Thomas Petit (Mobile Growth Consultant) talks about why most apps should do at least some paid user acquisition, how the UA manager role has evolved and how to think about marketing subscription apps in the post-IDFA area.
Organic is never enough to scale very significantly. Every single one of the huge apps that make it to the top of the charts are at least partly pushed by paid acquisition. Example: TikTok._ _You need a combination of a lot of paid acquisition, a lot of virality and a great product to stay at the top for long.
You should start working on organics way before your app is live: building an email list, content, community, etc.
You need to fuel your cohorts so you can work on improving retention. You want a minimum of new users every day so you can know if the latest release you ship is better than the previous one. If you don’t have a strong organic foundation, then paid UA is probably required.
Be careful about the biases that your early adopters can bring (if you have a strong organic baseline) because their behavior is not necessarily representative of the mainstream audience. This can make running paid UA healthy: to prove your hypotheses and optimize the product towards the mainstream audience.
You might want to exclude the cohorts for days where you get App Store features (if it represents a big portion of your audience), because the audience is often on the opposite end of early adopters and not representative of what your product should become.
If you have a very small budget (like $500/month), focus on organic or specific paid actions (e.g., promotion on Reddit, very specific keywords on ASA, etc.). Do not go to Facebook or Google, because they are machine-learning driven.
Building your owned and earned audiences is a must from Day 1 and all the way (even after you start paid UA). You need to find the ways that work for you to do that.
The only real lasting thing you get from App Store features is the “Editor’s Choice” badge that you can keep permanently on your listing.
(Jacob, Revenue Cat)
The core of the UA manager job today is producing, testing and deploying creatives. The role is also more strategic, as you have to think more like a data analyst so you can feed the right events to the machine learning algorithms.
Marketers look a lot at the impression to install ratio, but Facebook is now taking engagement more into account (“like” vs. “comment” or “shares”, view rates, etc.) You need to trigger these positive signals through your creatives and keep people engaged on the Facebook platform for the algorithm to favor you in the auction.
Be careful about using “free” in your communications: start a free trial, etc. because you might attract users that just want free stuff. It’s probably best to have less users coming in but more qualified.
Facebook’s algorithm wasn’t looking only at the data from your app, it was also looking at the data from pretty much all the apps in your category. With ATT, the duopoly is now attacked at the heart of their differential advantage.
With the new attribution system it’s going to be even trickier to optimize for the free trial event. Smart subscription advertisers often use a more complex event than the free trial, by layering it with some other things that are specific to their app. Example: free trials for users that have not cancelled within the first 2 hours + engagement.
Subscription apps with no free trials that don’t have the budgets to optimize for purchases have to find earlier signals to create a “bundle of events”: onboarding answers, accepting notification requests, etc.
Customizing too much the combination of events you’re optimizing for on Facebook can also prevent its algorithm from optimizing correctly, because the comparison with the data that it has from all the other apps becomes less relevant.
There is no single source of truth anymore. What the ad networks do to optimize and the way advertisers interpret it is going to be more complex. Example: Google might use a blend of signals (SKAdNetwork, the IDFA when available, Android app, first-party data around contextual placements and other apps, etc.) and advertisers are going to need to use different data sets to interpret what’s happening (SKAdNetwork – not the full picture, IDFA, media-mix model or probabilistic model, known behaviors from specific-channels – like how fast users go through onboarding).
Try to better understand the interdependencies between channels (e.g., email list, Instagram, offline, etc.) so you can assess how much you should be putting back into the paid acquisition machine.
Organic is never enough to scale very significantly. Every single one of the huge apps that make it to the top of the charts are at least partly pushed by paid acquisition. Example: TikTok._ _You need a combination of a lot of paid acquisition, a lot of virality and a great product to stay at the top for long.
You should start working on organics way before your app is live: building an email list, content, community, etc.
You need to fuel your cohorts so you can work on improving retention. You want a minimum of new users every day so you can know if the latest release you ship is better than the previous one. If you don’t have a strong organic foundation, then paid UA is probably required.
Be careful about the biases that your early adopters can bring (if you have a strong organic baseline) because their behavior is not necessarily representative of the mainstream audience. This can make running paid UA healthy: to prove your hypotheses and optimize the product towards the mainstream audience.
You might want to exclude the cohorts for days where you get App Store features (if it represents a big portion of your audience), because the audience is often on the opposite end of early adopters and not representative of what your product should become.
