ABOUT OUR GUESTS: Jayne Peressini | Claire Rozain | Philip Weiskirchen | Gadi Eliashiv | John Koetsier

ABOUT ROCKETSHIP HQ: Website | LinkedIn  | Twitter | YouTube


KEY HIGHLIGHTS

☂️ What percentage of the market is good at SKAN?

🪢 Are additional measurement methodologies required?

🧶 Probable performance improvements in SKAN 4.0.

🦺 What’s a feature that SKAN 4.0 should roll out?

🧣 Google’s Privacy Sandbox and how it differs from SKAN.

🎒 MMM & Incrementality Modeling.

🧢 How measurement methodologies can be affected by organizational leadership.

🙇🏻‍♀️ Skills required to prep for the future.

🚵🏻‍♂️ Top takeaways from all panelists.

KEY QUOTES

Is SKAN enough for measurement?

Philip: To clarify, running SKAN campaigns allows you to bid for users with IDFA, providing reporting on results for both opted-in and opted-out users.

So it’s not like we are leaving IDFA traffic or IDFA attribution behind. I believe SKAN is enough if you have figured out the setup, but figuring out the setup requires advertisers, MMPs and also DSPs to work closely together.

There are too many sources of data right now.

Jayne: Having worked in the industry for almost 20 years, we did everything, only taking 30% of impression-based conversions and various other attribution approaches, but many of them ended up being highly erroneous. The campaign performance was so far from the truth.

So I think less is actually more in this case. It’s just a matter of agreeing on the source of truth.

When to go for additional measurement methodologies.

Shamanth: If you are at scale, and on multiple channels, you need additional measurement methodologies. If you’re small and don’t have too much ambiguity in what you’re measuring, I don’t think you need all that much.

How is SKAN 4.0 going to be different?

Shamanth: If you’re a small advertiser, you are going to get a cost conversion value when you’ve got none in the past. Larger advertisers will also experience fewer null conversion values. I think one of the bigger challenges has been the loss of signal, which SKAN 4.0 is going to take some steps toward addressing, even if it’s not completely groundbreaking,

Is MMM the preference now?

Shamanth: I think the vast majority of folks are using some form of non-deterministic measurement. I think that specific method matters less than the fact that they are making some effort to do this. We’re leaning into a lot of MMM, just because we use an open-source tool that we are very comfortable with and that we have gotten the results from. 

There are third-party tools out there, the folks who run incrementality, there’s quite a bit folks who do model data just as those third-party tools that do that. Now, all of these are broadly put under the umbrella of non-deterministic measurement, which I think a lot of the savvy folks are leaning into.

Sometimes attribution depends on the leadership decision.

Shamanth: I’ve seen turf wars in companies where different groups have wanted to measure things differently. I have said before that LTV is a completely made-up number, depending on who’s talking.

FULL TRANSCRIPT

John 

Hello, everyone. This is the Kickass UA in the Privacy Era webinar. We have an amazing panel, and we’re going to have some interesting discussions.

Is performance marketing dead? It has evolved significantly. 

What strategies will work in the coming years? 

Will MMM, SKAN, privacy sandbox, first-party data, or partner/ad network stats be effective?

Our panel of experts will shed light on these topics. 

We have Jayne Peressini, a UA Consultant now. She’s had major roles at Dapper Labs and Electronic Arts. 

Shamanth Rao, CEO and Founder at RocketShip HQ and a prominent podcaster and influencer. 

Claire Rozain, Founder at Global Warming Games, formerly at Rovio. Recently, she joined Carry1st in the senior role, Africa’s leading publisher of mobile games and digital content. 

Philip WeisKirchen, Senior VP at Jampp with expertise in MMPs and Game Genetics. 

Lastly, Gadi Eliashiv, Co-Founder of Singular. He’s also super knowledgeable about mobile ad tech, growth data and privacy technologies and will introduce us to Singular.

Gadi 

I’m really humbled and excited to be a part of this panel. We at Singular are deeply involved in the changes taking place.

We are a measurement and attribution platform, and also do a lot of analytics. We help companies with their growth data. Our power comes from meeting customers and being able to work with some of the best brands in the world. I’m excited to explore how UA evolves in this different era of privacy. 

