--- title: Testing campaigns vs evergreen campaigns for app ads on Meta — forcing spend and graduating winners slug: meta-app-creative-testing-and-graduation summary: How subscription-app media buyers separate creative testing from evergreen campaigns on Meta — Samet Durgun's install-campaign funnel (200 ads on Android → 10 on iOS → 2 hero creatives), automated rules that cap spend per ad so every ad gets tested, and graduating only on cost per purchase after the trial window. domain: mobile-apps tags: [user-acquisition, meta-ads, creative-testing, campaign-structure, advantage-plus, learning-phase] related: [paid-social-creative-testing-framework, meta-qualified-trial-optimization-event, meta-ads-budget-math-for-apps, ad-creative-volume-and-hit-rate, meta-creative-testing-by-budget, meta-account-structure-2026, meta-scaling-winners, meta-andromeda-gem-lattice] sources: [yt:kXXFtl0ocU0, yt:8SfXEIFdB2M, yt:QdIGi-tKr28] updated: 2026-10-24 --- # Testing vs evergreen campaigns for app ads on Meta ## TL;DR - Keep **testing** and **evergreen** campaigns separate. Testing rules protect budget. Evergreen rules protect the learning phase (~50 conversions/week **per ad set**).[^1][^2] - **Samet Durgun's funnel (Dec 2025):** test lots of creatives in a manual **Android install campaign** → put the ~10 that take most spend into an **iOS install campaign** → the ~2 that take 60%+ of spend become "hero creatives" for a trial/purchase campaign.[^3] - In 2025–26 (Andromeda), many ads in an ad set get **zero spend**. Fix it with automated rules that pause each ad after a fixed spend. Samet uses ~$2 per ad on install tests; Anas Maqsood uses $200–500 depending on CPA.[^4][^5] - Samet: in an install test, **spend share is the KPI**. Trust where Meta puts the money over a lucky low CPI.[^6] - Anas: **graduate an ad only on cost per purchase**, after waiting out the trial window (e.g. 7 days), never on hook rate, hold rate or even cost per trial.[^5][^7] - Don't dilute: 3–4 campaigns × 5–6 ad sets on a small budget starves every ad set, and SKAN too.[^2][^8] ## Details **Samet's three-phase structure.**[^3][^9] 1. **Make many original concepts**: shoot yourself, ask people, use AI tools. Run them in a **manual** install campaign on **Android** (cheap), with broad targeting and no audience limits.[^10] 2. Of ~200 creatives, maybe 10 rise. Retest those in an **iOS install** campaign. Maybe 2 take most of the spend. 3. Move the hero creatives into a conversion campaign (trial or purchase) "you're not afraid to spend money on". This shows the app's real potential. - Bundle **3–5 ads per ad set** for manageability, or up to 10 if you manually stop leaders so others get exposure.[^6] - Don't let **Advantage+** pick creatives from install tests. Use it only after you have tested ~100 (ideally ~1,000) ads.[^10] - If Meta still fails with good creatives, look at pricing, onboarding and value. Meta is "a good mirror".[^9] **Samet at 100K+/month (Sep 2026).** The same idea with an extra gate: **install campaign** as first filter → **validation campaign** for up to ~7 days, so a lucky first-day ROAS doesn't promote an ad → **main campaign**. He also tracks **cost per winner** (what you pay to find each new winning creative) and prepares the next winners before current ones fatigue ([[scaling-meta-past-100k-for-subscription-apps]]).[^11][^12] **Anas's rules (Sep 2026).**[^5][^7] - With ~30–40 ads in one ad set (a common post-Andromeda habit), Meta concentrates spend on a few. Automated rules give each ad a fixed spend, then pause it. - After the test stops, wait the trial length to collect late conversions. Then compare ads across the **full funnel**, using MMP numbers for trials *and* purchases. Promote winners to the evergreen campaign. - An ad with a good cost per trial but a high cost per purchase must not graduate. ## Disagreements / caveats - **Install-optimised testing vs purchase-optimised testing.** Samet tests on install campaigns and judges by spend share. Anas tests on the business event and judges by cost per purchase. Steve P. Young warns that cheap installs often never pay ([[meta-ads-budget-math-for-apps]]). Samet's answer is to move only hero creatives into conversion campaigns, so install tests act as a first filter. - Samet's ~$2/ad and Anas's $200–500/ad differ by 100×. The first is a reach check on install campaigns, the second a conversion test. Size yours to your CPA. - Forcing spend overrides Meta's allocation, which some practitioners discourage. Compare [[meta-creative-testing-by-budget]] (separate ad sets per concept). - Samet: "this requires a lot of experience… take it with a grain of salt."[^10] ## See also - [[paid-social-creative-testing-framework]] — Ivy's 3-day tests and hook/hold targets - [[ad-creative-volume-and-hit-rate]] — why you need hundreds of creatives - [[meta-scaling-winners]] — what to do after graduation - [[people/samet-durgun]], [[people/anas-maqsood]] [^1]: yt:8SfXEIFdB2M @ 15:31 — Anas Maqsood [^2]: yt:8SfXEIFdB2M @ 17:03 — Anas Maqsood [^3]: yt:kXXFtl0ocU0 @ 08:30 — Samet Durgun [^4]: yt:kXXFtl0ocU0 @ 10:35 — Samet Durgun [^5]: yt:8SfXEIFdB2M @ 20:15 — Anas Maqsood [^6]: yt:kXXFtl0ocU0 @ 11:24 — Samet Durgun [^7]: yt:8SfXEIFdB2M @ 18:39 — Anas Maqsood [^8]: yt:8SfXEIFdB2M @ 26:33 — Anas Maqsood [^9]: yt:kXXFtl0ocU0 @ 09:43 — Samet Durgun [^10]: yt:kXXFtl0ocU0 @ 16:45 — Samet Durgun [^11]: yt:QdIGi-tKr28 @ 27:28 — Samet Durgun [^12]: yt:QdIGi-tKr28 @ 28:16 — Samet Durgun