How not to suck at SKAN: 15 SKAdNetwork tips from all-world experts
A year and a half after the introduction of SKAdNetwork in iOS 14.5, most mobile marketers are still not comfortable with SKAN. But if you want to maximize your growth opportunities on the most lucrative mobile platform, getting good at optimizing ad campaigns via the data you can derive from SKAdNetwork is not optional. The key to getting there might just be the right SKAdNetwork tips from the right growth leaders.
Our panelists included:
- Matthew Ellinwood, director of product management at Liftoff
- Puneet Gupta, co-founder and CPO at Kayzen
- Sandy Shen, mobile marketing lead at LinkedIn
- Shannon Fishman, senior manager and paid mobile labs lead at LinkedIn
- Vassil Georgiev, user acquisition team lead at Gameloft
- Eran Friedman, co-founder and CTO at Singular.
Here’s a summary of some of the top SKAdNetwork tips that we learned …
What’s the hardest part of SKAdNetwork?
Achieving accuracy in LTV and ROAS predictions is the hardest part of SKAdNetwork, according to hundreds of growth marketers who attended the webinar. The second hardest part, not accidentally, is completely related: getting enough measurement data despite SKAN’s privacy thresholds.
In order of the most severe, the top five challenges with SKAN are:
- Unreliable LTV/ROAS: 30%
- Beating privacy thresholds: 24%
- Finding the right conversion model: 13%
- No long term data or cohorts: 19%
- Limited D1 signals: 9%
As most user acquisition experts will agree, they’re pretty much all related. It’s hard to get reliable LTV or ROAS when you don’t have enough data, when the conversion model you’ve chosen isn’t predictive, and when you only have a very brief window of insight.
How do you feel about SKAN?
Unsurprisingly, that’s led to some emotional responses to SKAdNetwork, including some despair and some anger. We also asked How not to suck at SKAN webinar participants to tell us how they feel about SKAN.
A very small number of marketers feel like they are “crushing” SKAN. A fifth feel like they are lost, with no idea where to start. A small number — very small, fortunately — feel like there is no hope at all: SKAN will never work for them and their apps.
But a significant majority at least has hope:
- We have hope it will work for us: 67%
- We’re so lost, no idea where to start: 22%
- We’re crushing it: 9%
- SKAN will never work for us: 2%
I don’t have prior data, but I’m willing to go out on a limb: two thirds of marketers having hope that SKAN will work for them, that SKAN can be made performant for their user acquisition efforts … that’s significantly up from where it was a year ago or even 6 months ago. Which means that with the right SKAdNetwork tips, there’s a chance we can get the majority of mobile marketers feeling confident about their ability to grow their apps.
Feeding the hope: 15 SKAdNetwork tips from experts
1. Start with the right mindset
The old way you used to grow apps is over. IDFA availability is running around 20-25%, and it’s not going back to the 70-80% you had before App Tracking Transparency.
Understand that. Accept that. Lean into that. Get the right SKAdNetwork tips to make it work for you.
Then apply the resources you need to figure SKAN out, says LinkedIn’s Shannon Fishman:
“I think the biggest way that we have succeeded is by constantly testing and never giving up the hope that we can accomplish success with SKAN. I think part of that comes from really releasing the shackles of pre-SKAN mindsets and making sure that you have the right measurement, engineering, resources, and budget allocation to push forward your program goals.”
Release the shackles.
2. Zoom in to your app. It’s not just about marketing metrics
Marketing is all about the metrics, right?
It’s also about knowing your app. Knowing your user journey. Knowing your players or customers, and knowing how they live in your app.
This has been a core change for Gameloft since SKAN, says team lead Vassil Georgiev:
“On the micro level, I can definitely tell you that my team has been zoomed out quite a lot from the user profile and persona as a whole. And right now, they are quite more zoomed into the user journey. And on a macro level, in terms of crisis, it definitely brought the different key stakeholders around the company like growth data science, the growth managers, and the game designers closer than ever.”
Understanding your app and what people do inside it to a very deep level helps you understand what’s core about your growth loops, what’s critical to measure in SKAN conversion models, and what features and fun to focus on in your ads.
3. Get the full team on board, not just marketers
Marketing metrics is a marketing-only problem, right?
It never really was, but especially in the era of scarce data from SKAN and soon Privacy Sandbox, it’s more important than ever to engage the entire team. Especially when you’re starting out and need to do some brain surgery to start making SKAN performant.