John 

The first poll is on the future of marketing measurement. You can select multiple answers: SKAN privacy sandbox, MMM incrementality, probabilistic device-level fingerprinting, first-party data, or channel stats from ad networks/partners.

Results: 51% SKAN privacy sandbox, 60% MMM incrementality, 12% probabilistic device-level fingerprinting, 50% first-party data, and 25% channel stats.

The second poll is focused on MMM and incrementality. Which is good for long-term budget allocation – channel selection or tactical campaign creative performance? 

Results: 64% channel selection, 36% tactical creative performance.

Philip, in terms of performance marketing on iOS in 2023, let’s discuss SKAN. What’s the percentage of the market that you see that’s actually good at SKAN?

Philip 

At Jampp, approximately 65% of our clients in the iOS-heavy US market are successfully running SKAN campaigns. While this can be seen as a good percentage, there is still a debate about whether it represents the overall market.

But around 35% of our advertisers have either not started testing SKAN or chose to postpone it, indicating that it’s not a priority for them currently.

From my perspective, 35% is way too much, because we believe that SKAN is the only sustainable way to effectively reach an iOS audience by making informed and reliable marketing decisions.

John 

Claire, same question, what’s the percentage of the market that you think is good at SKAN?

Claire 

I think firstly, there is no good or bad. By switching to different companies, I saw different ways of doing things.

For some marketers, it definitely makes sense to invest resources in it. For some others, revenue is the most important thing. 

John 

Jayne, your thoughts?

Jayne

I think there are a lot of overly confident marketing groups out there that think they can get away with things that are just going to go away like fingerprinting, and are not even starting on SKAN yet. In my opinion, if you’re already testing it, you’re good. 

John 

Shamanth, same question.

Shamanth 

Yeah, I echo what Jayne said. You could be getting bad results because of the nature of your product and the way your product monetizes. 

John 

Philip, if you’re excelling at SKAN, meaning you’re running successful performance marketing campaigns with positive ROI and achieving decent rollouts, is that sufficient? Or do you require additional measurement methodologies?
 

Philip 

To clarify, running SKAN campaigns allows you to bid for users with IDFA, providing reporting on results for both opted-in and opted-out users.

So it’s not like we are leaving IDFA traffic or IDFA attribution behind. I believe SKAN is enough if you have figured out the setup, but figuring out the setup requires advertisers, MMPs and also DSPs to work closely together.

From our side, I can report that we are seeing good results for our customers when we run SKAN campaigns. On an average, at least 37% lower CPI rates in SKAN versus IDFA-only campaigns.

However, probabilistic attribution can still be leveraged by advertisers as it stands now. At least the methodologies that are more Apple-compliant when they complement the SKAN campaigns that they are running and it can help to get further and gain additional insights for smarter marketing decisions. But the long-term sustainability of this approach remains questionable.
 

John 

Jayne, do you need additional measurement methodologies?

Jayne 

No, I think there are too many. Having worked in the industry for almost 20 years, we did everything, only taking 30% of impression-based conversions and various other attribution approaches, but many of them ended up being highly erroneous. The campaign performance was so far from the truth.

So I think less is actually more in this case. It’s just a matter of agreeing on the source of truth.
 

John 

Shamanth, agree or disagree?

Shamanth 

I would disagree. But it depends on your scale and how big you are. If you are at scale, and on multiple channels, you need additional measurement methodologies. If you’re small and don’t have too much ambiguity in what you’re measuring, I don’t think you need all that much.

John 

Interesting. Claire?

Claire 

I agree with Jayne. We are in a market that is constantly changing. We are adding live ops, minigames, and so on. At the end of the day, I feel it’s a bit overrated to say, “Oh, you are going to have a good schema, and then you’re going to make revenue” because we know it’s not how it works.

John 

Excellent. Gadi, let’s talk about SKAN 3.0 versus SKAN 4.0. How much performance improvement do you think we’ll have?

Gadi 

While Apple has provided us with more value and capabilities, it often comes with certain limitations.

You get one postback. But then, you only get one bit of information like the cost conversion value. Or do you have more options for source IDs? This means that you can now maybe encode creative or geo or other parameters more freely. But that means that you need to pass on certain thresholds.