If you have a very small budget (like $500/month), focus on organic or specific paid actions (e.g., promotion on Reddit, very specific keywords on ASA, etc.). Do not go to Facebook or Google, because they are machine-learning driven.
Building your owned and earned audiences is a must from Day 1 and all the way (even after you start paid UA). You need to find the ways that work for you to do that.
The only real lasting thing you get from App Store features is the “Editor’s Choice” badge that you can keep permanently on your listing.
(Jacob, Revenue Cat)
The core of the UA manager job today is producing, testing and deploying creatives. The role is also more strategic, as you have to think more like a data analyst so you can feed the right events to the machine learning algorithms.
Marketers look a lot at the impression to install ratio, but Facebook is now taking engagement more into account (“like” vs. “comment” or “shares”, view rates, etc.) You need to trigger these positive signals through your creatives and keep people engaged on the Facebook platform for the algorithm to favor you in the auction.
Be careful about using “free” in your communications: start a free trial, etc. because you might attract users that just want free stuff. It’s probably best to have less users coming in but more qualified.
Facebook’s algorithm wasn’t looking only at the data from your app, it was also looking at the data from pretty much all the apps in your category. With ATT, the duopoly is now attacked at the heart of their differential advantage.
With the new attribution system it’s going to be even trickier to optimize for the free trial event. Smart subscription advertisers often use a more complex event than the free trial, by layering it with some other things that are specific to their app. Example: free trials for users that have not cancelled within the first 2 hours + engagement.
Subscription apps with no free trials that don’t have the budgets to optimize for purchases have to find earlier signals to create a “bundle of events”: onboarding answers, accepting notification requests, etc.
Customizing too much the combination of events you’re optimizing for on Facebook can also prevent its algorithm from optimizing correctly, because the comparison with the data that it has from all the other apps becomes less relevant.
There is no single source of truth anymore. What the ad networks do to optimize and the way advertisers interpret it is going to be more complex. Example: Google might use a blend of signals (SKAdNetwork, the IDFA when available, Android app, first-party data around contextual placements and other apps, etc.) and advertisers are going to need to use different data sets to interpret what’s happening (SKAdNetwork – not the full picture, IDFA, media-mix model or probabilistic model, known behaviors from specific-channels – like how fast users go through onboarding).
Try to better understand the interdependencies between channels (e.g., email list, Instagram, offline, etc.) so you can assess how much you should be putting back into the paid acquisition machine.
Organic is never enough to scale very significantly. Every single one of the huge apps that make it to the top of the charts are at least partly pushed by paid acquisition. Example: TikTok._ _You need a combination of a lot of paid acquisition, a lot of virality and a great product to stay at the top for long.
You should start working on organics way before your app is live: building an email list, content, community, etc.
You need to fuel your cohorts so you can work on improving retention. You want a minimum of new users every day so you can know if the latest release you ship is better than the previous one. If you don’t have a strong organic foundation, then paid UA is probably required.
Be careful about the biases that your early adopters can bring (if you have a strong organic baseline) because their behavior is not necessarily representative of the mainstream audience. This can make running paid UA healthy: to prove your hypotheses and optimize the product towards the mainstream audience.
You might want to exclude the cohorts for days where you get App Store features (if it represents a big portion of your audience), because the audience is often on the opposite end of early adopters and not representative of what your product should become.
If you have a very small budget (like $500/month), focus on organic or specific paid actions (e.g., promotion on Reddit, very specific keywords on ASA, etc.). Do not go to Facebook or Google, because they are machine-learning driven.
Building your owned and earned audiences is a must from Day 1 and all the way (even after you start paid UA). You need to find the ways that work for you to do that.
The only real lasting thing you get from App Store features is the “Editor’s Choice” badge that you can keep permanently on your listing.
(Jacob, Revenue Cat)
The core of the UA manager job today is producing, testing and deploying creatives. The role is also more strategic, as you have to think more like a data analyst so you can feed the right events to the machine learning algorithms.
Marketers look a lot at the impression to install ratio, but Facebook is now taking engagement more into account (“like” vs. “comment” or “shares”, view rates, etc.) You need to trigger these positive signals through your creatives and keep people engaged on the Facebook platform for the algorithm to favor you in the auction.
Be careful about using “free” in your communications: start a free trial, etc. because you might attract users that just want free stuff. It’s probably best to have less users coming in but more qualified.
