On May 21st, the share of conversion values (CV) with null entries in Apple’s SKAdNetwork postbacks changed dramatically.
While some media sources experienced a drastic drop in CV null values, dropping from a share of 80% to 20% over the course of one day, others demonstrated the opposite trend, displaying a spike from 0% to over 30% CV nulls:
The data, based on 47 million postbacks from 10 major media sources during the measured timeframe, suggests that Apple changed the logic that determines its privacy threshold on May 20th. As a result, an immediate impact on the CV null rates was seen the following day.
It’s important to mention that since Apple has yet to communicate anything regarding this matter, we can only make a confident assumption that this is indeed the reason.
In this blog, the second in our Inside SK series (exploring the ins and outs of SKAdNetwork), we’ll provide our take on this dramatic shift, its potential cause, implications, and recommended action items for advertisers.
The likely culprit: Apple’s privacy threshold change
As mentioned above, we believe Apple updated its privacy threshold on May 20th.
But what is this elusive privacy threshold, you ask? Great question. In its documentation, Apple only states: “The postback may include a conversion value and the source app’s ID if Apple determines that providing the values meets Apple’s privacy threshold.”
What does meeting this threshold actually mean? We can only assume by connecting a few dots at this point. More on that a little later on.
There’s a strong indication that Apple’s privacy threshold is essentially an attempt to prevent any way of identifying or even coming close to identifying a unique user.
In an aggregated solution such as SKAdNetwork, the lower the number of postbacks, the higher the chances to identify a specific user. Therefore, when granularity is as low as publisher level on a per-day basis, it’s possible for only a small number of postbacks to be included.
In such a case, when Apple believes there’s a risk of identifying a user, it will mark all daily postbacks from that publisher as null.
Needless to say, this has massive implications for advertisers. Because it’s the only piece of data that is available on post-install activity from SKAdNetwork, it informs predictions of future value, and therefore campaign optimization. Without CV, advertisers are left with install data alone, which carries very little weight or value without the needed post-install data.
Why did some networks experience spikes while others sharp drops?
We believe this is because each network has its own internal logic when it comes to figuring out the privacy threshold. They offer their advertisers guidance accordingly; each with its own set of campaigns hierarchy, ad-sets, publishers etc., so any change, such as the May 20th event, can bring about significant differences between networks’ performance.
For example, an ad network with hundreds or even thousands of publishers will look at the data on a publisher level, which can include very few postbacks. Other media sources like social networks, analyze their data at ad-set level because they have just one or a few publishers.
Facebook, for example, has recently advised its advertisers that the privacy threshold, as it applies to Facebook, is 128 installs per day per campaign.
Overall, and beyond the fact that Apple limits SKAdNetwork campaigns to 100, undoubtedly a change in its own right, the privacy threshold creates added complexity for networks. Supporting this, several ad networks we’ve approached on the matter acknowledged the problem, and confirmed they were working to reduce their share of null values.
Fresh insights on Apple’s privacy threshold
To help shed more light on the recent privacy threshold modification, we compared the share of CV nulls with the number of postbacks per campaign, and then again with the number of campaigns.
The data presented is based on the DAILY number of postbacks per app, campaign, and network. It does not, however, include the source app ID dimension, which we believe is included in Apple’s logic. Since we don’t know which other factors are involved, these numbers should serve to highlight general trends rather than specific benchmarks.
The data above shows a clear correlation between CV nulls rates to the number of daily postbacks per app, per campaign, where the percentage of nulls begins a dramatic drop from 80% to 0% after exceeding 10 postbacks.
On the flip side, when we look at the CV null ratio compared to the number of campaigns, we can see that the percentage of nulls rises from around 10% to more than 30% roughly after 57 campaigns. Although this is not a consistent trend, evident by the sudden dips around 61 and 71 campaigns, it is an upward shift nonetheless.
Recommended next steps
- We recommend that advertisers revamp their campaign structure to favor a high volume of installs per campaign across all networks they advertise on, and follow network guidance where applicable (e.g. Facebook).
- It might be beneficial for networks to seek clarity from Apple, so that they can update their campaign taxonomy recommendation accordingly, but also receive a proper heads-up when logic or values are updated in the future.
- Advertisers should be agile in executing on recommendations such as campaign aggregation, moving away from hyper-segmented campaigns.
The final word
Privacy thresholds are an important component in today’s privacy-centric reality. Enhanced transparency about how, what, and when they are put in place can help marketers optimize their campaigns while persevering the new privacy standard.
To ensure maximal relevance, we’ll be sure to update the post as more information becomes available, and continue to closely monitor, analyze, and shed more light on this and future trends.