So the real answer is, it depends from a product standpoint, and we’re excited about it because it’s more things to work with. It adds complexity, because of how it works with the second postback and the missing data. There are even more variations right now. We need to see the live data and what thresholds Apple is going to set.

John 

Shamanth, are you optimistic about SKAN 4.0?

Shamanth 

Yeah, it’s definitely going to be a step forward, as you will have much fewer null conversion values, compared to SKAN 3.0.

If you’re a small advertiser, you are going to get a cost conversion value when you’ve got none in the past. Larger advertisers will also experience fewer null conversion values. I think one of the bigger challenges has been the loss of signal, which SKAN 4.0 is going to take some steps toward addressing, even if it’s not completely groundbreaking,

John 

Gadi, what should marketers do to prepare for SKAN 4.0?

Gadi 

It depends on what your setup looks like. But what we recommend to our customers is to first update the SDK and if you use a measurement platform, make sure that that SDK supports SKAN 4.0. A lot of our big customers are already running tests right now.

We were just in MAU in Vegas and we met a lot of the big partners as well. They’re all talking about how to deploy this and run tests.

There are a few features in SKAN 4.0 for which you need alignment from the industry on how to use it, like the lock window functionality, which is just some anecdote.

This alone could cause a lot of issues if you don’t standardize correctly. So we took a stance at Singular on what to do or actually not to use it because it may cause issues in the data. So make sure your app is ready, build your models, and update your SDK.

John 

At MAU, I heard a lot of people saying how they are waiting to see what Meta will do and how they’ll treat it. We’ll discuss privacy sandbox and MMM, and address one more question on SKAN.

Last question on SKAN for now. If you could have one feature request for improving SKAdNetwork for marketers what would it be? 

Shamanth, you first.

Shamanth 

Retargeting, perhaps.

John 

Excellent. Jayne, one feature request for SKAN?

Jayne 

High-level data on where the users are when they’re not in our app.

John 

Very good. Claire, one feature request.

Claire 

Definitely no SKAN cohesion and normalization. Google and Apple agree together on something with the government. So we have a worldwide tracking framework, and we can focus on a project.

John 

Philip, your one request for SKAN 4.0.

Philip 

I did like the retargeting request. But I think I would pick the ability to tie the three prospects to the initial activity that generated the install, and also to receive them with a little less delay to make it a little bit closer to what we used to know as return prospects.

John 

Gadi your one feature request?

Gadi 

I support what Phillip said about being able to tie the postbacks. But even without that, I wish the second and third postbacks convey more data,

John 

We’re now going to talk about the Android Privacy Sandbox. We are going to bring up a poll. 

A privacy sandbox is a 360-degree solution because it offers the right measurement audiences and retargeting among a few other things.

Should Apple add features to SKAN to enable similar functionality? So basically, along the lines of what a couple of people have mentioned.

Results:

3% said no. Maybe we have some real privacy advocates.

97% said Yes.

Gadi 

There is one aspect where technically Google is removing GAID and Apple kept IDFA. Why would someone say no?

John 

Exactly. Now, 2023 might be a little early for a private sandbox, but not really. It’s all coming. There are tests happening right now pretty soon and Google is releasing more devices. Gadi, can you give us a brief overview of Privacy Sandbox and Android?

Gadi 

Privacy Sandbox is a broad initiative by Google to promote privacy, both on the web and mobile. They have the web version and the mobile version is for Android. You could say it’s their response to what’s happening with browsers and third-party cookies on Firefox and Safari. It’s been blocked forever now. Chrome is about to do the same. And then Apple was doing things like an escape network, and they have ITP on the web. That’s Google’s response.

In general, what’s really nice about this is that it’s comprehensive. So it’s not just for attribution, they try to solve other use cases like retargeting and then what’s impressive about it is, unlike the SKAN launch where it caused a lot of people to sour over it then, it was abrupt and went away too quickly, etc.

Google’s taking their time. They’re in beta. They’re working with partners and we work with them all the time. They’re slowly rolling this and they’re making sure everybody’s ready. 

John 

And the Google referrer Gadi, does it stick around? 

