iOS 14.5 aftermath: questions answered on SKAN, MMPs, privacy, and App Tracking Transparency
We’ve had a month with iOS 14.5. Time enough for at least some initial insights from mobile marketers and growth leaders about how it’s going, what’s working, and what’s still a complete minefield. (One of our recent guests used a slightly different term.)
We asked some of the best for their early reviews:
- Jayne Peressini, Senior director, marketing & growth, EA
- Paulo Esteves, Director of UA, Gameloft
- Eran Friedman, CTO and co-founder, Singular
- Phil Crosby, CPO and co-founder, Liftoff
Jayne and Paulo oversee millions in user acquisition and growth spend for global leading games and brands. Check out their thoughts, plans, and strategies in the full on-demand webinar here.
But we also had some questions during the live webinar that we could not answer in the time allocated. Here are those questions, along with answers.
iOS 14.5 aftermath: your questions answered
1. Can you explain more on “SKAN and MMP”?
SKAN is SKAdNetwork, Apple’s framework for privacy-safe deterministic attribution. MMP is an acronym for Mobile Measurement Partners, companies like Singular that have special access to major platforms like Facebook, Snap, Pinterest, Twitter, TikTok, and more to provide third-party measurement services.
2. Can you talk about the percentage of users who even allow apps to ask for consent?
We actually just published some data on that last week. A global average of 83% of people allow individual apps to ask for tracking permission.
(See that data, country-specific ATT and iOS 14.5 numbers, and much more information.)
We’re updating this at least every couple of weeks. Re-check the Singular blog regularly for new information.
3. Any feedback on PCM (private click measurement)? What kind of data is sent and when?
Private Click Measurement reports the kind of engagement (for example, a click), the source site and source ID, an “attribution” value for the website that wants to attribute incoming clicks, and some trigger data: a four-bit value that encodes up to 16 triggering events. PC also supports an eight-bit Campaign ID (values of 0-255).
Similarly to SKAdNetwork, the website postback/pixel is sent after a random timer of 24 to 48 hours.
This is currently app to web, not web to app. The Webkit team behind PCM says it’s interested in web to app, but hasn’t found a solution yet.
4. 19.4% of all iOS users have both seen and allowed tracking permission … what do you think users factor when giving consent?
In our latest data that’s now very slightly up to 19.7%. We’ll continue to monitor this over the next weeks and months.
- Men are more likely to allow tracking than women
- Younger and older people are less likely; middle-aged most likely
- Knowing and trusting the company asking for tracking permission is the most critical component of the decision
Also, however, education is key. People simply do not know what “tracking” for marketing measurement means, with many equating it to marketers knowing everything that they do online.
5. What does SKAdNetwork traffic look like? Have costs been impacted?
We’re between 60-80% SKAN-ready in terms of traffic, depending on who you talk to. We’re seeing a lot of fluctuation in CPMs as networks adjust to new methods of targeting, but there’s definitely been some spiking here and there.
6. We’re a B2C subscription app using app install campaigns. We currently measure our CAC:LTV ratio, with CAC being cost per paid subscriber. This will be pretty impossible for us as our 7-day free trial means we can’t set ‘trial convert’ as a conversion value. Has anyone navigated conversion events too far down the funnel for tracking via conversions? If so, what have you tried … what’s your strategy for alternative CPAs?
This is a really tough one.
Given that Apple is not going to change SKAdNetwork for this use case anytime soon, there are a number of options:
- Shorten your trial period (which, of course, has implications for people who test your subscription plan and for conversion rates while using it. Some of those implications could be positive, though.
- Choose channels that allow longer than 24-hour conversion periods so you’re not limited to one day of free trial.
- Find good proxies for conversion to paid. The best time to test this was 12 months ago; the next best time is now. But … go back to your data over the past year(s) and look for indicators that correlate with successful conversion to paid. Then test marketing optimization based on those indicators.
- Abandon deterministic measurement entirely and move to an incrementality model that takes in all your aggregated campaign and marketing output data and correlates it with trial launches and trial-to-paid conversions.
Not all of those options are easy or pleasant. My personal suggestion: find something in and around option three while working on long-term solutions in option four.
7. Can you speak to the percentage of inventory you’re seeing that is Do Not Track on iOS?
Our latest data on that is about 17.5%. See more details, including by country, here.
8. What is the privacy threshold right now?
No one outside of Apple knows the exact privacy thresholds. What we can say with confidence is that in late May, Apple significantly changed privacy thresholds so more data is available.
Short version: if you have campaigns with low numbers of installs — around 20-30 — you went from seeing maybe 10% of postbacks with conversion values to 80-90%.
9. Between SKAN data vs IDFA opt-in, which one is more accurate/reliable?
Tough question. This is dynamic and likely to change as time passes since on the one hand SKAN supports all iOS users regardless of their consent, but on the other hand not all publishers are SKAN-ready yet, and most of them don’t support view-through yet since it was just added in iOS14.5.
But as time goes by, SKAN is likely to become the more accurate source of truth as the number of Opt-In IDFA users decreases and SKAN’s coverage increases.