Facebook’s algorithm wasn’t looking only at the data from your app, it was also looking at the data from pretty much all the apps in your category. With ATT, the duopoly is now attacked at the heart of their differential advantage.
With the new attribution system it’s going to be even trickier to optimize for the free trial event. Smart subscription advertisers often use a more complex event than the free trial, by layering it with some other things that are specific to their app. Example: free trials for users that have not cancelled within the first 2 hours + engagement.
Subscription apps with no free trials that don’t have the budgets to optimize for purchases have to find earlier signals to create a “bundle of events”: onboarding answers, accepting notification requests, etc.
Customizing too much the combination of events you’re optimizing for on Facebook can also prevent its algorithm from optimizing correctly, because the comparison with the data that it has from all the other apps becomes less relevant.
There is no single source of truth anymore. What the ad networks do to optimize and the way advertisers interpret it is going to be more complex. Example: Google might use a blend of signals (SKAdNetwork, the IDFA when available, Android app, first-party data around contextual placements and other apps, etc.) and advertisers are going to need to use different data sets to interpret what’s happening (SKAdNetwork – not the full picture, IDFA, media-mix model or probabilistic model, known behaviors from specific-channels – like how fast users go through onboarding).
Try to better understand the interdependencies between channels (e.g., email list, Instagram, offline, etc.) so you can assess how much you should be putting back into the paid acquisition machine.
Notes for this resource are currently being transferred and will be available soon.
The best product is not always going to win.
[💎@02:42] Organic is never enough to scale very significantly. Every single one of the huge apps that make it to the top of the charts are at least partly pushed by paid acquisition. Example: TikTok. You need a combination of a lot of paid acquisition, a lot of virality and a great product to stay at the top for long.
On the other hand, paid UA alone is not going to be enough.
First thing to look at are the retention metrics.
Should apps wait to do paid UA until they have some product-market fit?
[💎@06:00] You should start working on organics way before your app is live: building an email list, content, community, etc.
The need to get started on paid UA depends on how fast you need to grow. Example: VC money requiring to be more aggressive vs. being bootstrap.
[💎@06:50] You need to fuel your cohorts so you can work on improving retention. You want a minimum of new users every day so you can know if the latest release you ship is better than the previous one. If you don’t have a strong organic foundation, then paid UA is probably required.
[💎@07:17] Be careful about the biases that your early adopters can bring (if you have a strong organic baseline) because their behavior is not necessarily representative of the mainstream audience. This can make running paid UA healthy: to prove your hypotheses and optimize the product towards the mainstream audience.
[💎@09:37] You might want to exclude the cohorts for days where you get App Store features (if it represents a big portion of your audience), because the audience is often on the opposite end of early adopters and not representative of what your product should become.
If you're ambitious, there will always be a point where you will be stalling a bit and usually that’s when people turn to paid UA.
Some companies rely only on paid UA from the start, because they want to understand immediately what kind of audience they would get. This is too extreme as well. You want both organic and paid UA.
Small budget according to Thomas: probably more than $20k/month, it’s hard to do anything under $5k (maybe with ASA). $10k/month is when you can start getting something significant and get the same traffic you might be able to get later.
[💎@13:15] If you have a very small budget (like $500/month), focus on organic or specific paid actions (e.g., promotion on Reddit, very specific keywords on ASA, etc.). Do not go to Facebook or Google, because they are machine-learning driven.
Owned audiences that you build for yourself or earned audiences (through PR, featuring) are things you need to do from Day 1 of the project. They also take more time to bring results.
[💎@16:15] Building your owned and earned audiences is a must from Day 1 and all the way (even after you start paid UA). You need to find the ways that work for you to do that.
We call Organics what’s not attributed. But in reality, it’s more what you spend time against in order to get exposure. Example: getting featured requires spending time on Apple relationships, on building features, etc.
[💎@18:35] The only real lasting thing you get from App Store features is the “Editor’s Choice” badge that you can keep permanently on your listing. (Jacob, Revenue Cat)
The impact of featuring has gone down now. Features are also shorter, and they’re presented differently. They used to bring 100k installs but now App of the Day can bring 5k installs. Plus, there’s a random factor and as mentioned above they also bring a particular kind of traffic.
Features are not a sustainable strategy. So, you have to assume you won’t get them.
You don’t want to be chasing features that Apple wants you to implement instead of a user-centric approach.
The intent of users coming in through features is not as high. Turning curiosity into a paid subscription is harder too.
