David Vargas

Last week, I was one of the first guys announcing the new setup that TikTok offers for iOS campaigns, where you can basically enable or disable SKAN and trust in ADC only. I was not going to miss the chance of being one of the first with the new update so I started a new campaign the exact same day when I posted that to see how good ADC can be compared to SKAN.Today, as usual, I am sharing insights and real data so you can draw your own conclusions 😉Here's what the new ADC campaign has generated for me in one AEO campaign based on Purchases for a mobile game (casual category) whose SKAN schema is simply formed by revenue buckets:👉Installs ADC: 618👉Installs SKAN: 270 (-56% than ADC)👉Installs ASC: Not available ❌ --> first time in my life I […]

🚨 TikTok has changed the iOS campaign setup 🚨 Today, I noticed a new option when creating an iOS campaign on TikTok: you can now enable or disable SKAN attribution.This is essentially the same approach Meta uses with AEM, but TikTok still gives you the option to activate SKAN attribution so you can view SKAN data directly in the dashboard (something that you can also do with Meta but not in TikTok if you disable this option)The question is: What signals are the campaigns using now? Previously, you couldn't disable SKAN when ADC was enabled so I assumed the campaigns were still relying on SKAN signals for optimization. But now, it looks like TikTok has taken things a step further. It appears the algorithm is ready to optimize based on probabilistic signal attribution only.Time to test!

Last Case Study before MAU! ✈️ Are you a gaming company monetizing through IAA?🕹️ Then this case study will definitely be of interest to you.Four months ago, I was among the first to publicly test Meta’s new adROAS campaigns (feel free to check my earlier posts). Since then, I’ve been consistently testing them across different games.Today, I want to share something that has happened not once, but three times—with three different games:👉𝐌𝐞𝐭𝐚 𝐚𝐝𝐑𝐎𝐀𝐒 𝐜𝐚𝐦𝐩𝐚𝐢𝐠𝐧𝐬 "𝐪𝐮𝐢𝐜𝐤𝐥𝐲" 𝐥𝐨𝐬𝐞 𝐞𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐜𝐲 𝐰𝐡𝐞𝐧 𝐬𝐜𝐚𝐥𝐢𝐧𝐠Don’t get me wrong— These campaigns work really well on Android and have delivered strong results for me. But it seems there’s a ceiling: as you start to scale spend, efficiency drops and the breakeven point gets pushed further out.Here are the three main issues I’ve observed:1️⃣ Efficiency drops despite continuous creative rotationIn this last case, I have tested +100 creatives with […]

🚨🚨Meta has officially introduced the Engaged-view Attribution as an option for your videos🚨🚨 The worst part? They put it as a tool to decrease your CPI 😂😂😂Man, this industry... you complete the sentence. 😂

I hadn't shared my thoughts on ASA’s recent move to SKAN yet. But now that the decision has officially taken effect today, I think it’s still a good time to discuss what I expect from this. Apple has decided to align their ad network aligned with the rest of channels by using SKAN but only limited to versions 1-3 which is in my opinion, an embarrassing decision. We must remember that SKAN 4.0 was launched in October 2022!!!! 🫠 If the creator of this attribution framework isn’t even able to implement the latest version in its own ad network, you can imagine how "ready" SKAN 4 really is…But putting that aside, let's talk about what we can expect from this pivoting decision and I will be quite straightforward with my opinion as usual:👉You must expect a drop in all SKAN […]

SKAdNetwork (SKAN) 4.0, the latest version of Apple's privacy-preserving framework for mobile app install attribution, was officially launched on October 24, 2022. Today, 1076 days later, Meta is adopting that version😂😂😂😂😂😂Thank you guys! You have really shown proactivity at the highest level.

📢SKAN can work better than AEM for some apps Yes, you read that right.AEM is not a silver bullet.It can fail when, for some reason, Meta is not able to fingerprint their users 😬(because yes, they obviously do fingerprinting😂)Today, I’m sharing a real case where this happened.We started a Trial Start campaign in early February using AEM to feed the algorithm with real-time signals which indeed helps the campaigns to optimize better.The campaign started to show more signals on AEM than SKAN during the first weeks but with the time, SKAN started capturing more CTA conversions daily (as shown in the graph), making a HUGE difference on the CPT for each attribution methodology.In my case:👉Cost per trial (last 14d) according to AEM: $71👉Cost per trial (last 14d) according to SKAN: $49🤯Difference: SKAN is 30% cheaper, making my campaigns profitable—unlike AEM.Despite […]

If you are targeting older audience segments (+35yo) --> This is what you are looking for 🎯 Today I am sharing a real data case with one of my clients.In this case, our core audience was men over 35. After setting up the initial campaigns in Meta, I started experimenting with the creatives as I usually do.We had all kind of angles and concepts (UGC, static banners and motion videos) within the same ad group because we had not more budget to create separate testing ad groups. However, by analyzing the engagement metrics of our videos, we identified the key elements that resonated with our core audience. I primarily rely on the following metrics to evaluate the performance of creatives::-CPA-CPI-Outbound CTR-Hook Rate-Hold Rate-Install Rate-IPM-CVR from Install to Target ActionWith these metrics, you can easily determine which creatives perform best and […]

📢If you run iOS campaigns on Google Ads and Meta using their modeling technologies simultaneously, you need to know this: 👉Google ALWAYS tends to overestimate conversions👉Meta ALWAYS tends to underestimate conversionsAnd to prove it, I will share a recent case I had with one new client which I recently audited and had NOT SKAN schema in place (they just had cv=0 for installs and that's it)First, let's analyze Google Ads numbers: They were running one AEO campaigns for "start_trial" in the US. Since they didn't set up any event in SKAN schema, these trials were purely modeled by Google.➡Google modeled installs: 537➡SKAN installs: 427➡ASC App Referrer: 419 installs With ASC and SKAN being so close, it’s hard to trust Google’s numbers. (I know ASC & SKAN are not attributed in the same way, and they're even different conversions)➡Google modeled trials: 139 […]