Gadi Eliashiv 

Maybe the broader question is, how is this different from SKAN? Or is it similar? And if I had to summarize it, it’s way better than SKAN.

You have similarities like it’s running on a device. So the idea is that you’re not just sending a level of data everywhere, it’s staying on the device. And they also have a component where it’s sending a post back similar to SKAN post back, but it’s a lot more granular.

They also added another mechanism, which is called aggregate double reports and we have a lot of blog posts on it. But if the prospect mechanism is not enough, they give you another mechanism, which is amazing.

Unlike iOS, a key mechanism based on our current understanding of how attribution works for clicks will still remain. When you go from one website to another, the target website knows where you came from.

So there’s the equivalent in the Play Store. That means that if somebody clicked on your ad and landed in your app, you’ll still have that information. That means that click-based attribution is still deterministic. 

People still need to go to the Privacy Sandbox. Google will be much more friendly for marketers, and they still made major strides toward privacy.  

John 

Philip, how does Privacy Sandbox differ from SKAN?

Philip 

The Privacy Sandbox will protect user privacy, but it won’t be as limiting as SKAN. It will offer advertisers a lot more information in comparison to what SKAN does. In the end, we should not forget that Google is an advertising business, after all. I believe they are trying to find the right balance between protecting user privacy, but also enabling advertisers to spend their marketing dollars with confidence.

We have to understand that SKAN is a single consolidated privacy-focused solution. While Privacy Sandbox is composed of different technologies, I think the one point that is very important to highlight is that Google is interested in building a solution that also allows advertisers to keep running retargeting campaigns and I believe that’s something where we are not the only DSP. 

John 

Absolutely. We saw in the polls and also from the panel that retargeting is a key ask that is gone right now, pretty much on iOS.

Okay, we’re going to move into the third segment of our webinar when talking about the future measurement: MMM, incrementality modeling, and a few other things.

We’ve talked about the platforms, privacy and safe measurement. They’re deterministic aggregates. They have some capabilities and some limitations. 

Is there media mix modeling incrementality modeling off of sets of data? Shamanth talked about it, what are brands exploring, and what are they using?

Shamanth

I think the vast majority of folks are using some form of non-deterministic measurement. I think that specific method matters less than the fact that they are making some effort to do this. We’re leaning into a lot of MMM, just because we use an open-source tool that we are very comfortable with and that we have gotten the results from. 

There are third-party tools out there, the folks who run incrementality, there’s quite a bit folks who do model data just as those third-party tools that do that. Now, all of these are broadly put under the umbrella of non-deterministic measurement, which I think a lot of the savvy folks are leaning into. 

John

Claire, what are you seeing?

Claire 

I’m seeing exciting things that I never saw before. It’s basically a driven attribution and not binary tracking. AI is taking over tracking and going beyond.

John 

So AI-driven attribution and measurement. But what methodologies and what data is AI taking in Claire?

Claire 

It’s basically that you have a learning phase, as old as machine learning, in order to train your tracking model. Then you have your own tracking model that is fitting and tailored. It’s a bit like the tailored algorithm that you used to have for programmatic but this time for tracking. 

John 

Jayne, what are your thoughts here?

Jayne 

So let’s say I say media mix modeling. That’s really just one ad and most advertisers still have it. But internally, we do nearest neighbor logic, and we do promo codes and have a survey. So I think there is a potential that AI is helping consolidate that into one.

But like I always say, you just need one source of truth. If you’re doing multiple treatments at multiple different points, you’re getting further away from the ultimate goal of one source of truth. 

For instance, you’ve had four treatments, and even after Singular or whatever MMP you use, you really need to make sure that you’re as close as possible with them. Otherwise, your campaigns will not be optimized the way that you want them to, and it just has ramifications downstream.

So I would say to stop at up to three treatments. I would say even post-MMP, you need to get closer to your MMP and try to remain consistent and just work with them. 

A lot of it too is the fact that you’ve been potentially succumbing to what your finance or product team wants. By that point, your campaigns are not even close to being optimized to what you have reported within Singular and even worse by trying to appease multiple different parties within a company that is now looking at four different types of attribution.