People have also already tried a lot of apps in each category
2015-2020: Facebook has been driving a huge part of the discovery. But now, discovery might be coming back to the App Store.
Job as a UA manager has already shifted since 2018, with the automation made by the networks taking over adjusting bids, deciding which creatives are shown to who, etc.
[💎@30:15] The core of the UA manager job today is producing, testing and deploying creatives. The role is also more strategic, as you have to think more like a data analyst so you can feed the right events to the machine learning algorithms.
Creatives are probably ¾ of the job: whether it’s talking to influencers, talking to a new agency, etc.
Examples of companies automating creatives in a very efficient way: copy.ai for copy, trypencil.com (you send your current assets, “brief the AI” and they create ads).
It is still a human task but part of it will be automated.
Who works on paid UA, product-led growth, understanding retention, etc.? For early stage it’s probably the same person that touches a bit all of this. This is how each person can find what they are the best at (for example, Thomas stepped away from creatives to make a bigger difference in onboarding, monetization, etc.).
A UA manager just focused on UA is now a terrible idea. It took a long time for people to realize that what matters is what happens down the funnel.
Acquisition and monetization are now more intertwined.
“Acquisition has become a business model competition”.
The difference you can make is on the creatives or in the monetization.
For games it used to be about whale-hunting: looking for a few very high-spenders. Facebook knows where the wales are, so it’s a matter of convincing the algo to drive them to you: through high bids and creative relevancy.
[💎@40:18] Marketers look a lot at the impression to install ratio, but Facebook is now taking engagement more into account (“like” vs. “comment” or “shares”, view rates, etc.) You need to trigger these positive signals through your creatives and keep people engaged on the Facebook platform for the algorithm to favor you in the auction.
“It’s your job to make Facebook happy”
The logic of bidding for subscriptions is different. The biggest networks didn't care about apps compared to gaming, DTC brands, etc.
Part of the role of a UA manager for a subscription app is to recognize that the FB algorithm is not made for subscriptions.
Something really useful for UA for a subscription app is to have multi-event funnels: are people engaging with the product, starting the trial, etc. You have to teach the FB algo a particular way. Example: Thomas used to map trial start to purchase (instead of “add to cart”) in order to get users that generally purchase (note: this does not work anymore).
[💎@46:08] Be careful about using “free” in your communications: start a free trial, etc. because you might attract users that just want free stuff. It’s probably best to have less users coming in but more qualified.
Attribution on apps is unique because there is the App Store blackbox between the ad click and the app open: no cookies, etc.
Initially there were workarounds, then the IDFA started being used. MMPs became absolutely critical for marketers.
Now is a very defining moment for the industry because acquiring users is going to be done in a very different way.
Facebook is not able to do attribution based on its own resources, even if it’s the platform displaying the ad and then having the Facebook SDK in the app. Pre-ATT, Facebook was not even showing ads to users if there wasn’t an IDFA.
Even though Apple might not be able to see what Facebook is doing on the backend, Thomas believes Facebook will not match data between apps and leverage that for optimizations, even if they could.
FB and Google were so powerful because they had data to know exactly what users were doing and could use this feedback loop to optimize for specific events.
[💎@56:30] Facebook’s algorithm wasn’t looking only at the data from your app, it was also looking at the data from pretty much all the apps in your category. With ATT, the duopoly is now attacked at the heart of their differential advantage.
Not everybody uses the same event, especially those that don’t have a huge budget. You need to figure out the event that has enough completion for the network to learn but enough correlation with long term goals like LTV. For many people that’s the free trial, but the free trial event can also be misleading.
[💎@59:15] With the new attribution system it’s going to be even trickier to optimize for the free trial event. Smart subscription advertisers often use a more complex event than the free trial, by layering it with some other things that are specific to their app. Example: free trials for users that have not cancelled within the first 2 hours + engagement.
[💎@1:01:00] Subscription apps with no free trials that don’t have the budgets to optimize for purchases have to find earlier signals to create a “bundle of events”: onboarding answers, accepting notification requests, etc.
[💎@1:02:21] Customizing too much the combination of events you’re optimizing for on Facebook can also prevent its algorithm from optimizing correctly, because the comparison with the data that it has from all the other apps becomes less relevant.
SKAdNetwork is the fundamental basis for attribution and every advertiser is going to use it.
You can modelize from the IDFA opt-in, but the opt-in needs to happen on both sides: on the source app (which displays the ad) and in your app.
[💎@1:06:00] There is no single source of truth anymore. What the ad networks do to optimize and the way advertisers interpret it is going to be more complex. Example: Google might use a blend of signals (SKAdNetwork, the IDFA when available, Android app, first-party data around contextual placements and other apps, etc.) and advertisers are going to need to use different data sets to interpret what’s happening (SKAdNetwork – not the full picture, IDFA, media-mix model or probabilistic model, known behaviors from specific-channels – like how fast users go through onboarding).
Fingerprinting is a subpart of probabilistic. Algolift assigns probability at the aggregated level based on what’s happening inside your app, and that’s compliant.
You live and die by the platform, so taking the risk to upset Apple is big.
On the UA front, don’t panic. Revisit the impact of marketing on everything, paid or not.
[💎@1:15:24] Try to better understand the interdependencies between channels (e.g., email list, Instagram, offline, etc.) so you can assess how much you should be putting back into the paid acquisition machine.
The best product is not always going to win.
[💎@02:42] Organic is never enough to scale very significantly. Every single one of the huge apps that make it to the top of the charts are at least partly pushed by paid acquisition. Example: TikTok. You need a combination of a lot of paid acquisition, a lot of virality and a great product to stay at the top for long.
On the other hand, paid UA alone is not going to be enough.
First thing to look at are the retention metrics.
Should apps wait to do paid UA until they have some product-market fit?
[💎@06:00] You should start working on organics way before your app is live: building an email list, content, community, etc.
The need to get started on paid UA depends on how fast you need to grow. Example: VC money requiring to be more aggressive vs. being bootstrap.
[💎@06:50] You need to fuel your cohorts so you can work on improving retention. You want a minimum of new users every day so you can know if the latest release you ship is better than the previous one. If you don’t have a strong organic foundation, then paid UA is probably required.
[💎@07:17] Be careful about the biases that your early adopters can bring (if you have a strong organic baseline) because their behavior is not necessarily representative of the mainstream audience. This can make running paid UA healthy: to prove your hypotheses and optimize the product towards the mainstream audience.
[💎@09:37] You might want to exclude the cohorts for days where you get App Store features (if it represents a big portion of your audience), because the audience is often on the opposite end of early adopters and not representative of what your product should become.
If you're ambitious, there will always be a point where you will be stalling a bit and usually that’s when people turn to paid UA.
Some companies rely only on paid UA from the start, because they want to understand immediately what kind of audience they would get. This is too extreme as well. You want both organic and paid UA.
Small budget according to Thomas: probably more than $20k/month, it’s hard to do anything under $5k (maybe with ASA). $10k/month is when you can start getting something significant and get the same traffic you might be able to get later.
[💎@13:15] If you have a very small budget (like $500/month), focus on organic or specific paid actions (e.g., promotion on Reddit, very specific keywords on ASA, etc.). Do not go to Facebook or Google, because they are machine-learning driven.
Owned audiences that you build for yourself or earned audiences (through PR, featuring) are things you need to do from Day 1 of the project. They also take more time to bring results.
[💎@16:15] Building your owned and earned audiences is a must from Day 1 and all the way (even after you start paid UA). You need to find the ways that work for you to do that.
We call Organics what’s not attributed. But in reality, it’s more what you spend time against in order to get exposure. Example: getting featured requires spending time on Apple relationships, on building features, etc.
[💎@18:35] The only real lasting thing you get from App Store features is the “Editor’s Choice” badge that you can keep permanently on your listing. (Jacob, Revenue Cat)
The impact of featuring has gone down now. Features are also shorter, and they’re presented differently. They used to bring 100k installs but now App of the Day can bring 5k installs. Plus, there’s a random factor and as mentioned above they also bring a particular kind of traffic.
Features are not a sustainable strategy. So, you have to assume you won’t get them.
You don’t want to be chasing features that Apple wants you to implement instead of a user-centric approach.
The intent of users coming in through features is not as high. Turning curiosity into a paid subscription is harder too.
People have also already tried a lot of apps in each category
2015-2020: Facebook has been driving a huge part of the discovery. But now, discovery might be coming back to the App Store.
Job as a UA manager has already shifted since 2018, with the automation made by the networks taking over adjusting bids, deciding which creatives are shown to who, etc.
[💎@30:15] The core of the UA manager job today is producing, testing and deploying creatives. The role is also more strategic, as you have to think more like a data analyst so you can feed the right events to the machine learning algorithms.
Creatives are probably ¾ of the job: whether it’s talking to influencers, talking to a new agency, etc.
Examples of companies automating creatives in a very efficient way: copy.ai for copy, trypencil.com (you send your current assets, “brief the AI” and they create ads).
It is still a human task but part of it will be automated.
Who works on paid UA, product-led growth, understanding retention, etc.? For early stage it’s probably the same person that touches a bit all of this. This is how each person can find what they are the best at (for example, Thomas stepped away from creatives to make a bigger difference in onboarding, monetization, etc.).
A UA manager just focused on UA is now a terrible idea. It took a long time for people to realize that what matters is what happens down the funnel.
Acquisition and monetization are now more intertwined.
“Acquisition has become a business model competition”.
The difference you can make is on the creatives or in the monetization.
For games it used to be about whale-hunting: looking for a few very high-spenders. Facebook knows where the wales are, so it’s a matter of convincing the algo to drive them to you: through high bids and creative relevancy.
[💎@40:18] Marketers look a lot at the impression to install ratio, but Facebook is now taking engagement more into account (“like” vs. “comment” or “shares”, view rates, etc.) You need to trigger these positive signals through your creatives and keep people engaged on the Facebook platform for the algorithm to favor you in the auction.
“It’s your job to make Facebook happy”
The logic of bidding for subscriptions is different. The biggest networks didn't care about apps compared to gaming, DTC brands, etc.
Part of the role of a UA manager for a subscription app is to recognize that the FB algorithm is not made for subscriptions.
Something really useful for UA for a subscription app is to have multi-event funnels: are people engaging with the product, starting the trial, etc. You have to teach the FB algo a particular way. Example: Thomas used to map trial start to purchase (instead of “add to cart”) in order to get users that generally purchase (note: this does not work anymore).
[💎@46:08] Be careful about using “free” in your communications: start a free trial, etc. because you might attract users that just want free stuff. It’s probably best to have less users coming in but more qualified.
Attribution on apps is unique because there is the App Store blackbox between the ad click and the app open: no cookies, etc.
Initially there were workarounds, then the IDFA started being used. MMPs became absolutely critical for marketers.
Now is a very defining moment for the industry because acquiring users is going to be done in a very different way.
Facebook is not able to do attribution based on its own resources, even if it’s the platform displaying the ad and then having the Facebook SDK in the app. Pre-ATT, Facebook was not even showing ads to users if there wasn’t an IDFA.
Even though Apple might not be able to see what Facebook is doing on the backend, Thomas believes Facebook will not match data between apps and leverage that for optimizations, even if they could.
FB and Google were so powerful because they had data to know exactly what users were doing and could use this feedback loop to optimize for specific events.
[💎@56:30] Facebook’s algorithm wasn’t looking only at the data from your app, it was also looking at the data from pretty much all the apps in your category. With ATT, the duopoly is now attacked at the heart of their differential advantage.
Not everybody uses the same event, especially those that don’t have a huge budget. You need to figure out the event that has enough completion for the network to learn but enough correlation with long term goals like LTV. For many people that’s the free trial, but the free trial event can also be misleading.
[💎@59:15] With the new attribution system it’s going to be even trickier to optimize for the free trial event. Smart subscription advertisers often use a more complex event than the free trial, by layering it with some other things that are specific to their app. Example: free trials for users that have not cancelled within the first 2 hours + engagement.
[💎@1:01:00] Subscription apps with no free trials that don’t have the budgets to optimize for purchases have to find earlier signals to create a “bundle of events”: onboarding answers, accepting notification requests, etc.
[💎@1:02:21] Customizing too much the combination of events you’re optimizing for on Facebook can also prevent its algorithm from optimizing correctly, because the comparison with the data that it has from all the other apps becomes less relevant.
SKAdNetwork is the fundamental basis for attribution and every advertiser is going to use it.
You can modelize from the IDFA opt-in, but the opt-in needs to happen on both sides: on the source app (which displays the ad) and in your app.
[💎@1:06:00] There is no single source of truth anymore. What the ad networks do to optimize and the way advertisers interpret it is going to be more complex. Example: Google might use a blend of signals (SKAdNetwork, the IDFA when available, Android app, first-party data around contextual placements and other apps, etc.) and advertisers are going to need to use different data sets to interpret what’s happening (SKAdNetwork – not the full picture, IDFA, media-mix model or probabilistic model, known behaviors from specific-channels – like how fast users go through onboarding).
Fingerprinting is a subpart of probabilistic. Algolift assigns probability at the aggregated level based on what’s happening inside your app, and that’s compliant.
You live and die by the platform, so taking the risk to upset Apple is big.
On the UA front, don’t panic. Revisit the impact of marketing on everything, paid or not.
[💎@1:15:24] Try to better understand the interdependencies between channels (e.g., email list, Instagram, offline, etc.) so you can assess how much you should be putting back into the paid acquisition machine.
The best product is not always going to win.
[💎@02:42] Organic is never enough to scale very significantly. Every single one of the huge apps that make it to the top of the charts are at least partly pushed by paid acquisition. Example: TikTok. You need a combination of a lot of paid acquisition, a lot of virality and a great product to stay at the top for long.
On the other hand, paid UA alone is not going to be enough.
First thing to look at are the retention metrics.
Should apps wait to do paid UA until they have some product-market fit?
[💎@06:00] You should start working on organics way before your app is live: building an email list, content, community, etc.
The need to get started on paid UA depends on how fast you need to grow. Example: VC money requiring to be more aggressive vs. being bootstrap.
[💎@06:50] You need to fuel your cohorts so you can work on improving retention. You want a minimum of new users every day so you can know if the latest release you ship is better than the previous one. If you don’t have a strong organic foundation, then paid UA is probably required.
[💎@07:17] Be careful about the biases that your early adopters can bring (if you have a strong organic baseline) because their behavior is not necessarily representative of the mainstream audience. This can make running paid UA healthy: to prove your hypotheses and optimize the product towards the mainstream audience.
[💎@09:37] You might want to exclude the cohorts for days where you get App Store features (if it represents a big portion of your audience), because the audience is often on the opposite end of early adopters and not representative of what your product should become.
If you're ambitious, there will always be a point where you will be stalling a bit and usually that’s when people turn to paid UA.
Some companies rely only on paid UA from the start, because they want to understand immediately what kind of audience they would get. This is too extreme as well. You want both organic and paid UA.
Small budget according to Thomas: probably more than $20k/month, it’s hard to do anything under $5k (maybe with ASA). $10k/month is when you can start getting something significant and get the same traffic you might be able to get later.
[💎@13:15] If you have a very small budget (like $500/month), focus on organic or specific paid actions (e.g., promotion on Reddit, very specific keywords on ASA, etc.). Do not go to Facebook or Google, because they are machine-learning driven.
Owned audiences that you build for yourself or earned audiences (through PR, featuring) are things you need to do from Day 1 of the project. They also take more time to bring results.
[💎@16:15] Building your owned and earned audiences is a must from Day 1 and all the way (even after you start paid UA). You need to find the ways that work for you to do that.
We call Organics what’s not attributed. But in reality, it’s more what you spend time against in order to get exposure. Example: getting featured requires spending time on Apple relationships, on building features, etc.
[💎@18:35] The only real lasting thing you get from App Store features is the “Editor’s Choice” badge that you can keep permanently on your listing. (Jacob, Revenue Cat)
The impact of featuring has gone down now. Features are also shorter, and they’re presented differently. They used to bring 100k installs but now App of the Day can bring 5k installs. Plus, there’s a random factor and as mentioned above they also bring a particular kind of traffic.
Features are not a sustainable strategy. So, you have to assume you won’t get them.
You don’t want to be chasing features that Apple wants you to implement instead of a user-centric approach.
The intent of users coming in through features is not as high. Turning curiosity into a paid subscription is harder too.
People have also already tried a lot of apps in each category
2015-2020: Facebook has been driving a huge part of the discovery. But now, discovery might be coming back to the App Store.
Job as a UA manager has already shifted since 2018, with the automation made by the networks taking over adjusting bids, deciding which creatives are shown to who, etc.
[💎@30:15] The core of the UA manager job today is producing, testing and deploying creatives. The role is also more strategic, as you have to think more like a data analyst so you can feed the right events to the machine learning algorithms.
Creatives are probably ¾ of the job: whether it’s talking to influencers, talking to a new agency, etc.
Examples of companies automating creatives in a very efficient way: copy.ai for copy, trypencil.com (you send your current assets, “brief the AI” and they create ads).
It is still a human task but part of it will be automated.
Who works on paid UA, product-led growth, understanding retention, etc.? For early stage it’s probably the same person that touches a bit all of this. This is how each person can find what they are the best at (for example, Thomas stepped away from creatives to make a bigger difference in onboarding, monetization, etc.).
