IDFA Changes: UA & Attribution

How can advertisers prepare to manage UA in iOS 14?

Adam

[💎 @06:07] Look at the way you're currently marketing and try to build a framework using contextual modeling (using device graph, app bundle ids, time of day, IDFAs, etc. to signal you're on the right track). Train these models today based on ML. It will help you in the near-term (however the less IDFAs are available in the future the more your model will decline).

Angela

If you're able to evaluate the data today (IDFAs, last click, etc.) and how you can decide media buying based on this model it will help you for the transition to non-deterministic.

Claire

It's been predicted but nobody thought it would happen so fast. Understand what is happening and that you won't have user-level permission data post iOS14.

[💎 @08:37] As an app publisher, communicate with all the teams within the company: management, data science team, product team to integrate/update all the SDKs and make sure everybody is ready.

Angela

Today we call it deterministic but there are some things we already do not understand.

Kalia

[💎 @09:49] Embrace the culture of experimentation and understand on what you are getting incremental lifts.

At the high level, reset expectations with your team and create an environment of transparency. You want all internal teams in the same mind space.

Charles

MMPs' job is to server advertisers

[💎 @11:47] Advertisers want to interpret signals: deterministic signal, probabilistic signal. SKAdNetwork (even though it is delayed and not real time) is going to be balanced with other signals when making decisions.

If it's SKAdNetwork inventory you still want impressions and clicks to be trafficked, even if it's probabilistic because you want to integrate this as part of the input you have.

Adam: Is there a way to really leverage SKAdNetwork?

Charles

[💎 @14:32] It is not an either/or between SKAdNetwork and probabilistic signals. Both have to be synthesized in a tool for advertisers.

You can build custom tech yourself but the ones the companies that are the best equipped are the ones that will be able to spend. MMPs are focused on the synthesis (maybe just 5-10% SKAdNetwork inventory) and cross-referencing the signals.

Sources of data and limitations of SKAdNetwork

Charles

[💎 @17:04] Kochava is going to be linking conversion values to the post-install events that advertisers are already used to. So the question for advertisers is: what is most important to track for you with these 6-bits?

Security is also important. Example: if it becomes public knowledge that "play video" is a particular bitstream that is important for Disne and if it's possible to synthesize that conversion value that is sent, it could be a high conversion value that is not authenticated. How do you protect the advertiser from exposing private information that they don't want to have exposed, but at the same time get them enough information so that they can scale?

There will be a lot of innovations coming from the space.

Claire

At the moment advertisers run a lot of campaigns with split tests, etc. and look at the creative level.

SKAdNetwork only provides 100 campaigns and 1 conversion event that will be shared. The way we used to work with deep events will be really tough to have. Aiming for ROAS will be hard.

Adam

[💎 @20:32] SKAdNetwork's timer is per event so if you have several events together (like install, conversion, completed tutorial etc.) every event resets the clock. You might end up with 20 events with a 5-hour lag between each of them yet you're not going to get a post-back bundle for all these events until 3-4 days from now. Bottomline: if you want a lot of information, it will not be as usable as it is now for campaign optimization.

Applauding this decision are the ones that are looking at long-term effects of channels, not performance marketers that are used to looking at D7 ROAS and more immediate metrics.

You still have Android and can leverage past-experiences. New guys coming in will struggle to scale and it will be very contextual-based.

Angela

[💎 @21:47] You have to look at the bigger picture (no longer last-click revenue) and find how to interpret all the signals and how you read your data. Marketers have to be more educated on data and how to take long-term decisions.

Incrementality

Adam

ROAS is the pinnacle but also created the beast there is today and a "whales chasing" problem.

What ROAS created however is the ability to choose one publisher over the other.

[💎 @23:42] Until now, ROAS was used to decide between channels. The concept of incrementality (lift over base) will now become even more critical.

Example: what happens if I turn Facebook down 40% in Germany? Do I see incremental value, a decline in revenue, or things relatively the same?

Charles

Box-balancing the signals to come to a conclusion in terms of incrementality, regardless of platforms.

[💎 @25:27] CTV is like mobile was back in 2011: there is tons of inventory and it's not easy to track. Everyone is going to be buying on CTV.

Advertisers' media mix is going to change.

Apple Search Ads

Claire

[💎 @26:22] All advertisers are going to be shifting to ASA: there's going to be more cannibalization of paid vs. organic and also much more competition.

Need to adapt marketing strategy around incrementality and if you have older products you need to do some offline acquisition.

Tooling

Kalia

For Disney Streaming Services, preparing is about what exists in the space to understand what's happening and what they can do about it: either build or buy based on the pain points of the marketers in the team.

Team thinks in the long term and don't focus too much on the granular and focus on the trends in all channels.

Question is: how do you consolidate all the information in one place to make it easier to analyze? Will there be multi-tools to streamline it for advertisers.

What does iOS 14 do to the media mix model?

