Discover more from Growth Gems
💎 Growth Gems #78 - Paid UA and Data (SKAN 4.0)
Try out the free Growth Gems newsletter👇
This week I’m sharing gems on:
These insights come from Alex Bauer, Eric Seufert, Jesse Pujji, and Simon Lejeune.
🥇 TOP GEM OF THE WEEK
Data: SKAN 4.0 release
The mobile industry’s favorite topic is back.
After the WWDC session on SKAN 4.0 was published, I shared several insights in Growth Gems #71.
Apple surprised-released SKAN 4.0 last week, and a lot has been published about it already.
One piece of content that stood out is the discussion on the MobileDevMemo podcast between Alex Bauer (Head of Product & Market Strategy at Branch) and Eric Seufert (Analyst and Strategy Consultant at Heracles Media): SKAdNetwork 4.0 is live. What does that mean for mobile advertising?
It’s particularly insightful because it goes beyond the technicalities and gives some pointers on how to react.
Before we start, here is some context if you’re not that familiar with the topic 🤯
💎 Apple extended the length of the randomization timer. It used to be 24 hours, and it’s now up to 144 hours (6 days). This will still make it difficult to make any time-based cohorting, which people were looking forward to with the multiple postbacks.
⛏️ Going Deeper: below is Alex’s take on the changes to the postback delay logic and the “locked postback window”.
💎 With SKAN 4.0, you get to cycle back some early-stage data for the ad platforms while also registering mid-stage data via the 2nd postback window and late-stage data via the 3rd post window.
⛏️ Going Deeper: it’s important to note that you can not tie the three postbacks at the user level.
💎 With the coarse conversion value, you get signal when you launch a new campaign (or change an existing one) that helps you decide if you should “invest in data” to get past the privacy thresholds, or stop the campaign.
💎 With the conversion locking window, you can decide to start the randomized timer early on.
💎 Once you know a campaign works, you don’t need immediate feedback as much because you’re not making as many changes, so using the locked conversion window might not be as valuable. There still could be swings in audience reaction to your ads, creative fatigue, etc., but you’re less worried about a “big shock”.
🤔 My 2 cents: I agree there, but also keep in mind that the decision is not made at the campaign level. And that adding/removing the locked conversion window might disrupt your campaigns.
💎 The crowd anonymity tiers set on install are the brain behind privacy in SKAN 4.0. Previously, it was only two buckets, and we were talking about “privacy thresholds”. Now:
Tier 0 is equivalent to the worst-case scenario in SKAN < 3.0.
Tiers 1/2/3 control how many postbacks you will receive and how many digits of the source identifier you will receive (2, 3, or 4). They also control which version of the first postback conversion value you get (fine grain or coarse grain).
⛏️ Going Deeper: the following diagram from the Apple documentation depicts the relationship between the tiers and relative crowd sizes.
Below is a very helpful table from Branch’s webinar on how the crowd anonymity tiers control the postback data you get. It’s unclear whether the tiers are “linear” or not, and we will need data to know.
💎 It’s only in the first postback that you’re eligible for more than two digits for the source identifier. Postbacks 2 and 3 are locked to only two digits.
⛏️ Going Deeper: we saw in Growth Gems #71 that the source identifier can now have 4 digits (decimal, not binary). Ad networks can encode any campaign breakdown they wish to encode within these 4 digits and later decode it: campaign ids, geo, placement, ad creative, etc.
In the Twitter space he had last week, Eric shared the extra digit might be best used for ad creative identification.
💎 Web-to-app conversion is for Safari only. You won’t get it from Chrome, Firefox, or a standard embedded webview. It will most likely be possible to get it from Safari controller because WebKit is active there by default.
⛏️ Going Deeper: Eric believes Apple might have limited it to Safari because app install campaigns show up when you google something. So Apple decided to treat other browsers like different apps and probably implemented SKAN 4.0 at the app layer for Safari.
Safari penetration is high (AppsFlyer mentions 93.65% of iPhone users have Safari), but it’s not universal.
