SKAN 3 in review: 16 insights we learned
- What have we learned?
- What mistakes did we make?
- What do we need to know today, given SKAN 3 is still the main game in town?
You can now watch that webinar on-demand (and I highly recommend it). The panelists sharing their insights and tips were:
- Edouard Favier from A Thinking Ape
- Vanessa Simmons from Feedmob
- Santiago Casais from Smadex
- Noah Gerard-Grossman from Unity
- Shamanth Rao from Rocketship
SKAN 3 in review: bye-bye granularity, and the stages of grief
As we all know, the big change with SKAN is the loss of granularity.
“Anything that’s user level reporting or targeting, all of that is now kicked up to be campaign level aggregate,” said Vanessa Simmons, ad operations senior team lead at Feedmob. “It really changes how you’re visualizing your data, how you’re doing your optimizations, all of your performance.”
This changed everything and, as Shamanth Rao says, everyone in the industry had to move through all the stages of grief when SKAN 3 came:
Once the industry moved through to acceptance, however, we all learned some key lessons. Here are 16 mentioned in the webinar …
What we learned this past year
We all learned much more than we ever thought we’d have to about privacy-safe marketing measurement.
Some of the key learnings from our experts in the SKAN 3 in review webinar include:
- Adapt your KPIs to what SKAN can provide
“We had a client who turned off all of iOS and spent a really long time working with their internal teams to build out their internal first party data, making sure that their mapping was going to be something that their system would recognize … their main event was something that happened at day 45, way outside of the postback window. But because they had spent the time working with all the necessary parties, they were able to get that one postback and compare it to that first party data,” Simmons says.
- Focus on immediate post-install events
“Capture an event that’s going to track as closely as possible to user LTV and convert within the first 24 to 48 hours,” says Gerard-Grossman.
- Study your users to learn which events are predictive of future value
“It really depends on the type of apps you have but for us we focus most on in-app purchases,” says Favier. “We quickly realized that it was by far the most important signal for us and the model has to be basically 100% focused on that.”
- Lean on your first-party user data, which hasn’t changed
“Other signals, of course, are your own first party data,” I added at one point. “What am I seeing? What can I sort of create cohorts out of? What performance am I seeing from there? How can I add that or layer that over top of my SKAN results?”
- Test on low impact apps in your portfolio
“If you have the chance to have a backend portfolio of apps where you can test things, do it,” says Favier. “That’s probably the best way.”
- Use ATT-yes users to help inform SKAN users
“Look at different signals you can get for your campaign,” Favier says. “It doesn’t necessarily have to be the SKAN signals. It can also be a deterministic approach: try to improve ATT approval, the number of IDFAs you’re getting.”
- Build volume within campaigns
“Start with a high level of installs per day,” says Casais. “This will enable you to cross all thresholds and all campaign IDs and get as much granular data as possible.”
- Be patient
“Day one, there’s going to be a new campaign launched, a lot of new impressions, but installs won’t even be coming in yet,” says Gerard-Grossman.
- Minimize non-essential changes to avoid data delays
“Minimize changes within campaign IDs, or know how that’s going to affect,” says Casais. “The data will be more precise and you won’t have to deal with install delays when changing these campaign IDs.”
- Understand that some apps are perfect for SKAN
“Their main KPI was a registration, a sign up, something that happened almost immediately after installing,” says Simmons. “And it was a really great indicator that this person was going to complete X, Y, and Z afterwards … so running on SKAN, it was really nice.”
- Lean on partners to help understand and use SKAN
“Discuss with some of your other ad partners,” Favier says. “I’m sure they will be willing to help and that will make your life a lot easier and it will allow you to test as much as you can.”
- Tailor your conversion model to your app
“You’re going to lose data,” says Favier. “It’s going to happen, but you have to look more at the gains over the loss of that. And a better conversion model can significantly change the performance of your campaign, especially with networks that rely a lot on SKAN.”
- Use multiple data sources to validate and enrich SKAN data
“Use multiple data sources because there’s no one single deterministic source now,” says Rao.
- Don’t be afraid to learn
What’s important? “Learning all the details, understanding all the terms, understanding kind of the nitty gritty of SKAN and how it’s changing as well as how network support for it is changing,” says Gerard-Grossman. “And then taking the next step to apply it to your specific app and making sure that your conversion model is set up right, that you’re making use of it as best as possible.”
- Use your conversion scheme to segment users
“The goal of the schema should be to separate out high value users versus low value users,” says Rao. “And from that perspective, just a revenue schema is the best. You could use non-revenue events preceding revenue, like sign up, complete onboarding, etc., but really the bulk of it should be revenue.”
- Try different types of testing
“It’s mostly testing different SKAN models, different conversion models and different types of conversion models,” Favier says. “So you can switch the timers between the conversion models, you can vary the events, you can vary the value of the events.”
Much more in the full SKAN 3 in review webinar
There’s so much more in the full webinar, and it’s available now, for free, on demand. Go get it right here: