8 limitations of SKAdNetwork for mobile marketing measurement
SKAdNetwork may be the only game in town for mobile marketing measurement and attribution on iOS 14.5 and following for the foreseeable future. And it offers some significant benefits: deterministic attribution of app installs on the iOS App Store. But there are also some major SKAdNetwork limitations.
The IDFA, after all, offered attribution data that was deterministic too.
But it also provided device-level granularity. Deep linking functionality with deep linking measurement and tracking. Extensive post-install app events for conversion reporting and ad campaign optimization. Data on ad impressions. Data on ad clicks. And, of course, an identifier for retargeting and look-alike audiences.
That’s in the past.
But there’s still some value in understanding as much as possible about SKAdNetwork limitations so that mobile marketers can optimize around them. I spent a few minutes with Singular CTO Eran Friedman to understand those SKAdNetwork limitations and get some insight into how to maximize the data that SKADNetwork does supply.
8 SKAdNetwork limitations
1. Distributed postback data
Mobile measurement partners invented the concept of the postback for mobile attribution about a decade ago. It’s simply a digital notification of an event. MMPs collected them for all the different ad networks and media partners a marketer used, and marketers automatically had a centralized repository of app installs, attribution data, ad network contributions to their marketing success, and more.
Under SKAdNetwork, when someone sees an ad for an app, clicks on it, and installs the app, a postback is sent from their mobile device directly to an ad network. If you, like most mobile advertisers at scale, use 10 or 20 or even more ad networks, your mobile app marketing attribution data is now scattered across all those companies’ servers.
(Note: Singular has fixed that problem with our SKAN solution.)
2. Limited granularity
The old way of doing mobile attribution on iOS gave perfect granularity. As long as someone didn’t set Limit Ad Tracking on, their IDFA was fully available to advertisers, publishers, and all the various layers of the adtech stack. That gave you device-level data on ad views, clicks, installs, and post-install activity and conversions.
That’s still possible, but it requires app-by-app permission for everything to work as it used too. Getting IDFA permission is great, but if it’s only on one side of the advertiser-publisher relationship, it’s insufficient. You’ll need permission on both the publisher app and the advertiser app to make it really work.
In SKAdNetwork, granularity is limited:
- No device-level data
- No creative-level data
- Only 100 campaigns (and ad networks can use most of their for internal tracking)
- Publisher data
- Only 24 hours of post-install conversion data, unless you update the conversion time
- Only six bits of post-install conversion data
3. No retargeting
In the iOS 13 and earlier era of IDFA, you could retarget former app users or people who have your app installed but aren’t opening it. Using their IDFA, you could target ads with offers to them, incentivizing them to return, reinstall, or reengage.
In the iOS 14.5 era of SKAdNetwork and scarce IDFAs, retargeting is essentially toast. SKAdNetwork does have support for a redownload flag so you can know when you’re getting someone back, but you can’t really target them on other apps or via ad networks like you used to be able to.
4. No look-alike campaigns (or … look-alikes with reduced effectiveness)
IDFA made look-alike campaigns possible. Using your attribution data, app events, and conversion values, you could build lists of your best users or your top customers. You could then export this list of IDFAs and send them to your ad partners, telling them to find more users like these.
Because ad networks and major platforms like Facebook, Google, Twitter, Pinterest, Snap, and TikTok have a lot of data about ads that devices with those IDFAs have clicked on and apps that they’ve installed (plus more data in many cases) they could find you more people like the best users and customers you already had.
Guess what: no IDFA, no look-alike campaigns.
“Things like lookalike campaigns, for example, rely completely on the IDFA,” says Friedman. “So they’re just completely unsupported by SKAdNetwork.”
Now, platforms are doubtless exploring ways they can offer look-alike-like products to advertisers, but the reality is that without IDFA, there’s going to be some new level of uncertainty, probability, and modeling in the audiences that they assemble. So there’s an inevitable loss of granularity and deterministic targetability, which will degrade effectiveness to some degree.