A UA manager just focused on UA is now a terrible idea. It took a long time for people to realize that what matters is what happens down the funnel.
Acquisition and monetization are now more intertwined.
“Acquisition has become a business model competition”.
The difference you can make is on the creatives or in the monetization.
For games it used to be about whale-hunting: looking for a few very high-spenders. Facebook knows where the wales are, so it’s a matter of convincing the algo to drive them to you: through high bids and creative relevancy.
[💎@40:18] Marketers look a lot at the impression to install ratio, but Facebook is now taking engagement more into account (“like” vs. “comment” or “shares”, view rates, etc.) You need to trigger these positive signals through your creatives and keep people engaged on the Facebook platform for the algorithm to favor you in the auction.
“It’s your job to make Facebook happy”
The logic of bidding for subscriptions is different. The biggest networks didn't care about apps compared to gaming, DTC brands, etc.
Part of the role of a UA manager for a subscription app is to recognize that the FB algorithm is not made for subscriptions.
Something really useful for UA for a subscription app is to have multi-event funnels: are people engaging with the product, starting the trial, etc. You have to teach the FB algo a particular way. Example: Thomas used to map trial start to purchase (instead of “add to cart”) in order to get users that generally purchase (note: this does not work anymore).
[💎@46:08] Be careful about using “free” in your communications: start a free trial, etc. because you might attract users that just want free stuff. It’s probably best to have less users coming in but more qualified.
Attribution on apps is unique because there is the App Store blackbox between the ad click and the app open: no cookies, etc.
Initially there were workarounds, then the IDFA started being used. MMPs became absolutely critical for marketers.
Now is a very defining moment for the industry because acquiring users is going to be done in a very different way.
Facebook is not able to do attribution based on its own resources, even if it’s the platform displaying the ad and then having the Facebook SDK in the app. Pre-ATT, Facebook was not even showing ads to users if there wasn’t an IDFA.
Even though Apple might not be able to see what Facebook is doing on the backend, Thomas believes Facebook will not match data between apps and leverage that for optimizations, even if they could.
FB and Google were so powerful because they had data to know exactly what users were doing and could use this feedback loop to optimize for specific events.
[💎@56:30] Facebook’s algorithm wasn’t looking only at the data from your app, it was also looking at the data from pretty much all the apps in your category. With ATT, the duopoly is now attacked at the heart of their differential advantage.
Not everybody uses the same event, especially those that don’t have a huge budget. You need to figure out the event that has enough completion for the network to learn but enough correlation with long term goals like LTV. For many people that’s the free trial, but the free trial event can also be misleading.
[💎@59:15] With the new attribution system it’s going to be even trickier to optimize for the free trial event. Smart subscription advertisers often use a more complex event than the free trial, by layering it with some other things that are specific to their app. Example: free trials for users that have not cancelled within the first 2 hours + engagement.
[💎@1:01:00] Subscription apps with no free trials that don’t have the budgets to optimize for purchases have to find earlier signals to create a “bundle of events”: onboarding answers, accepting notification requests, etc.
[💎@1:02:21] Customizing too much the combination of events you’re optimizing for on Facebook can also prevent its algorithm from optimizing correctly, because the comparison with the data that it has from all the other apps becomes less relevant.
SKAdNetwork is the fundamental basis for attribution and every advertiser is going to use it.
You can modelize from the IDFA opt-in, but the opt-in needs to happen on both sides: on the source app (which displays the ad) and in your app.
[💎@1:06:00] There is no single source of truth anymore. What the ad networks do to optimize and the way advertisers interpret it is going to be more complex. Example: Google might use a blend of signals (SKAdNetwork, the IDFA when available, Android app, first-party data around contextual placements and other apps, etc.) and advertisers are going to need to use different data sets to interpret what’s happening (SKAdNetwork – not the full picture, IDFA, media-mix model or probabilistic model, known behaviors from specific-channels – like how fast users go through onboarding).
Fingerprinting is a subpart of probabilistic. Algolift assigns probability at the aggregated level based on what’s happening inside your app, and that’s compliant.
You live and die by the platform, so taking the risk to upset Apple is big.
On the UA front, don’t panic. Revisit the impact of marketing on everything, paid or not.
[💎@1:15:24] Try to better understand the interdependencies between channels (e.g., email list, Instagram, offline, etc.) so you can assess how much you should be putting back into the paid acquisition machine.