John 

So, Jayne, I’m going to ask you to dig into that a little more. Because I find what you’re saying is interesting. If I have multiple sources of information, and they all say different things, I don’t know what’s true anymore or how to act or operate or optimize or anything like that.

I guess the other side is, if I can somehow consolidate or if I can somehow see different aspects of the truth. If you’re only going to pick one, how do you choose?  

Jayne 

I think there’s never going to be a perfect solution. I’ve been in industries where timeliness is actually more important than the fidelity of the data that you’re getting. It can be very different if you’re more of an evergreen game that doesn’t have seasonality. You might want more fidelity, and timeliness is not as much of an impact. So you just have to make some concessions.

But at the end of the day, you need to consolidate data into one. And if you have more sources, it truly confuses people internally. And I’ve been in enough companies to have seen this repeat itself over and over again.

It just causes more confusion than if you just were to stick to one and then consistently come back at a regular cadence. When do we actually come back and refine it?

Claire 

That’s so interesting. Who is taking the decision usually to change the tracking or the attribution? Is it changing so much because of leadership?

Jayne 

It really does depend on the leadership. You have marketing organizations that are either embedded more in the tech side or more in the creative or marketing or ops side. Depending on the stage of the company, it might be more of a CFO lead. That is a different conversation than if you were to talk of a product-led company.

If you are a leader in your marketing group, it is truly on you to fix this stuff and to make sure that you’re being very consistent in consolidating the data. If your attribution is bloating, it’s more of an impact on you. As a leader, you are not being able to persuade and get the company to align with your vision. And that is a battle that you need to solve.

John 

Interesting. What’s your response, Shamanth?

Shamanth 

I’ve seen turf wars in companies where different groups have wanted to measure things differently. I have said before that LTV is a completely made-up number, depending on who’s talking.

John 

How do all the measurement methodologies fit together? Do they plug in at different times for different teams? Or should you try and consolidate around one?

Gadi 

In every organization, even for us, as we’re a B2B business, you have business development and sales who want to have different ways of measuring different things, etc. versus marketing. So it happens everywhere.

I had to take myself out of that decision of attribution. I asked my co-founder to be the arbitrator, so he’s just deciding logically, but it is hard and it happens a lot.

I think everybody should strive for consolidation. The challenge is that the SKAN situation really broke a lot of things that were the norms before. People had the last touch and IDFA. And then SKAN came in and some of the numbers didn’t make sense. People are trying to get SKAN good enough so that it’s reliable, and works on all channels. 

But then people are refusing to accept SKAN because it’s not as good as what we had before. 

So we’re trying to make the best SKAN solution we can and make it super advanced. But on the flip side, we have to explore other methods. Our view is that even today on iOS you can have a lot of signals. You have your partial data faded but you want to use that because it’s valuable information.

For the people that opted in, you have the SKAN data, and maybe you have some first-party data. So you try and build a cohesive view on iOS, that starts to use similar signals that can be explainable. But the output has to be a unified view.

There is another effort, which is media mix modeling. We announced a beta first-to-market solution for our customers because a lot of people like it. They like it because it’s beyond just iOS, and you can measure all the channels.

It’s a different model. There’s a statistical model – it takes aggregate inputs, and you could say that it’s resilient to privacy changes and it depends on the organization. Some prefer to focus, and some always look for innovations.

Our long-term vision is to consolidate all into a single number and call it the hybrid measurement. In phase one, we may enable some more measurement methods, because of a lot of demand from the market. Phase two will be where we will be bringing it all into a single place.

But it has to be meaningful. It can’t just be a trick where the data is completely off and you’re just doing something weird to consolidate it. We need some more iterations on people trying different methodologies before we try and consolidate it. Some people ask for 10 different methods and some channels really want to push MMM, because they’ll look better in that.

So you have a lot of forces working on the measurement ecosystem right now because the status quo has been destabilized. Simplicity will weigh in eventually. 

John 

Exactly. Some channels prefer MMM. Shamanth can mere mortals do it?

Shamanth 

I’d like to think I am a mere mortal. I don’t know advanced programming and I and many folks on my team are able to run MMMs. We use open-source code that we’re able to customize.

