AEM models Installs & Registrations better than Adjust

Here is the last case I had with 𝐀𝐄𝐌 𝐌𝐞𝐭𝐚 𝐜𝐚𝐦𝐩𝐚𝐢𝐠𝐧𝐬 for mobile gaming since Adjust sent the email about how they were expanding the coverage of AEM on October 9th (check my last post to see that).

For some reason, this is the perfect example because one day before this email came, I started one AEM campaign based on purchases and using a 7D CTA window for this client (yes, this is a sweet coincidence that you are all taking advantage off😉) so let’s see how the campaign has performed.

During the first week of the campaign (08.10 to 14.10), I see the following data on AEM:

👉331 installs
👉288 signups
👉69 total purchases

If I look at Adjust dashboard (not SKAN), I see:

👉178 installs
👉171 signups
👉20 total purchases

If I look at SKAN (keeping in mind that we didn’t receive all the postbacks yet), I see, a 6.5% null rate:

👉277 installs
👉242 signups
👉8 users who triggered any purchase (I just have a revenue model SKAN schema)

From all this data, I get the following conclusions:
🟢AEM is modelling the installs better than Adjust’s own modelling (only 19% more installs vs the -55% that we see between Adjust and SKAN)

🟢AEM is modelling the registrations better than Adjust’s own modelling (only 19% more installs vs the -41% that we see between Adjust and SKAN)

🔴AEM is modelling too many purchases compared to both Adjust’s own modelling and SKAN. Indeed, I am still missing postbacks from SKAN as we are not waiting enough time to receive all the 1st postbacks but I don’t expect more than what Adjust is modelling (20 purchases) so in this specific case, 𝐀𝐄𝐌 𝐢𝐬 𝐟𝐚𝐢𝐥𝐥𝐢𝐧𝐠 𝐰𝐡𝐞𝐧 𝐢𝐭 𝐜𝐨𝐦𝐞𝐬 𝐭𝐨 𝐦𝐨𝐝𝐞𝐥 𝐭𝐡𝐞 𝐞𝐯𝐞𝐧𝐭 𝐭𝐡𝐚𝐭 𝐰𝐞 𝐚𝐫𝐞 𝐨𝐩𝐭𝐢𝐦𝐢𝐳𝐢𝐧𝐠 𝐟𝐨𝐫!

This leaves us with a bitter feeling because it seems that AEM is improving the modelling with upper funnel metrics but is still lost when it comes to down-funnel events, which is what everyone is looking for.

🟢On the other hand we have Adjust with a more pessimistic view of the performance but also more accurate in my opinion, because it gets way closer to the amount of purchases that I hope from the app according to my experience after 1 year working with this client.

𝐖𝐡𝐚𝐭 𝐢𝐬 𝐦𝐲 𝐭𝐚𝐤𝐞 𝐡𝐞𝐫𝐞? 𝐖𝐡𝐚𝐭 𝐝𝐨 𝐈 𝐫𝐞𝐜𝐨𝐦𝐦𝐞𝐧𝐝? 👇

Making decisions with real data is a huge advantage against the delay that SKAN has so I would continue analyzing the 3 scenarios but make the main optimizations based on what Adjust models since this is the closest scenario to reality and especially, this is an AEO campaign based on purchases, not in upper funnel events!

I hope you liked this one, see you soon! 😉🤝