From Sandy Shen, mobile marketing lead at LinkedIn:
“We connect two sides of engineering to ensure everything looks correct … helping us avoid back-and-forth investigation between integration or campaign setup when there are no conversion values coming in. So it did take us a while to see conversions show up. But we can focus more on brainstorming campaign setup.”
Get all hands on deck: product, engineering, data, and marketers. Stay on deck through the implementation and testing process. Brainstorm together. It’s everyone’s job to figure it out and land the plane.
4. Be patient
I know it’s the least favorite thing for mobile marketers to hear. You’re in tech, which is fast-paced, and you’re in mobile, which is perhaps the fastest-paced part of tech.
But patience is critical.
From Sandy Shen again:
“Be patient, let your new campaign run for at least two weeks.”
When you’re testing new conversion models you need to let the old data flush out of the ecosystem to get a clean comparison. So you might have to restrict spend for a period of a month or more when you’re first implementing SKAdNetwork. Test on a fraction of your spend — while spending enough with the channels you select to surpass privacy thresholds — and go though some iterations.
When you’re happy with your results and can spend confidently, open the spigots back up.
5. Model conversions to fix missing data
When too much of your data is censored due to privacy thresholds, conversions and installs look very expensive: more spend spread over fewer events.
One of the most important SKAdNetwork tips, therefore, is to use tech like Singular’s SKAN Advanced Analytics to model missing data. Now your conversions and installs return to expected levels.
From Singular CTO Eran Friedman, speaking about a customer he worked with:
“What really helped them is the use of our SKAN advanced analytics product, which focuses on modeling conversions to basically eliminate the privacy thresholds and provide longer cohorts, seven-day runs essentially. And that finally showed them equivalent performance for the SKAN campaigns as what they were used to.”
This particular customer paused spending because the results were unpredictable. Metrics were unreliable. CPIs were off the charts. Using modeled metrics that lean on first-party data to provide estimated performance within specified confidence intervals changed everything.
6. Know what matters most
With free IDFAs for all, mobile marketers could measure almost anything and everything, and frequently did. And you could measure retroactively, to see what metrics might be more powerful.
Those days are gone. Now you have to pick specific events and/or revenue and use just a few data points to predict overall results.
From Sandy Shen at LinkedIn:
“On the LinkedIn side, we understand what matters most important to us …”
Before you can understand what matters most, you have to (see #2 above) know what people do in your app. Then study that and see what events have the most predictive power. Now you can start setting up conversion models.
This will get somewhat easier in SKAN 4, with more data. But it will come at the cost of significant added complexity.
7. KISS: keep it simple …
SKAN is already complex, and getting more so in version 4.
So keep it simple.
Again from Sandy at LinkedIn:
“After SKAN, we started to prioritize our existing metrics, and only kept a few. Because as we know, the more metrics you want to track, the longer delay you will see in your SKAN reporting.”
Pick the most important. Use SKAN Advanced Analytics to hedge your bets a little with mixed models so you get a few events and some revenue data. But don’t boil the ocean and try to every everything instantly.
8. Leverage Android visibility
There are some things you can learn on Android that might help on iOS. The obvious benefit: you still have full advertising ID visibility in the full scope of the customer journey with GAID.
From Gameloft’s Vassil Georgiev:
“In the era of SKAN 3.0, we still have the luxury of having visibility on the Android apps. So by having both pre-install data like impressions, clicks and costs, and also posting some data like conversions and re-conversions on Google Play, you can shape any kind of hypothesis and test it out. And in case of success, you can actually get the right SKAN model to implement.”
Android offers a massive population for testing, no matter what country you’re in. Some things you learn there will apply to your iOS users.
9. But don’t lean on Android too much
With such huge numbers of people in both the iOS and Android worlds, there’s definitely some crossover. But with average iPhone prices being higher than Android phones and iOS apps being roughly twice as lucrative as Android, there are obviously also significant differences.
Again from Gameloft’s Vassil Georgiev:
“Don’t underestimate the power of creatives, because I can tell you that we see different performance from Android to iOS. So creative testing still matters.”
In other words, don’t think you can figure it all out on Android and copy-paste on iOS. Unfortunately, it’s not that simple.
10. Build signal by how you buy ads (and where you buy them)
Some SKAdNetwork tips aren’t about SKAdNetwork at all.