If you have larger budgets and more sophisticated teams, you can certainly do much more advanced models. I would say to get started, you can have data in a spreadsheet and open source code, or even without the open source code, you can have a spreadsheet-based model to just understand and get the hang of how to run MMM. You can set up a linear regression model, that’s the very basic form of MMM.

So the short answer is yes. We did a webinar earlier where some folks asked for a couple of YouTube videos we have made, so feel free to check out the YouTube videos we have. It will help to understand and set up MMM using the open-source code that we have.

John 

Gadi, because you mentioned that Singular released an MMM product, is there something that more mere mortals can do?

Gadi 

Yeah, you could use open-source software. Eventually, these algorithms have been around forever. Algorithms like this existed for 20 years or so. And, with some of the models, you could get more sophisticated or less, but I also think we can make it easier. That’s where the job of the vendor comes in. 

In MMM, a big portion is onboarding and getting all your data and continuously feeding data.

The stuff that Singular has is really good at aggregating, normalizing, and standardizing. So for us, it makes a lot of sense to have a product that does that, ideally, as much tech base as we can, and not hand-holding a bunch of analysts tweaking numbers, and that’s the direction we’re going. I think you could do it with open source. But I think that we could probably make it easier for you because it houses all the data you need anyway.  

John 

Claire, so marketing teams are looking at all these things that are options for them. What do they need to do now? How do they prepare and what skills and data and tools do they need to learn or acquire?

Claire 

Speaking about tracking, if you think long term, you should invest in your own database and first-party AI and algorithm. As Jayne said, if you want to be successful on Google, you need to be compliant with their Sandbox. But if you have your own algorithm, you are linked to old folks. 

John 

Jayne, your thoughts on prepping for the future?

Jayne 

When a marketer tells me or gets frustrated about what’s going on with SKAN and privacy, I think to myself, what an opportunity that is for a marketer.

What I mean by that is a lot of UA teams and marketers have lost a lot of credibility in the last few years by spending money when they shouldn’t be and by making really bad decisions.

I think this is an opportunity to realign teams, create the attribution model and be the one that decides all of these things. That ultimately will give you the most power over your budgets and over the business decisions that you want.

Gradually get the company around it. Get your CFO involved. Get your product team involved, but make it like a group effort. Lead it and you’ll be the one that ends up in a way better spot and be able to call a lot more shots than you did today.

John 

I think we’ve seen that over the past year. Shamanth, what is your top takeaway for people on this topic?

Shamanth 

Non-deterministic attribution is the present and the future and people should get used to the new reality.

John 

And get comfortable with uncertainty and knowing that you don’t know it all. 

Gadi, what’s your top takeaway?

Gadi 

Yeah, I think it actually relates to what Jayne said, that you have to lead a certain way. Make sure that you innovate and invest. AI people have done amazing things taking stuff they built in-house and pouring that into frameworks like SKAN and privacy sandbox.

My takeaway in the last few years was to build a team to handle privacy. Now I’m saying that don’t just build a team, be ready and innovate.

John 

I love that it makes me think of GPT for an AI that Claire brought up. We’ve seen GPT for tasks like investing and people have actually put millions of dollars in GPT-4 running the investment fund. 

Jayne, what is your top takeaway?

Jayne 

I want to see a lot more failure in this industry, with a lot more people making this a really hard thing to solve. We should start getting messy with it and start testing a bunch of stuff. Don’t take too much advice because I think that too much advice is going to lead to a lot more confusion.

John 

Excellent, Claire, do you have anything different for us?

Claire 

You need to be a positive kickass marketer. You need to be spectacular, open to actual and upcoming opportunities, and watch your monetization activities closely.  

John 

Excellent. And Philip, the last top takeaway for advertisers and marketers?

Philip 

It’s still quite important to work closely with your MMP partner and potentially also with your DSP partner to figure things out. It’s a team effort that will lead to success in the end, because only by truly understanding how these privacy initiatives work, advertisers will be able to make the right marketing decisions for their mobile apps. Collaboration is the key here.

John 

Excellent. I think you need SKAN and Sandbox. You’re adding the network to the top-of-funnel data, your first-party data especially. 

Philip, what networks are currently supporting SKAN 4.0?

