Using AEM & ADC for your iOS campaigns on Meta and TikTok

WELCOME TO CHAOS 😂

This is a case I wanted to share, as it’s a situation I’m facing more and more lately: SKAN is reporting higher than probabilistic attribution methodologies, as you can see in the charts.

I think this can really help people to understand how to deal with different data scenarios when using AEM & ADC for your iOS campaigns on Meta and TikTok Ads respectively.

Context of the setup:

✅ One campaign in facebook optimized for an upper funnel event, using AEM
✅ One campaign in TikTok optimized for the same upper funnel event, using ADC
✅ One campaign in facebook optimized for a lower funnel event, using AEM
👉 Appsflyer as the MMP
👉 3 scenarios: Appsflyer classic attribution, ADC/AEM & SKAN (I have not modeled ASC data to make it simpler)

❌ Important note: when you enable ADC for TikTok, your classic attribution on the MMP becomes useless — it just mirrors ADC data.

What are we facing?

💀 Meta continuously underreports through AEM when compared to SKAN (CTA only)

💀 TikTok, which used to overreport through ADC, shifted to the same situation as Meta (underreporting vs SKAN) once we started reducing budget.

This leads to different conclusions depending on which data point you use. For example:

🔷 Facebook Campaign 1 shows a better CPA than Campaign 2 when looking at AEM, but it’s much worse when looking at SKAN.
🔷TikTok Campaign 1 shows a much better upper-funnel CPA on SKAN, but performs worse when compared to the lower-funnel SKAN data of Facebook Campaign 2.
🔷We have also seen how Meta campaigns, despite showing much higher CPA on AEM compared to ADC, actually achieve better CPA according to SKAN.

What do I know? 💡

– From my experience, I’m 99% sure that AEM is underreporting (I’ve seen this across many accounts, and it’s also clear when you look at the number of App Referrer installs in ASC).

– ADC also underreports a bit (based on ASC + blended), but not as much as AEM.

– SKAN is stable and deterministic, but as I’ve shown many times, it also underreports

What do I do? 🎯

As I explained in an article I wrote for RevenueCat a few months ago, I use an adjustment factor to align reality in both directions — but I ALWAYS prioritize SKAN when doing my math.

👉 If my math works with SKAN numbers (even knowing SKAN underreports), I can be confident about scaling or shifting more budget toward those campaigns, since they’ll bring higher value/efficiency.

👉 I still use ADC & AEM data for creative optimization, since those are the signals Meta and TikTok algorithms actually optimize against.

👉 If SKAN math doesn’t work, I apply the adjustment factor. If adjusted math suggests shifting the budget, I’ll do it and closely monitor blended results to confirm the impact.

Hope it helps! The post got too long and I don’t have more space to write! 😂
Have a nice week ahead!