5. No real multi-touch attribution
The good news about SKAdNetwork limitations is that Apple continues to iterate the framework. And in iOS 14.6 we got a little gift: loser postbacks.
(OK, that’s my private name for them.)
Most postbacks are “winner” postbacks: a notification that your ad network’s ad impression was successful. It generated a click and the click resulted in an app install. With iOS 14.6 and SKAdNetwork 3.0, however, we got a little gift from Apple: postbacks to up to five other ad networks whose ads were seen but not clicked on, or, at least were not responsible for the app install.
In other words: loser postbacks.
That’s good, but it’s not really sufficient for MTA. You need more data across more platforms and devices to really enable multi-touch attribution (though it will help in measuring incrementality). Full MTA probably requires data from the web, where we now have app-to-web measurement in SKAdNetwork but don’t yet have good web-to-app data, data from deeplinks, and data from entirely non-Apple and iOS sources.
The reality, of course, is that in the last-click-dominated mobile attribution world of the past decade, we didn’t have MTA anyways. We want it, and maybe even need it — everyone knows that last-click is a horrifically myopic way of looking at marketing measurement — but we still don’t have a real path to it with the current SKAdNetwork limitations.
6. Web to app measurement
With Private Click Measurement, Apple gave us a tool to measure app to web journeys in SKAdNetwork. That’s great, but we also need tools to measure web to app journeys.
The unfortunate part: that’s much harder.
You can measure app to web because you have a known environment: an app on iOS with access to the SKAdNetwork framework. On the mobile web you might be in mobile Safari, in which case Apple could insert some SKAdNetwork-relative code, but you might be in Chrome or Opera or Firefox or some other browser that Apple does not control.
Rock, meet hard place.
7. Fraud, fraud, fraud
The Apple postback for app installs on the iOS App Store is cryptographically signed. Great. But the payload of post-install conversion data about what happens after the install is not. Oops.
First off, post-install conversion data could be faked. (Of course, a fraudster would have to know your conversion schema to make it stick.) But secondly, there’s no country or geolocation data, so you could be getting “high quality U.S. users” who are actually in Kazakhstan.
And third, a fraudulent media partner could just replay legitimate SKAdNetwork postbacks repeatedly. Unless you’re regularly checking the details on your postbacks, you simply wouldn’t know.
In fact, Singular is seeing duplicate SKAdNetwork postbacks in the wild right now, Friedman says.(Here’s how Singular is fighting SKAdNetwork fraud, by the way.)
8. Data fragmentation (or siloing)
So imagine you’re all set up. You know exactly what you need to do for SKAdNetwork, and you do it. You’re getting the data you need, and life is good, right?
“SKAdNetwork data doesn’t live in a vacuum,” Friedman told me. “We’re going to have a mix, basically. You have SKAdNetwork data. You’re still going to have IDFA users … users who have opted in.”
And then there’s campaign data too … the data on the campaigns and spend and creative that you’ve set up with ad networks.
Marketers have to connect all that data to be able to answer questions that need answering: how much am I spending? What am I getting for all my efforts? How many total installs am I getting across all my partners? Adding all of that and tying results to spend is a significant challenge.
Part of the challenge: deciding what dataset you’re using to optimize campaigns. If it’s largely SKAdNetwork, can you leverage partial IDFA data for insights? And, how are you going to analyze creative … with SKAdNetwork campaign IDs? But if so: do you have enough data
Data silos complicate data analysis. Marketers need tools and guidance to properly tackle them.
Maximizing the value of the SKAdNetwork data you do get
So how do you minimize the SKAdNetwork limitations and maximize the value of the data that you are getting?
It starts with defining your conversion model, Friedman says. Then, ensure that you get clean data in a single place that you can analyze both separately and together with all of your other app attribution data. Analyze it for campaigns and partners that are doing well, and any additional insights you’ve inserted into your conversion models.
Also, however, validate your postbacks for both accuracy and fraud.
That all sounds like a lot of work, and it can be. The easy answer, however, says Friedman is to simply use Singular SKAN.
Need help with SKAdNetwork?
We’d be happy to walk through your plans, your strategies, and how Singular can help.