💎 This is a system reset for SKAN. You can do the bare minimum to support the functionalities of 4.0 but to take advantage of the additional signal options, you have to rebuild your system from scratch.
💎 Three criteria have to be satisfied for SKAN 4.0 to work:
Users must be using iOS 16.1 and above
The app that is showing the ad AND the app that is advertising need to have built their app with the new iOS 16.1 SDK
The ad network needs to support triggering their ads with the version 4.0 ad signature (final gate between broad adoption)
💎 The fallback to SKAN 3.0 is seamless for ad networks: if they don’t update their signing mechanism, they don’t have to worry about SKAN 4.0 messing up your system. This means we will have two different versions of data for the same advertised app for a long time.
💎 The lack of fingerprinting enforcement is not creating any urgency for ad networks to move to SKAN 4.0 overall, which might slow adoption.
⛏️ Going Deeper: Eric believes that once Apple cracks down on fingerprinting (he expects this to happen by the end of the year), there will be a race to adopt SKAN 4.0. Both he and Alex seem to think this could take around six months.
💎 Ad networks most likely want to wait to see what Facebook does before investing time into building out their adaptation of SKAN 4.0 (e.g., how they treat conversion values, a recommendation for the locked conversion value window, etc.): advertisers will not go out of their way to go with a different adaptation than Facebook.
⛏️ Going Deeper: if you don’t yet have Facebook campaigns running, should you wait for all this to be sorted out?
I asked Alex what he thought, and he said there isn’t any reason to wait: we don’t know how quickly Facebook plans to introduce SKAN 4.0, but he’s pretty sure they'll fully abstract the details in any reporting you'd see (either on their platform or via an MMP), and that we might not even know when they flip over to the new version.
Alex believes that the likely change is that once Facebook has SKAN 4.0 implemented, we'll get somewhat better data granularity in the reporting (more breakdowns by campaign and ad set, better quality modeled conversions, etc.).
💎 There will be a few very tech-forward advertisers that see absolute optimization of SKAdNetwork as a competitive advantage and might dedicate a team to it. For everyone else, this is annoying and is an opportunity for products to facilitate the implementation.
Paid UA: postback period, early-stage vs. late-stage focus
Let’s go back to some higher-level growth stuff.
I liked the insights from Simon Lejeune (VP Growth at Wealthsimple) at MAU (cf. Growth Gems #74), so I started digging a bit more and found Paid Growth with Jesse Pujji and Simon Lejeune from Demand Curve’s Growth Summit.
I didn’t know Jesse Pujji (CEO at Kahani), but he also had some good perspectives.
💎 Focus on the payback period rather than LTV. Be conservative.
⛏️ Going Deeper: Jesse shared that the CAC/LTV ratio or payback period you set is a business decision. Both might be possible but will lead to different activities.
💎 Start the conversation by aiming for a first order/month payback period, and look at what could make this happen: a promo, an annual plan, etc. Work backward from that until you find a reason to extend that payback period (e.g., trying something new, growing faster, etc.).
💎 People make a false dichotomy: you can grow or be profitable. But some initiatives can help you grow and be profitable. Since you have limited time/resources, focus on these first. Example of changing this “gravity law”: Shopify “take the quiz”. Influencers used to be one, SMS is still under-leveraged, etc.
💎 If you see installs coming from content creators, and they have the right tone/concept, reach out to see if you could use it as a paid ad or copy/paste the idea to recreate it.
💎 In an early-stage company, you’re trying to find what works. The point is not to test because testing is a means to improve stuff that works.
💎 Offering new/additional products is a huge area for growth. Eventually, the growth from new customers decreases, and when that happens, cross-selling can keep you growing.
💎 You can build custom plugins on Figma to send creatives directly to Facebook/Snapchat/Twitter ads via API connections.
And before I leave, a quote that fits quite well given today’s edition, from the incrmntal podcast:
“An MMP used to be a commodity. Now it’s more about education.” - Thomas Petit (Growth Consultant)
See you next time. Stay savvy!