How do apps scale paid ads without losing money?

By gr0w Research · Updated · From 185 founder interviews · 5 min read

Short answer

In an analysis of 185 founder interviews, 44 founders used Meta ads, and the ones who scaled profitably followed the same loop. They tested many cheap creatives, kept only ads above their break-even ROAS (above 1x, with 2.2–2.5x on Meta for AbMax), and raised budgets about 20% every few days. They also tracked results with their own attribution data, not only the ad platform's numbers. Founders also said most creatives lose money: Boost App Social said 1 in 85 ad videos carried the spend.

44 of 185
founder interviews where Meta ads drove growth at some stage
2.2–2.5x
Meta ROAS AbMax's founder said he ran at before moving creatives to TikTok
20%
maximum budget increase every few days, per two founders who scaled TikTok ads
1 in 85
ad videos that got the spend and worked at Boost App Social

What ROAS should you target before scaling paid ads?

Break-even ROAS, 1x after platform fees, is the floor, and founders who scaled aimed well above it. Benji at Snag tests UGC videos on Meta and keeps them if ROAS is above 1 [1]. Lino Leiden said AbMax's Meta ads ran at 2.2–2.5x ROAS and the same creatives on TikTok reached 4–5x, which he credited to TikTok's younger audience [2]. His co-founder George said a $24 CPA on Meta was close to 2x ROAS for them, and each app has its own break-even CPA [3].

Compute unit economics first. Frederick James subtracts CPA from realized LTV, then Apple's 15% (with the Small Business Program), RevenueCat's 1% and roughly 20% EU VAT [4]. Will, who runs TikTok ads, kills variants above his CPA target of about $16 per subscription against about $30–35 LTV [5]. Paywall consultant Vahe Baghdasaryan suggested picking a price that gives a payback period within about a month [6]. Reflectly shows the risk: it reached about $3–4M ARR, but ad spend matched revenue, so it was close to break-even [7].

What cost per install or trial is acceptable?

There is no universal number. It depends on what a paying user is worth, and founders reported very different figures.

  • Cheap installs: Lotz said Prayer Screen got 30–50 cents per install in the US after he found a winning hook organically [8].
  • Mid-range: Cedric at Pep AI said about $2.50 per download was profitable, and he wanted to stay under $5 [9].
  • Per purchase: George said AbMax's TikTok cost per purchase averaged about $15 [3].

Cost per trial can mislead. Marcus Burke, a Meta ads consultant, said trials from young, low-quality audiences are cheap but convert poorly down-funnel [10]. Judge campaigns on cost per purchase, not trial. Yev at Sizzle AI used a winning ad as steady traffic to run paywall and onboarding experiments, and cost per purchase fell from about $500 to $300 [11]. On Google, Steve at Journable said you need a daily budget of about 10x your target event cost, because the algorithm needs about 10 conversions a day [12].

How many ad creatives should you test?

Test far more than feels comfortable, because most will fail. Alex at Boost App Social said one video in 85 got the spend and worked, and about $800K of his $1.5–1.8M ad spend went to finding such winners [13]. Casper at Payout said maybe 20–30% of ads are profitable [14].

Volumes founders gave:

  • ListKit: almost 20 new creatives a week, 10 static and 10 video [15].
  • Sizzle AI: with a $5K budget, up to 100 variations. On TikTok, run 20 at the $20 minimum and turn each off after $5 [11].
  • Social (app studio): about 4,000 AI video ads and about 100 A/B tests a month [16].

To lower testing cost, Ryan Thorp's team gave every video a $20 test, graduated winners to bigger budgets, and put 90% of spend behind the top one or two evergreens [7]. Some founders post creatives organically first and only boost the ones that hit [7]. Snag uses 50,000 views as the bar for a Meta test [1].

How fast can you raise an ad budget?

About 20% every few days, unless the platform is optimizing to a CPA target. Frederick James said he never raised budgets more than 20% every few days. He then used TikTok's goal-based budget with a target CPA, and spend rose from about $200 to $1K a day, taking him from $5K to $10K a month in about a week [4]. Will raises winners 20% every three days until they cap out [5].

Benji tests a new creative at about $50 a day. If ROAS is above 1, he steps to $100, $200 and $300 a day until ROAS turns negative [1]. He also warned that returns diminish, so doubling spend does not double profit [1].

Going too fast backfires. George said faster scaling raised AbMax's CAC until it evened out, and his unsolved problem was producing enough creatives to scale spend quickly [3]. Burke splits ad sets by media type, format and message, with budgets at ad-set level. Otherwise Meta can put about 80% of spend on one ad [10].

How do you track attribution accurately?

Use your subscription and analytics data next to the ad platform's, and expect gaps. Frederick James set up tracking through RevenueCat before running Apple Search Ads [4]. Steve sent install, paywall view, trial and purchase events with values to Google Ads before spending [12].

Clear 30's founders built a dashboard showing return on ad spend after the Apple tax and trials by channel. They also asked users "how did you hear about us" in onboarding to balance Meta against TikTok spend [17] [17]. Yev compares platform numbers with his own database across several attribution windows [11]. Attribution can still break: George hired a contractor because AbMax's setup was "cooked" once he added clipping campaigns, though ads-only tracking stayed simple [3].

