The Truth About SKAdNetwork Nobody is Talking

Believe me when I say that the title of this article is 100% accurate. No thumbnail. This is not for you if you expect to find a magic solution or a hack to your SKAN problems. Difficult problems require complex approaches and great minds to solve them.

I’m not going to explain the basics concepts of SKAN, Conversion Values, Timer Mechanisms, ATT, the impact of a post-IDFA world, etc. So, you may not understand this article if you are new to this world. The main goal is understanding how to coexist with measurement technologies, specifically Apple’s attribution tool.

Let’s begin with the #1 question I receive every day: Lucas, how do you use SKAN in your favor and how do you recommend I set up my Conversion Value Schema? Are you ready for the secret answer? Well, IT DEPENDS.

Know Your Limits

It won’t be the same if you work with a small team in a new startup as if you work with a team of 5 UA managers or 10 data guys. My take here is simple if you are just starting and don’t have enough resources, PLEASE DON’T COMPLICATE yourself. Keep things simple, focus on one or two main KPIs, and start. SKAN it’s designed for the elites (further in the article, you’ll get the point).

I know that ad networks, MMPs, agencies, etc., will tell you to use the 63 bits to maximize the data you receive in the postbacks. But maybe that’s not the case for you, and it’s okay.

Focus on in-app events that happen in the first 24 hours window after the first open event fires, and when setting up the conversion value schema, test a simple design. What do I mean by simple design? Just use in-app events; if you feel comfortable, try with revenue or funnel; if possible, avoid ranges/buckets, don’t use intervals for the activity window, and use a maximum of 20 to 30 conversion values.

Also, it is crucial that if you are working with a mobile measurement platform, the events you included in your conversion value schema must be the same as you added to the in-app event postback section.

For example, for fintech apps, you could focus only on the connect card event or first trade. For new shopping apps, you could start testing the purchase events with one or two revenue ranges. Focus more on revenue or a few-level games for small mobile game apps.

One quick and friendly reminder. Everything I’m saying here is based on my experience working with multiple apps of different sizes. There’s no single answer; whatever works for me may not work for you.

On the other hand, if you find yourself in a situation where you are spending more than 100K a month at least and have a dedicated team of data engineers/scientists, it would be easy to maximize the use of the 63 bits depending on your revenue model. In this case, you could use predictive LTV modeling and utilize different data sources (SKAN, SRN’s modeled conversions, IDFA cohorts, Backend Data, Consoles, Product Analytics, Incrementality, MMM, etc.) to create the ideal conversion value schema.

For example, you could implement more advanced strategies like adding up to 6 in-app events for one single conversion value or use a split strategy where a single conversion value can contain revenue ranges and an in-app event funnel priority.

As you can see, there are limitless options for approaching SKAN, but remember that the best strategy is just to start and be prepared to fail/test.

Use SKAN Conversion Values Cons in Your Favor

As a UA/performance guy, I want to give you some tips to consider to maximize our current limitations. Get ready to take notes because I learned these 8 things by making mistakes.

1. Avoid the timer mechanism: I don’t mind if your most important event happens 5 days after the install; you should eliminate the option of updating your CVs. Why? Because you will lose time/money and won’t be able to optimize your campaigns on time. Be sure to receive the postback before the 72 hours window.

But wait. A typical scenario would be an app with most of its IAPs happening after 24 hours. This is a problematic scenario for new startups because you will have to think of a conversion value schema optimize to cut the timer mechanism after 24 hours. In this case, I highly recommend using as many different data sources and modeling a “single source of truth” to create predictive models that will help you to map key events and strategic revenue ranges. The issue with this is that you will make probabilistic conclusions affecting different countries and user profiles. Therefore, the data you will see inside your BI tool won’t have a deterministic matching. Not everything that glitters is gold 🙂.

Finally, a 24 hours conversion window will help you optimize for custom events within your user acquisition campaigns and control media buying costs while maintaining high user value.

2. Simplify the mapping: more data = more noise. Remember that if you don’t have the resources to build and maintain a robust data stack to track SKAN postbacks, you should avoid sending information that’s not relevant. Keep it simple and feel comfortable testing a different approach (but don’t forget about the 24 hours cooldown period that ad networks require after you update the CV Schema.)

3. Intervals are only for the pro: avoid setting up intervals of 12 hours or 24 hours in case you configure a longer activity window inside your MMP. I tested in the past, and it was a waste of time and effort for the teams to materialize the impact of receiving this information in our user acquisition performance.

4. Use only one CV Schema: always utilize a single conversion value mapping. You can’t have a separate CV Schema for different ad networks. This means that your mapping should look the same in Meta and Snapchat. Using an MMP will help you unify this.

5. Be aware of multiple SDK’s updating the CV: if you are using the Facebook SDK and want to migrate to an MMP SDK, make sure to plan this transition. Remember to pause your campaigns (Meta has already removed the cooldown period) and deactivate your current SDK. Only one single ad tech can update/call your conversion values at the same time.

6. Think twice before you change your CV Schema: an optimal testing period for a CV mapping is at least 10/14 days. Also, remember that you will have to pause your campaigns for at least 24 hours before testing the new schema (except Meta). So be smart and have a genuine reason before updating your CV schema.

7. IDFA-cohort is your friend: use these deterministic attributed cohorts to model your data. These real users can help you understand the big picture that SKAN is missing.

8. Prepare for max liquidity: keep in mind the daily threshold to receive the full postback. Multiply your average CPI x 125 to calculate the minimum daily budget for each iOS campaign.

New Era, New Rules

One of the things about the mobile marketing industry is that everything changes fast. SKAN helped us understand how to go beyond the numbers and the overall business we are part of. It has taught us to sync with different teams and a bit of aggregated data, be more innovative at making decisions lacking user-level information and prepare for the new shifts toward privacy-centric regulations.

As you may know, SKAN 4.0 has already been implemented by Apple and will change the game forever. The fun part is that what you read above will be useless.

The SKAN game is an interesting one. There is no correct answer; only the brave ones see the light at the end of the road.

Good luck, my colleagues. The game has just begun.