SKAN and the mobile measurement landscape

Learn how to navigate the ever changing mobile marketing industry from the lens of a user acquisition manager. Recent user privacy measurements may have altered the amount of data we use to inform optimization decisions, but predictive analytics can address those limitations with insights based on early measurements.

Hey, welcome back to another episode of LevelUp Academy. I’m Michel Hayet, a senior product marketing manager at AppsFlyer. In our previous episode, we gave a high-level overview of predictive analytics. Now, we’re discussing SKAN and the new mobile measurement landscape. Then we’ll put it all together in the final installment of the series. So let’s dive in and talk about app marketing as it stands today.

Mobile marketing is an ever-shifting industry. In the last three years in particular, the pace of change has really accelerated. Especially when it comes to user privacy and the way that data is shared and consumed.

Let’s take a minute here and put ourselves in the shoes of a user acquisition manager. For some of you, this may not take a lot of imagination.

The amount of information we need to think about for each optimization decision is insane. Is this campaign ROI positive or negative? Is there a suspicion of fraud? What media channels provide the best value? What bid should I place? When should I make the decision? And this is just the tip of the iceberg.

In an industry where time equals money, these decisions need to be made fast and with as much accuracy as possible. Up until recently, this task was a lot easier because you could access pretty much any data point.

Remember when you could use unique user identifiers to measure LTV at any given time in the user journey? Basically, you got access to an endless stream of real-time metrics that let you optimize performance based on clear and actionable KPIs.

But the world is changing

SKAN and App Tracking Transparency privacy limitations have made attribution measurement significantly more difficult

How? Well, first of all, you need to wait twenty-four to forty-eight hours to get a postback confirming users installed your app. So for that time, you’re basically a sitting duck with no insight into the user journey. Second of all, the ATT and user consent requirements practically eliminated the ability to identify specific users. And third, you’re limited to just one postback. Which means your install data and a conversion value are grouped into one event, leaving you with fewer insights.

Let’s pause for a moment here. I just mentioned conversion values, which are a key part of UA under SKAN.

What are these SKAN conversion values?

They’re the way that Apple lets developers and ad networks attribute in-app events to the install. In other words, they measure post-install activity and tie it back to the install. And the app developer sets these values based on relevant business KPIs.

For example, if one of your key metrics is average revenue per user, you’d look to measure revenue and the amount of users. So to get to these two metrics, you need to use the right conversion values, like a value for users who retain until day 30 or another for users who spend $5 by level 6. See how your conversion values should match up directly with your target KPI?

Okay, back to SKAN. All of these changes are great for maintaining user privacy. But they made user acquisition significantly more difficult.

Now UA managers have to rely on data that’s limited to very specific events. And that can only be measured in the very early stages after users install the app.

This can all be fine and good if your core business relies on actions that occur within the initial twenty-four to seventy-two hours. But what about cases where your key events happen beyond day three? If you’re a subscription-based app with users who usually convert at day ten, then you need predictive analytics. You’re really only looking at these later-day events to measure user value.

This new reality basically means that optimization decisions have to be made within a very short time frame. And they have to be delivered in a very specific format – the conversion value., All while respecting user privacy, meaning they can’t rely on user identifiers.

Here’s the good news.

All limitations are addressed and solved with predictive analytics

The smart algorithm can drive accurate insights based on early measurements collected over a relatively short period of time. These insights can be delivered in the format of your choosing – in this case, conversion values. And while user identifiers could be an additional layer of information to add to the model, predictive analytics in no way rely on it. Basically, predictions can be completely anonymous.

So that about covers where the mobile industry stands now, with a brief introduction into the role of predictive analytics. In the next part, we’ll go even deeper into how predictive analytics addresses the SKAN challenge. Tune in as we bring these two worlds together.