It’s important to note that SKAN postbacks go to ad networks, which then redirect them to MMPs like Singular. Singular aggregates all that data from all your partners in one place, checking it for accuracy and fraud.
Realistically, we know that SKAdNetwork underreports data, particularly for organic installs. Conversions may take longer than a week and never get reported. Web to app flows aren’t supported. Privacy thresholds will block some conversion data. Other limitations in SKAdNetwork may allow more installs to never become visible.
Ultimately, what you need from your mobile measurement partner is what Singular is doing: bringing together SKAdNetwork data, IDFA data from earlier iOS versions, and IDFA data from 14.5+ versions where people agree to tracking, and all your data from marketing campaigns … and build a comprehensive picture of inputs and outputs, paid and organic efforts as well as results, and provide an intelligent framework for optimization and allocation decisions.
10. How are you checking install amounts and conversion values reported by SKAdNetwork? Basically, what do you compare against what?
The key is to collect conversion postbacks from all the networks you are working with. The collection process is done differently for each network. Some networks share the raw postbacks, while other networks, like Facebook, only share them in aggregate via an API.
That’s why an SKAdNetwork aggregation solution like Singular is needed. Singular collects all the postbacks and then decodes conversion values back to meaningful events or revenue metrics, which enables performance analytics, comparing performance across channels and campaigns, and of course optimization.
11. Why not just assign one creative to one campaign in order to isolate the creative you’re looking to test? It’s probably more work but might replicate what you used to do.
That’s totally an option, and it’s one that some people are using.
But it comes with a significant downside: killing the ability of the network to run creative selection based on creative performance metrics like engagement and CVR. This means you as a marketer will run significantly longer cycles and do more work to figure out which creative works best. And of course, it’s harder on networks like Facebook, which only allow you to use nine of SKAdNetwork’s available campaigns.
(One of the reasons Facebook wants so many is to do creative optimization.)
Using tools like Singular’s Creative Analytics you will still be able to analyze the upper funnel performance per creative, like impressions and CVR, but you won’t be able to utilize SKAN for ROAS. For analyzing creative ROAS, you would need to run campaigns with a single creative or rely on cases where user-level data is provided like Android and iOS users who provide double ATT consent (on both the publisher and advertiser app).
12. Do we see any data on which apps/categories/developers see the higher opt-in rates?
Option rates generally seem to be in the 15-25% range, with the most common rate around 20%. We’re seeing some variation in verticals, but much of it is likely due to different rates of asking for ATT between verticals.
We’ll see more data on this over time.
13. For conversion values, is it better to track fewer/earlier funnel conversion events (so you can get SKAN back faster) versus deeper events that may take longer to get back?
Pretty much all of the data I’m seeing from top marketers suggests absolutely yes. Rovio is working on a three-day conversion cycle, and many others are defaulting to D1, thanks to how some of the major platforms require it.
Check Jayne’s and Paulo’s full answers in the webinar.
14. How accurately are you receiving purchase conversions?
We haven’t run the data on this yet, but plan to. Two sources of leakage are privacy thresholds and conversions that occur after the SKAdNetwork postback.
Ultimately, to find that leakage, you’ll have to isolate campaigns, check the data you’re getting, compare it to actual results in your app from measured cohorts over time, and assign a multiplier.
15. How do you know which conversion settings are better?
Singular offers a way to toggle between conversion models: revenue, conversions, engagement, or funnel. You can also model conversions without code changes in your app: speak to a Singular representative about how.
16. How will these updates affect buying traffic based on CPA on iOS?
Put simply: they make it harder. There’s less data, delivered slower and at a lower granularity. They don’t make it impossible, but you’re going to have to work with SKAdNetwork and Singular to get the most out of the data you do get.
17. What is the future of Apple Ads as a channel?
That’s more for Apple to say than Singular. What we can say is that it’s been an extremely high-quality source for app installs in the past, and that is only likely to be more true in an SKAdNetwork reality.
Apple Search Ads, of course, does not use SKAN. Apple views data it collects on iPhone and iPad owners as first-party data, and therefore not something that is shared across companies and apps, and needing to be obfuscated via SKAdNetwork.
That gives Apple very precise targeting capabilities … and makes ASA-sourced installs generally very high quality.
18. Question for Jayne: What type of influencers are you looking at to push your titles? And how do you measure performance with influencers?
From Jayne Peressini: We look at influencers as an ecosystem. Top influencers are handled by our partnership and competitive gaming teams while mid and long-tail influencers are something we manage. It’s performance based just like the other channels we buy on. And similar to other channels, we look at trying to reduce overlap of influencers and their communities so we get as much reach and scale as possible. We measure on unique reach + ROAS.
Looking for more insight on iOS 14.5 and SKAdNetwork?
Singular was the first marketing measurement company to support SKAdNetwork, and we have the most comprehensive solution for growth professionals looking to boost their apps. Book some time with us today, and we’ll learn what your goals are, and outline how Singular can help you achieve them.