On April 26 2021, Apple officially began their rollout of the App Tracking Transparency (ATT) prompt, giving iOS users the choice to opt in or out of data tracking on mobile apps. For Facebook Ads specifically, this change has massive implications for how Facebook optimizes, targets, and reports on web conversion events.
Big changes to Facebook Ads include:
As performance marketers who analyze these data points every single day: this roll out has massive implications for how we do our work.
So at Thesis, we spent months in preparation. We completed all the necessary action-items for our clients which included verifying domains, prioritizing pixel events via Aggregated Event Measurement (AEM), and installing server-side tracking via Conversions API (CAPI).
And when the day of reckoning finally came… it was oddly still. The rollout still has not occurred in full (Facebook refuses to share exact numbers), but what we do know is that the outlook appears… rough.
Only 5% of iOS users are opting in to tracking.
Many questions remain. So as best as we can, knowing full-well that our responses are biased and that we still don’t have all the answers, we’re going to give you our perspective on the top 7 questions that we’ve seen about iOS14 and Facebook Ads.
As long as you’re optimizing for purchase using a Conversion campaign, Facebook’s algorithm fundamentally knows (better than you) which devices to put your money towards to get purchases. So, officially, our stance is “no”. This is not a high confidence test.
However, the thing that could complicate this question is your attribution window.
As of May 2021, the only attribution window that includes iOS opt-out users (using AEM) is 1 day click. If you’re using any other attribution window (1DC1DV, 7DC, or 7DC1DV) your campaigns are not currently accounting for iOS users that have opted out of tracking.
Which means that if you’re using 7DC, and you have a high amount of iOS users that are converting on 1DC then you could be losing tons of signal and purchase data.
Because of this caveat (which Facebook aims to change soon) I could see a world in which we would test splitting these devices out. We’ll be waiting a few months until we’ve gotten word that more people have received the ATT prompt, which has not rolled out in full.
As a part of our overall Thesis-strategy (which is considered a modern-day-best practice), we focus WAY less on audience testing as a whole. At the scale point that many of our clients are at, we continue to find the best, consistent results from utilizing broad audiences and focusing on creative testing.
With that said, we have seen clients find short-term efficiency from utilizing interest targeting in 2021. These clients also use detailed targeting expansion, which is becoming the norm for interest targeting as a whole. From our perspective: you’re telling Facebook where to start with a certain subgroup of people, but it’s still broad.
Lookalikes, on the other hand, haven’t been as successful the last few months. But we have seen short-term lifts from using lookalike expansion and the stacked approach.
This is a really hard question to answer. Some agencies and media buyers build their entire strategies around using certain bidding strategies, so no matter what we’ll be giving a really biased answer here.
With that said, we have found it really hard recently to find efficiency using cost caps. We’ve tested it on a few clients and found that it choked spend. Since many of our clients need to maintain volume, they need to spend a certain amount of money every single day to maintain revenue and business goals.
We have found that bid caps were better suited to maintaining those spend goals, while still helping to control costs.
Again, this is incredibly subjective. And as it relates to iOS14, there appears to be no major changes here as of yet.
If you haven’t already, look into consolidating your retargeting campaigns. So instead of breaking out website visitors from add to cart or view content, I would instead combine all of these audiences into a single ad set. This will help you exit the learning phase quicker since they’re not in separate ad sets, which each need 50 conversions per week.
Another thing to note is that your social engagers audiences will be unchanged. So even if you found that instagram or Facebook engagers didn’t work for you in the past, give this another shot. We prefer to combine Facebook and Instagram engagers into a single ad set.
An additional thing you can do to bulk up your retargeting is to utilize custom audiences, like uploading an email list of non-purchasers. Facebook has recently announced that match lists via custom audiences will continue to match as usual. A few things to note here:
This is a really tough question to answer. The question of attribution is heavily reliant on your buying cycle: do most of your customs convert on day 1? Day 5? If you have a longer buying cycle, then you’d want that 7DC attribution so that you could get a more accurate view of the conversions your ads are driving.
However, the reason why this is such a hard question to answer is that in Facebook Ads Manager: attribution and optimization windows are the same thing.
So if you are using a 7DC attribution, you are telling Facebook to find people who will convert within 7 days of clicking on an ad. Conversely, if you told Facebook to optimize for 1DC… the algo would concentrate on finding people who are likely to convert within a 24 hour period.
If you think that you are on the incorrect attribution window, you could test 1DC vs 7DC over a period of a few weeks, using a 3rd party tracking like GA to analyze that data to see what is actually working.
And if you decide to change your attribution window… be sure to recalculate your CAC and ROAS goals.
Something that we are openly bullish on is NOT using 1 day view for any attribution window. Again, a biased point of view, but we’ve done incremental lift tests that have shown in most cases, these view through conversions aren’t incremental.
This could really hurt your brand because while Facebook is super smart, it will take the easy way out when it comes to attributing conversions.
The short answer is yes. With the caveat that there is that there has always been a discrepancy between Facebook Ads Manager and GA data, so expecting a 1:1 comparison for the data lost is not what you’re going to get. But GA is better than nothing and can be used to help you make sense of what is happening with your campaigns.
For now, we’re leaning on GA to identify trends of our ad performance as a whole. We triangulate the GA data with what we see in Ads Manager and Shopify. We’re not monitoring this for day-to-day data on singular ads at this point, so we’re still using Ads Manager to evaluate ad-specific performance.
Another thing we strongly suggest in regards to tracking is to install CAPI. This will help identify more conversions that you will be able to track in Ads Manager via server-side tracking.
As of May 2021, the ATT prompt has not rolled out to all users. Because of that we have not felt the full impact of iOS14.
Because of that, we have not had to change our approach to creative testing, which leans heavily on testing our ads against a broad audience so that they have the best potential to scale.
Interestingly enough, this might not be something that we have to change. Our contacts at Facebook said that this spectrum is where they suspect we’ll see the biggest degradation of results:
What this means is that Facebook suspects that broad audiences overall will be affected the least by the iOS14 rollout.
Since that is where we do most of our creative testing and we have not seen a big swing in results on broad for a majority of our clients, our approach to creative testing remains the same.
With the release of TikTok’s Creative Center, we’ve been taking a deep dive into the Top Ads section to find some trends in what is working on the platform. It’s a great resource for brainstorming new ad ideas for your ads.
At the suggestion of our Snapchat rep, we tested Snapchat's Lifestyle Categories that an account indexed strongly with to see how they performed relative to our standard broad targeting.