Adam

Some companies are just not going to put the pop-up in their game because they don't want the loss in DAU so groups of users will go "dark". The pop-up is wrongly worded and "predatory".

[💎 @32:35] People will start looking at sub publishers. We will see a come back of private marketplaces allowing to find apps that have interesting users for you. It's going to go back to "I know your game has good traffic because it has the same kind of users as my game".

Some companies are about: build yourself a contextual model today to be able to determine things in the future. But it's going to be up to gaming companies and if they display the pop-up. If they don't then it will truly be a world without IDFAs.

Charles

Will the pop-up display on the demand side or the supply side?

Claire

It's essential to highlight that we're shifting from behavioral targeting to contextual targeting.

[💎 @35:17] We have seen an increase in ARPU in the last 2 years thanks to AEO. Working with AAA ads (Automatic Ad Creation - dynamic ads) without any audience or suppression list is going to be a big shift in how you do your user acquisition.

[💎 @36:00] A lot of budget is spent on AEO/VO campaigns. Right now you can just say "we want X% ROAS". How will Facebook know how to reach that target without knowing the information? 50% of the CPM on the Facebook Audience Network is related to behavioral targeting and we need transparency on the amount of data networks will actually have after iOS14.

Remarks/suggestions

Kalia

Embracing environment of experimentation. You have to reset and think uniquely about testing things and solving problems.

Angela

Excited about the changes. You have to re-invent yourself.

Adam

The world we're about to step into has already existed. Don't expect what you've expected for the last 6 years because it won't exist.

• Expensive CPIs, not a lot of indicators of value, you will spend money in places you probably shouldn't
• Companies that have hardcore games or require a lot of ad revenue are going to suffer immensely on iOS
• [💎 @40:57] iPhone 6 and below will not be able to update to iOS 14 so you might be able to leverage that.
• Explain to your boss what it's going to look like. Don't expect what you've expected for the last 6 years because it won't exist.

Q&A

Derek from Tinder - Is finger printing going to be part of the solution? Will it be allowed?

• Charles: Kochava is intentionally deviating from the language of finger printing because the spirit of what Apple articulates as finger printing is an identifier that lives on the device and can be regenerated on the device bases on attributes.
• What Kochava does id take signals like IP, user agents and other contextual flags to create something "probabilistic" with an attribution waterfall configurable by the client. The advertiser will be able to define which signals/input they want to use: the marketer owns the end decision. Example: IP not used.
• Adam: [💎 @46:46]  Apple will not allow what we consider to be finger printing as a viable solution and say that it is not in the spirit of what they are trying to do. Finger printing will exist out of the gate but Q1/Q2 next year users that never opted in will complain they are being tracked and it will signal the end of fingerprinting.
• Andre from Admiral Media - Apple has a strategy that has not been discussed at all. They will not allow anything close to fingerprinting. We are limited to 100 campaigns and will be limited in terms of the data we can work with.

Any advice on naming conventions?

• Angie: you'll have to be more thoughtful of the context and how you do things vs. fingerprinting.
• Claire: Apple wants us to use Apple sign-in in applications, to use Apple pay, all purchases in-app etc.
• The naming convention will need to be optimized to better track campaigns.
• Better to embrace SKAdNetwork because nobody else than Apple will have as much data. It's an Apple device so follow what Apple is putting in place and says.
• Charles: important to differentiate probabilistic vs. finger printing, and that's when configurability of the waterfall is important. There is a lot of inventory already where advertisers are trying to understand what is happening. SKAdNetwork is one signal but even if it's "SKAdNetwork only" there will always be context around that.
• Campaign naming is a function of who owns the SKAdNetwork id that's serving the media. The way Apple is presenting it you are supposed to be an ad network or a publisher, not a DSP.
• Adam: [💎 @54:56] Mobile web is a great way to counter balance the effects of iOS 14 on app campaigns and you're going to be able to take what you learned to that environment. However it is really difficult to target these users. How do you find the right users for your product on web?

How do eCommerce businesses will deal with that? Email address, IP address, card data, etc. Can we still use that?

• Charles: [💎 @56:22] The Apple documentation now references SSO (Single Sign On). If the ATT pop-up is a pre-requisite to any SSO element (third party) then everyone will be showing the pop-ups and everyone might start agreeing. This MAY bring the adoption rates way up.

• The part of the documentation Charles is referring to

Hussein from Liftoff - How do we deal with audience suppression post iOS 14 for user acquisition campaigns where you might be advertising to your users?

• Kalia: IDFA is not the only thing you can use to suppress ads. You can look for companies that bridge this gap. Do you do offline identity matching? For clients you already know converted, how do you find them in other places? It's going to become a bigger industry but it's tricky to do.
• Charles: every advertiser should have their first-party data safe and secure and be able to do consent-driven, privacy-based matching with the first-party data of other partners.
• Claire: SAN won't get information from other applications but within the environment you will still have everything. Facebook has Whatsapp and therefore your phone number to be able to target you and advertisers will now ask information to users during onboarding to have some data about them.