TikTok Spark ads have no download button, so SideShift's founder said CPMs are about all you can measure there. Meta has the button and better attribution [18].

What can go wrong when you scale?

Founders named four risks: platform dependence, account bans, expensive mistakes and thin margins.

  • Bans: Alex said Boost App Social was banned 6–8 times, and it duplicated one ad by mistake and burned $8K [13].
  • Single channel: he said rising CPMs could kill profitability overnight with no organic channel [13].
  • Thin margins: Frederick James made about 30% margin on paid ads only, which he said made the app hard to sell [4].
  • Early losses: a first-timer may burn up to about $1,000 before ads return [4].

The library is thin on exact scaling rules for large budgets. Most numbers here come from apps spending $200 to $10K a day, and all are self-reported.

Paid-ads numbers founders said (self-reported)
Founder / appTarget or resultScaling rule
AbMax (Lino Leiden)Meta 2.2–2.5x ROAS; TikTok 4–5x [2]Ads behind one equity creator's content
Snag (Benji)ROAS above 1 to keep an ad [1]$50/day test, then $100, $200, $300 [1]
Pep AI (Cedric)About $2.50 per download, under $5 [9]Move proven organic winners to Meta
Habit tracker (Frederick James)Target CPA on TikTok [4]20% every few days, then goal-based budget
Journable (Steve)Target CPA, e.g. $15 trial [12]Daily budget of 10x target event cost
Reflectly team (Ryan Thorp)$20 test per video [7]Winners graduate; top 1–2 get 90% of spend
Boost App Social (Alex)1 in 85 videos worked [13]$100/day up to $10K/day

Frequently asked questions

What is a good ROAS for an app?

Above 1x is break-even on revenue, and founders wanted more. AbMax's founder reported 2.2–2.5x on Meta and 4–5x on TikTok [2]. Subtract Apple's cut and other fees before counting it as profit [4].

How much should I spend testing each creative?

Founders spent $5 to $50 per creative. Sizzle AI turns TikTok tests off after $5 [11], Reflectly's team used $10–20 [7], and Snag uses about $50 a day [1].

Should I post creatives organically before running them as ads?

It is the cheaper route. Ryan Thorp said posting organically first saves test spend, while going straight to ads gets answers faster [7]. Pep AI and Prayer Screen moved organic winners to Meta [9] [19].

Is it safe to rely only on paid ads?

Founders said it is risky. Boost App Social had 6–8 bans and exposure to rising CPMs [13], and a paid-only app made about 30% margin [4].

Can cheaper markets make ads profitable?

Yes, in one case. Tomer ran Cal Buddy ads in Israel, where CPMs are far lower than in the US, and reached about $20K a month in four months [20].

Sources

Founder interviews on YouTube; numbers are as the founders said them. How this research is done.

  1. [1]I Built A $30K/Month App: Here's My Exact Process · Snag (Starter Story)
  2. [2]This crazy simple app makes $50K/month · AbMax (Starter Story)
  3. [3]Meet the ex-wrestler making $41k/mo with AI apps · Wrestle AI, Brawler, AbMax (Superwall)
  4. [4]I Built a $10K/Month App in 88 Days (Starter Story)
  5. [5]The Lazy Method: How I Build $10K/Month Apps · Glowly, Overlo (Starter Story)
  6. [6]Our No-Code Apps Made $50M Using THIS Paywall (Superwall)
  7. [7]Meet The Guy Dominating The App Store · Reflectly, Done (Do Habit), The Mindfulness App, Flo (Superwall)
  8. [8]I flipped 4 apps and made $500,000 · Bible Buddy, Magic Music, Toxic Traits, Pray Screen (Starter Story)
  9. [9]I speedran an ai app to $100k in 70 days. · Pep AI (Superwall)
  10. [10]How to Build a $2M/Month App (Copy Me) (Superwall)
  11. [11]How I Grow Viral Apps to 1M+ Downloads (just copy me) · Sizzle AI (Superwall)
  12. [12]I Built a $100K/Month Android App · Journable (Starter Story)
  13. [13]Copy This Facebook Ads Strategy, It'll Blow Up Your App · Boost App Social (Superwall)
  14. [14]The Stupid Simple Content Strategy That's Making Apps Go Viral · Payout (Superwall)
  15. [15]I Built A $1M SaaS In 87 Days · ListKit (Starter Story)
  16. [16]$16M AI CEOs: How to Make $1.3M/Month With AI Apps · Social (app studio): Font Maker, Lipstick, Celebs, Autopaste, Hotspots, WiseSam.ai (Superwall)
  17. [17]I Built a $30k/mo App That Helps You Quit Smoking · Clear 30 (Superwall)
  18. [18]#1 UGC EXPERT Reveals How To Market Consumer Apps · SideShift (Superwall)
  19. [19]He Built a Simple $120k/Yr App, Then Sold It For Multiple 6 Figures! · Prayer Screen, Bible Buddy, Magic Music, Toxic Traits (Superwall)
  20. [20]I Copied A Huge App. Now I Make $80K/Month · Cal Buddy (Starter Story)