--- title: Which event should Meta optimise on? Qualified trials and enriched events slug: meta-qualified-trial-optimization-event summary: How subscription apps pick the Meta optimisation event that has both volume and quality — start trial vs purchase budget math, "qualified trials" (trial not cancelled within 1–2 hours), backend-enriched events above ~$100K/month, and how to A/B test events at ad-set level. domain: mobile-apps tags: [user-acquisition, meta-ads, signal-engineering, free-trial, optimization-event] related: [meta-ads-budget-math-for-apps, meta-sdk-event-validation, signal-engineering-and-conversion-event-audit, meta-app-creative-testing-and-graduation, skan-aem-capi-for-app-campaigns, web2app-ad-optimization-events, paywall-by-acquisition-source] sources: [yt:8SfXEIFdB2M, yt:kXXFtl0ocU0, yt:QdIGi-tKr28] updated: 2026-10-24 --- # Which event should Meta optimise on? ## TL;DR - Meta is "obedient": it finds whoever triggers the event you pick. Optimise on start trial and you get trial-starters, including impulsive ones who cancel within the hour.[^1][^2] - The event needs **volume and quality**. Start trial has the volume. Purchase has the quality, but it is often too expensive and too delayed for a young app.[^3] - **Qualified trial**: send the trial event only if it isn't cancelled within ~1–2 hours. That's the default recommendation when you're starting out.[^4][^5] - Above ~**$100K/month**, move to **backend-enriched events**, e.g. a trial plus 3–4 sessions in the first 24 h. Send them server-to-server via the MMP and map them in Meta.[^4] - The event is **locked when an ad set launches**. Test events with two identical ad sets in one campaign that differ only in the event.[^6] - Don't optimise on onboarding completion. Send the most important event within minutes of install.[^7] ## Details **Budget math (Anas Maqsood, Sep 2026).** Meta needs ~50 conversions per week **per ad set**, not per campaign:[^3][^8] - $20 cost per trial → ≥$1,000/ad set/week ≈ $4K/month. - $100 cost per purchase → ≥$5,000/ad set/week. - Samet Durgun's lower floor: at least **5 events a day** per campaign.[^9] By Sep 2026, for 100K+/month spenders, he asks for **10 events a day in every ad set** ([[scaling-meta-past-100k-for-subscription-apps]]).[^14] **Why trials leak.** Anas sees trial→paid of 15–25% typically, 35%+ for very good apps. The fix is less about trial length and more about the event Meta learns from.[^1] Samet's example: if a younger audience gives cheaper trials that don't convert 3–7 days later, "you're burning money".[^10] **Qualified trial definitions in use:** - Adapty sends the trial event after **2 hours** if it wasn't cancelled, and says some clients saw campaign performance improve.[^5] By Sep 2026 its Attribution product's default was **3 hours**, adjustable. Murat Menzilci pitches it for 7-day-trial apps that can't use ROAS optimisation.[^15][^16] - Samet: "trial started and not cancelled within the same hour". If you sell a yearly-with-trial and a weekly-no-trial plan that convert similarly, **bundle** "trial not cancelled in 1 h **+** purchase" into one custom event.[^10] - Waiting the full 7 days for trial→paid is real, but the delay plus purchase-level budgets make it hard for new apps. A trial with a short delay "is the real event", just cleaner.[^4] **What purchase optimisation does to your funnel (Samet's experience).** Purchase campaigns bring far more intent. Organic install→paid of 2% can become **10–20%** on a Meta purchase campaign, and a 10% trial rate can become **~30%** on a trial campaign. CPI usually rises a lot, up to ~**5×** the install-campaign CPI. A few accounts see similar CPIs either way.[^11] **Cheap trials vs valuable trials (Murat Menzilci, Sep 2026).** Two users can both start a trial, one cancelling at once and one converting to annual. Send plain trial events and Meta treats them as the same audience. His slide: campaign A had the cheaper cost per trial (~$8), but campaign B, with dearer trials, had the better LTV and ROI. More signals are not better signals.[^17][^18] The next step up is a **predicted-LTV value** sent with the event, which tells Meta to find valuable users rather than trial-starters. Adapty computes and sends pLTV itself.[^19][^20] Samet's caveat: pLTV hasn't always helped in his accounts, because the model itself has to be sound.[^21] **Testing events.** You can't send Meta a sequence of events per user and have it learn from all of them. Pick one. To compare, launch one campaign with two identical ad sets (same ads), one on start trial (control) and one on qualified trial. Then carry the winner into creative testing.[^6] Keep it simple: a simple setup that works beats a complex one that doesn't.[^12] **Diagnosis.** If install→trial is low, look at onboarding first. But the cause can also be the event: Meta may be finding the best *trial* users rather than the best *paying* users.[^13] ## Disagreements / caveats - Steve P. Young's starter advice is plain start-trial optimisation with a 50-trials/week budget ([[meta-ads-budget-math-for-apps]]). The qualified-trial view says that's fine for volume but leaks quality. The two fit together: start-trial first, qualified trial once volume allows. - Both guests spoke on Adapty's channel, and Adapty sells qualified-trial events. The "performance improved" claim is unquantified. - Anas says the 50/week learning threshold matters only for evergreen campaigns, not test ad sets (see [[meta-app-creative-testing-and-graduation]]). ## See also - [[signal-engineering-and-conversion-event-audit]] — audit what actually reaches Meta first - [[skan-aem-capi-for-app-campaigns]] — how the event reaches Meta after ATT - [[web2app-ad-optimization-events]] — the same event ladder for web funnels - [[scaling-meta-past-100k-for-subscription-apps]] — the wider 100K+ playbook - [[people/anas-maqsood]], [[people/samet-durgun]], [[people/murat-menzilci]] [^1]: yt:8SfXEIFdB2M @ 03:57 — Anas Maqsood [^2]: yt:8SfXEIFdB2M @ 04:48 — Anas Maqsood [^3]: yt:8SfXEIFdB2M @ 07:15 — Anas Maqsood [^4]: yt:8SfXEIFdB2M @ 08:57 — Anas Maqsood [^5]: yt:8SfXEIFdB2M @ 06:25 — Murat Menzilci [^6]: yt:8SfXEIFdB2M @ 10:41 — Anas Maqsood [^7]: yt:kXXFtl0ocU0 @ 19:18 — Samet Durgun [^8]: yt:8SfXEIFdB2M @ 15:31 — Anas Maqsood [^9]: yt:kXXFtl0ocU0 @ 13:48 — Samet Durgun [^10]: yt:kXXFtl0ocU0 @ 18:27 — Samet Durgun [^11]: yt:kXXFtl0ocU0 @ 21:00 — Samet Durgun [^12]: yt:8SfXEIFdB2M @ 12:19 — Anas Maqsood [^13]: yt:8SfXEIFdB2M @ 39:11 — Anas Maqsood [^14]: yt:QdIGi-tKr28 @ 25:03 — Samet Durgun [^15]: yt:QdIGi-tKr28 @ 20:18 — Murat Menzilci [^16]: yt:QdIGi-tKr28 @ 46:57 — Murat Menzilci [^17]: yt:QdIGi-tKr28 @ 05:10 — Murat Menzilci [^18]: yt:QdIGi-tKr28 @ 09:42 — Murat Menzilci [^19]: yt:QdIGi-tKr28 @ 07:05 — Murat Menzilci [^20]: yt:QdIGi-tKr28 @ 42:36 — Murat Menzilci [^21]: yt:QdIGi-tKr28 @ 41:47 — Samet Durgun