How to run iOS campaigns with Google

Background

I ran the App Ad Sales team at Google for ~6 years and during that time I was the go-to iOS person. I spent a lot of time understanding the intricacies of SKAdNetwork and the impact it had on advertisers, publishers and networks. Most advertisers I worked with were unaware of how unique Google is when it comes to iOS and what they needed to do to be successful. So, if you’re going to run app campaigns with Google, here’s what I want you to know.?

1) Focus on videos. Lots of videos.

Unlike Google’s Android inventory (which includes the Play Store), the majority of Google’s iOS inventory is video. And there’s LOTS of different formats. YouTube for example, has unskippable 6 second videos (bumper ads), YouTube Shorts (akin to TikTok), and the traditional skippable video formats. Then there’s AdMob inventory which is a mixed bag of video formats.

Because Google’s App Campaigns product does not allow advertisers to dictate where their ads serve, it’s imperative that you have ample video coverage to accurately test iOS campaigns on Google.?

2) Understand the shortcomings of ATT for Google

Before I dive deep into this topic, let’s do a quick refresher: Networks can still receive user IDFAs if users opt-in to ATT on both the publisher's app and the advertiser’s app. These users/conversions are considered “deterministic” because we know exactly who/what they are and where they came from. Ah, the good ol’ days!?

In talking to a lot of growth teams, I hear that they use deterministic data to extrapolate network performance. Here’s why I tell all my clients NOT to do this…

A client recently brought me on to diagnose their Google iOS campaigns as they “were not performing as well as Meta campaigns.” After learning they were using deterministic data to compare network performance, I dug into SKAN to see how the data compared. Their deterministic CPI showed Google’s CPI was 7x higher than Meta. But in SKAN, their Google CPI was actually 65% cheaper than Meta.?

So why the huge discrepancy? As mentioned in the refresher, deterministic data requires users to opt-in to ATT. Herein lies the problem: the majority of Google’s inventory either doesn’t allow for the ATT prompt or omits the prompt altogether. Let’s dig into that a bit more…

First, let’s break down Google’s iOS network. There’s 3 primary traffic sources: Search, YouTube and AdMob.?

• Search: The vast majority of Search inventory comes from mobile web, not Google’s Search app. Because this inventory sits in a browser and not an app, there’s no option to prompt for ATT and thus, no option for Google to collect IDFAs.?
• YouTube: Most YouTube traffic is through the app, however, back in 2021, Google made the decision to NOT prompt for ATT on YouTube. So, no IDFAs there either!
• AdMob: Since AdMob is a collection of publisher apps, it’s up to the publisher if they are prompting for ATT. This is the only meaningful source for collecting IDFAs on Google’s network.

If you’re going to run iOS campaigns with Google, know that Google’s deterministic data will likely look abysmal in comparison to other networks and you’ll need an alternate way of measuring network performance.?

3) Be careful how you test creative & keep NULL rates low

For the seasoned iOS advertiser, keeping your NULL rates low is a no brainer. But I’ve talked with countless advertisers who are unaware of the NULL implications, so I’m here to say it again!

NULL postbacks are SKAdNetwork’s way of saying “we think you’ll have a way of figuring out who these users are, so we’re only going to tell you that they installed the app and nothing else.” Which means these users could have gone on to take meaningful action in your app and you wouldn’t know.?

What I see with a lot of advertisers who aren’t familiar with Google App Campaigns, is they utilize campaigns as a way to test creative. I get it. Google’s creative testing options are not great (this was the #1 complaint from customers during my time at Google). But this strategy hurts you in that 1) it creates competing campaigns (you’re bidding against yourself), and 2) it produces more NULLs.?

So here’s what I recommend:

• Consolidate campaigns as much as you can
• Aim for 100 installs/day or 30+ post-install events/day?
• Use Ad Groups for creative testing, not campaigns

4) If you tested Google pre-October 2023, try it again.?

One thing that frustrated me with the release of SKAN 4.0 was how little the web-to-app tracking feature was mentioned. But it made sense why this was glossed over; almost all the other networks’ inventory is on apps, so web-to-app tracking wasn’t a huge deal. But that’s not the case with Google…

Google’s Search inventory is almost entirely on web. Yes, some people have the Google Search app, but for the most part, people use browsers for searching. Prior to October 2023, if you had any Search traffic in your Google iOS campaigns, none of those conversions would show up in SKAdNetwork.?

I can’t tell you the number of times I joined a call with a customer, only to reveal to them that 99% of their inventory was through Search and therefore SKAN wasn’t capturing any of Google’s conversions. SKAN 4.0 changed that and ensured Google was getting credit where it was due….at least in Safari. Chrome conversions still roam free!?

5) Test “On Device Measurement” (if you can)

If you’ve made it this far, you’ve learned that Google doesn’t have a ton of IDFAs to optimize and model off of. On Device Measurement (ODM) changes that. So if you collect user emails or phone numbers, here’s your chance to make Google work better for you!?

Like any ML-based product, Google’s App Campaigns function best when they have high quality data. ODM provides more of that data in an Apple-compliant way. I won’t go into all the technical and privacy aspects of ODM, but if you have a decent volume of user emails or phone numbers, it’s definitely worth testing.?

6) Don’t take Google’s word. Measure with SKAN.

Google is a self-attributing network (SAN), meaning it models conversions and reports on them in their platform. Using Google reporting to measure iOS performance is fine if you’re ONLY running on Google. But for most advertisers, Google is just one of many networks used for iOS user acquisition. So if you’re running on multiple networks, do yourself a favor and get familiar with SKAN.?

Here’s an example: Say you’re running iOS campaigns on Meta and Google, and using both networks’ reporting to determine performance. Meta’s platform reported 110 installs and Google’s showed 100. But when you look at your internal reporting, you only see 200 installs.

Does that mean that Meta took credit for too many? Or maybe Google’s number is inflated? You can’t really know, which is why you need to use a single source of truth for attribution.?

SKAN isn’t perfect and leaves much to be desired, but it does provide a (somewhat) level playing field when doing attribution.?

The Wrap Up

Google has the largest network on iOS but it’s often misunderstood because it behaves differently than other networks. So if you’re looking to test it out or understand it better, reach out and let’s see how we can work together.?

Author Bio

Ashley Black is the founder and owner of Candid Consulting, LLC, an app consultancy and marketing agency. Ashley has been in the AdTech industry for 15 years, and spent the last 6 years at Google leading the App Ads team. She has worked with thousands of apps across all industries, helping companies optimize their monetization strategies and improve marketing efficiencies.?

Ashley Black

[email protected]

candidconsultinggroup.com?