Because you don’t just get marketing signal from ad delivery like clicks, impressions, viewability, and CTR. And you don’t just get signal from deeper-funnel metrics like installs, events, conversions, engagement, and purchases.
You also get signal by the very act of choosing ad partners.
From LinkedIn’s Shannon Fishman:
“It’s important to also reflect on your media mix as another lever that you’re using to make sure that your campaign targeting, your geo segmentation, your creative is really able to drive the value that you need at your core geos and at your business goals … I think that is an easy thing to overlook. It’s just thought, oh, I think they’re all doing the same. They’re not. They’re definitely not. And I think that’s something that advertisers really need to be okay with being flexible on.”
In other words, different ad partners target different audiences in different regions at different stages of different journeys.
Use that as part of your incrementality testing.
11. Get down into the weeds
Just like you have to dive down into your app and familiarize yourself with those waters in a very intimate way, you also have to get into the minutiae of your ad campaigns, including SKAN.
From Liftoff’s director of product management Matthew Ellinwood:
“In order to drive results from channel to channel and understand what you’re seeing, you probably need to get down into those execution-level details for the campaigns that you’re putting in place.”
Getting your hands dirty pays dividends. Each ad network deals with the challenges of wrenching marketing signal from SKAN in a slightly different way. Understanding that will help you optimize campaigns.
12. Try more conversion models
How many conversion models have you tried for SKAdNetwork? Successful brands are probably testing more.
From Singular’s Eran Friedman:
“I saw that for customers running SKAN at scale, the average number of conversion models that they’ve tried was 5. So they’ve literally, on average, tried at least 5 conversion models until they found the one that’s optimized for their use to be able to scale. Which probably means that … I’m just gonna put something out there and it’s gonna work out of the box.”
If successful brands average 5 conversion model tests, some of the outliers have tested 7 or 10 models before finding one that works.
Keep testing. There is no finish line, and there is no going back to the way things were. You figure out how to make SKAN function, or you abandon performance marketing on iOS.
13. Don’t forget ad monetization revenue in your conversion models
Some apps earn most of their money via ads. For others, it’s only 25%.
Either way, it’s important data when assessing return on ad spend. So ensure it’s part of your calculations, especially if your apps rely heavily on ad revenue.
From Vassil at Gameloft:
“You need to be very careful when you set up the ad buckets in your revenue model. Basically, if you ignore revenue from ads, it’s not going to work. And you’re going to learn it the hard way.”
Learning the hard way sucks. Let others make mistakes, and measure your admon as well as your IAPs.
14. Find strong-signal proxy events
If you run a subscription app or an IAP-heavy game that doesn’t monetize massively on D1 or even by D7, you need to find proxy events that signal future revenue potential with as high predictability as possible.
From Vassil again:
“If you have strategy games that have longer LTV and don’t monetize so fast, you always struggle with providing enough signal during the first week of the campaign so it can get out of the learning phase. So this is the moment that you can think about modeled conversion, and some kind of engagement or proxy event … towards the purchase itself, that will actually make the SKAN work for you.”
Did I mention something about knowing your own apps and games better than even your customers previously …
15. Start ad campaigns narrow, not wide
Don’t spread your ad dollars too wide at first, especially if you have smaller budgets. Instead, concentrate them to maximize signal.
From Puneet Gupta, co-founder and CPO at Kayzen:
“The way you are setting up campaigns in any of your networks eventually boils down to your targeting criteria, right? And, for me it’s 2 stages: Either you go wide, or you go narrow. Our suggestion usually is that if you are starting new, and you don’t have a lot of learnings, start narrow. It just makes sense from a budget perspective … but more importantly from just a privacy threshold perspective, it just makes sense that the probability of getting more data from SKAN is higher if you’re getting 1,000 installs from 10 different publisher apps versus the same 1,000 from 500 different apps.”
Clearly once you have SKAN down pat, you need to expand your focus and test additional channels and partners. But until you know you can do so predictably and profitably, focus your spend narrow to get the most data back from SKAN.
So many more SKAdNetwork tips in the full webinar
This is just a small sample of the insights from our panelists. Access the full webinar here, for free, to get all of them. And, of course, to get them in context.
You can be successful at SKAN. We’ve seen it with hundreds of apps and brands. But it does take some work … and the right partners.
Talk to an expert at Singular if you have any questions.