UPDATED for SKAdnetwork (SKAN) 4.0: Deciding the Optimal Conversion Value and Window Setup
The details of SKAdNetwork (SKAN) 4.0 are out in the wild. As an advertiser, it’s time to start thinking about how to shape your setup to get the most actionable insights from your SKAN 4.0 data to drive performance on iOS.
In this post we want to guide you through some early thinking on how to setup Conversion Values optimally to deliver the iOS user acquisition insights you need. There have been many posts summarizing the initial SKAN 4.0 announcement at WWDC and the actual release as part of iOS 16.1. As a primer, this visual from Sara Camden on the InMobi DSP team gives a good overview of some of the key parts of SKAN 4.0.
For the purposes of this post we’re going to assume you have a base understanding of previous versions of SKAN and what the major new features are as part of SKAN 4.0, so we dive directly into what you need to start thinking about as an advertiser in terms of Conversion Values and Windows.
Here’s some key high-level thinking, before jumping into specific thoughts:
- Don’t start with all of SKAN 4.0s functionality. SKAN 4.0 comes with a wealth of new features and flexibility, which is great for the industry. However, you won’t need to use everything. For example, not everyone tracks every metric and uses every feature that is possible in their MMP. At least, not in a meaningful way. Some of it adds unnecessary complexity that you don’t need. That’s the same with SKAN 4.0: using everything won’t be right for every app. If you start by looking at everything that’s possible you will quickly end up tangling yourself in complexity and delivering unusable reporting data.
- Start instead by imagining your ideal reporting. First, begin by thinking about what your ideal output would be at the simplest level from a reporting perspective. What are the key events that you’re optimizing towards? What are the KPIs that your campaigns are judged on? What cohort windows are you using? Starting with this will help you avoid adding unnecessary complexity during implementation.
- Focus on coarse-grain Conversion Values as the measurement foundation. Whilst the 6-bit fine-grain conversion value adds more detail – which will be useful to get extra colour when available – you will need to normalize it back to the coarse-grain low, medium and high value. Fine-grain postbacks will only be available for the first postback covering days 1 and 2 post-install and there’s a good chance that a high volume of the first postbacks you receive will be coarse-grain. The good news is that fine-grain Conversion Values can easily be normalized back to the aggregate coarse-grain value. The shift in mindset here is to using attribution to give you a quick idea of what’s working and what’s not at a campaign level versus a historical mindset of deterministically tracking everything down to a granular level.
The advertisers we saw have most success with SKAN Conversion Values in its earliest format started by tracking whether six events had occurred or not, with the simplest type of Conversion Value schema. From this stage they added more complexity gradually, however, they started by keeping it simple and thinking about the six events that correlated most strongly with long-term revenue.
That’s much the same here. Starting your SKAN 4.0 journey with the simplest setup and layering on complexity when you think it will enrich your reporting insights will work best for most advertisers. Typically, we see customers optimizing towards three key KPIs / events across 3 cohort windows. Those three KPIs and cohort windows are likely the best place to start when setting up SKAN 4.0. For example, a fintech app might be optimizing towards registrations, first deposit and $50+ deposit on D1, D7 and D30 cohort windows. Or a subscription app might be optimizing towards registrations, free trials and paid subscriptions on D2, D7 and D35 depending on their free trial length. It typically looks something like this based on app category:
With a little bit of data analysis it’s quite easy for most mobile apps to understand the three key conversion events and cohort windows in their monetization funnel to then start their SKAN 4.0 journey from there.
Top three tips for setting Conversion Values and Windows
So now we’ve established a focus on getting simple, understandable and consistent data to deliver actionable optimization insights, we really have three key tips when it comes to setting up SKAN 4.0 Conversion Values and Windows.
1.Decide on three event buckets and stick to that throughout.
With your low, medium and high value for coarse-grain Conversion Values, there’s a temptation to use different values for each across the three windows. However, as we mentioned most apps are really optimizing towards three key events and in the spirit of keeping it simple it’s likely best for most advertisers just to have the same event for low, medium and high across the three windows.
This will make it easier for deciphering the performance of a campaign in your reporting. By using different events across all windows, you’re adding unnecessary complexity to your reporting, when all most advertisers really care about is three metrics.
2.Use two bits of fine-grain values for coarse-grain normalization
To normalize fine-grain and coarse-grain Conversion Values, it’s worth reserving 2-bits of your fine-grain conversion value to define whether it’s a low, medium or high coarse-grain value. You can then use the other 4 bits to track additional signal, whether that’s conversion events that have a lower correlation with long term monetization than your three key events, revenue buckets or both.
Ultimately, this means you can easily normalize all postbacks to coarse-grain and have clean optimization signals in your reporting across a high volume of installs.
3.Lock windows as early as possible.
Whilst you have three windows 0-2, 3-7 and 8-35 days post-install, the first window has a 24-48 hour delay and the second two have a 24-144 hour delay. That is a long time waiting to understand the performance of your campaigns. If you’re running a new campaign that is going badly wrong you don’t want to have to wait long to find out.
That’s why we’d suggest most advertisers lock the windows as early as possible on a fixed day. This way you get consistent data as early as possible to make smarter optimization decisions. It’s worth analyzing how long it takes for the majority of conversions to have happened for a specific event. Then you can set the window for a day post-install when you’re likely to have captured 80%+ of conversions. For example, for the hypothetical subscription app below you might want to set the windows to lock on day 2, day 4 and day 10.
This analysis is worth doing for the three KPIs you’ve identified in your app and then identifying the earliest you can set the conversion window to lock whilst still capturing the majority of conversions. This should give you directional numbers that you can extrapolate for longer-lead conversions. The big ‘”watch out” here is that once you lock a window, anything that happens after that window is locked and before the next window is open will not be tracked. That’s why it’s best to lock on a consistent day when you feel you will have captured most conversions. This way you have a consistent apples-to-apples comparison across campaigns. For an app category like ecommerce or marketplace where purchases may be more spread out locking windows early may not be possible.
We believe the reality of the third postback for many apps is that it won’t necessarily be used for optimization, but more for validating CPAs, since the data will take so long to arrive.
The general jist of what we’re suggesting for SKAN 4.0 is keep it simple to start and then layer on complexity from there. Once you have usable directional data that gives you timely actionable insights you can incrementally add functionality to enrich the data. Don’t start by using every feature of SKAN 4.0 and every possible metric you can track as you might find the data on the other end is unusable. Start by understanding your three key conversion events and 3 cohort windows and work back into SKAN 4.0 from there with a focus on getting the data back as quickly as possible without missing too much signal.
Whatever way you look at it, the data you get back from SKAN 4.0 is going to be more complex than ever. To turn it into actionable insights you are going to need to manipulate and adjust it and unify it with different datasets like media cost, predictive models and potentially layer Media Mix Models at a cross-channel level. Unifying and normalizing these datasets to deliver actionable insights is exactly what Appsumer was built for. If you haven’t yet setup a free account to try it for yourself now would be a good time to get this reporting infrastructure before layering on more complexity. You can grab your free account here.
Subscribe to our weekly Editions
Original thinking and essential industry news for ambitious app marketers.