Philip 

I can only speak for programmatic channels, I believe they would all support SKAN 4.0. I’m not sure about ad networks, but I also see a tendency that there are some channels out there that still rely on probabilistic attribution. They might even push their customers to use probabilistic attribution for attribution.

John 

That’s awesome. Gadi, what’s the slowdown here?

Gadi 

I think the slowdown is that it’s a new model that adds some features but adds a lot of complexity. I would make sure that first of all, it doesn’t break my existing revenue stream.

So they have to fix the product. They have to make sure all their optimizations know how to take advantage of SKAN 4.0, running some early tests already, and then making sure that this will deliver better results, so it’s a risk. 

The amazing thing about this is Apple just released a new version and WWDC, and then it’s a year’s worth of work to make that complicated system function at its peak, because that’s what everybody’s aiming for. How do I maximize profit over these bits of information I’m getting from SKAN? That’s the real complexity.

By the way, WWDC is like a week or two out, hopefully, we’re not gonna get SKAN 5.0 because I don’t think Apple would release it before that’s fully implemented.

John

Shamanth, I’m going to ask you the next question. The anonymous attendee is asking about CAPI, and event APIs. How do we feel about server-to-server signals back to ad networks, and if it will help with targeting or optimization on iOS? 

Shamanth 

On e-commerce, it’s certainly been effective. I just haven’t seen that being done or tried as much for apps. In theory, it should work. I think the practical implementation is much more of a challenge.

John 

Interesting, do we suspect SKAN 4.0 is over or under-counting installs attributed to certain sources?

Gadi 

Let’s say, you’re sending two installs per source ID, you’re really not using SKAN the right way. That’s where the partners come in because the data impacts them. And I’m sure by now everybody optimizes that, but theoretically, it could happen.

I will say from the flip side, there are partners and maybe everybody feels that SKAN doesn’t do them justice. And that’s why I think people are looking for other methods. There is a real concern that it does under-represent the impact of a given channel.

So this is not me speaking to Singular, but more from what I’m hearing from partners, I think some of them believe that it’s under-counting, under-presenting, or doesn’t show the full ROI potential here. It might not be just counting installs, it might be just under-representing the value.

John 

Okay, interesting.

Philip 

Could it also be because maybe it was over-represented before with the previous solutions?

John 

Exactly. Long attribution windows?

Gadi 

There’s an entire talk of people saying the old attribution wasn’t perfect either. So you don’t need to mourn it too much like even before.

John 

Jayne, this question is from an anonymous attendee: Our SKAN data doesn’t populate immediately for media campaigns, there’s like a two to three-day delay. As a result, our campaigns’ learning phases take a while to optimize, then we burn through the budget on lower-performing spend. Can you share your best practices for Meta SKAN campaign expansion testing that helps mitigate this inefficient budget burn?

Jayne 

I think that’s about your campaign structure if anything. What I’ve noticed is that if you’re running too many campaigns at once, and obviously you’re just burning your budget, rather than starting a little bit, I would say be more refined on your testing and then go from there.

So I would cut the number of campaigns or the hierarchy of your campaigns and try to be as really consistent with your budget as possible.

I’ve run budgets with hundreds of million dollars, and it really is that you need to consolidate. Even if that means being more patient with your testing as you can’t test everything at once. 

Be specific about what you want to test. Test certain creatives, maybe your structure around geo, or whatever it is, but don’t do everything at once. If you have a finite budget, especially with Facebook, I would just consolidate and refine my testing and have a longer roadmap for what I want to test.

John 

Shamanth really quickly. What about incrementality? Is this the same concept as MMM? And is incrementality at a granular level, like a campaign, even achievable?

Shamanth 

No, you cannot do very granular measurements for incrementality. Conceptually, incrementality is a little bit different. It’s more of an experimental paradigm. 

You could certainly use MMM to measure incrementality. But when I think about incrementality, it’s more like I increased budgets in Italy today as an experiment. What happens to my revenue in Italy on this metric? So that’s the experimental paradigm, which is incrementality MMM is somewhat more combating.

John 

Wonderful. I want to thank everybody who attended this webinar and spent some time with us. I want to thank every panelist. You guys were amazing. See you next time and have a great day.