gap is why D0 ROAS can fall while CPI stays flat.
So the work starts upstream of the campaign: a purchase signal the platform can actually optimize on, predicted value in the bid where the payback window is long, and a ROAS framework set by country so budget follows payback rather than cheap installs. That layer is its own discipline, and at Videa it ran country by country and took D0 ROAS from 20% to 43% while spend scaled past $200K a month.
Where creative comes in
Once the signal is clean, creative is the lever. EatBetter went from $10K to $180K in monthly recurring revenue in two months by turning influencer content into paid acquisition on Meta, with the paywall tuned alongside it. Blended ROAS went from 70% to 230% in the same window, and a $2M user acquisition financing round from Leus followed in July 2026, underwritten on the account. How I make UGC ads for apps is the production side of that.
The questions that come up
Whether to remove the free trial, which I have argued both ways with numbers. Whether to trust Meta, RevenueCat or Adjust when they disagree. Whether your iOS organic is actually paid Meta traffic. Each of those is a post, because each one deserved the numbers.
Who it is for
- Subscription apps with live revenue and a paid budget around $100K a month, or funded to reach it
- Teams whose Meta spend has plateaued on campaigns optimized for trials
- Apps where the ad platform, RevenueCat and the MMP each report a different number
- Founders who want the person reading the data to be the person running the campaigns
What you get
- Meta, Google App Campaigns, Apple Search Ads and TikTok, run daily
- A purchase signal, and predicted value in the bid, that the platforms can optimize on
- A ROAS framework set by country, so budget follows payback
- Creative testing with a steady pipeline of new concepts
- Weekly revenue, ROAS and payback against target
Questions people ask
Which channel should a subscription app start on?
Meta, almost always, because creative testing moves fastest there. I open Google, Apple Search Ads and TikTok once the funnel holds and the numbers are stable. The channel comparison has real D28 ROAS from one app across all four.
What ROAS should I expect?
The numbers from my own accounts are on the benchmarks page, each linked to its case. I do not quote industry averages because I cannot verify them.
Should I remove the free trial?
It depends on what the trial is doing to payback and to the signal. Here is what happened on an account that removed it, and the honest answer is that it is a test rather than a rule.
How do you handle the gap between Meta, RevenueCat and Adjust?
By reconciling them rather than picking one. Each measures something different, and the gap between them is itself a diagnostic. Which number to trust for which decision is written out.