Advertising on iPhones is About to Change Forever

How to make you customer acquisition thrive in the new iOS 'era'.

Hello 👋

Martin here. Welcome to another edition of Founders’ Hustle!

I write about the “hustle” of entrepreneurship and startup building frameworks.

Today I’m sharing methodologies startups can utilize to operate viable customer acquisition initiatives for iOS apps once Apple drops a game-changing update— its App Tracking Transparency framework.

I cover:

  • Why is App Tracking Transparency a big deal? 🤔

  • What impact it could have on your idea/app/startup. 📊

  • Approaches you can take to thrive in this new iOS ‘era’. 📈

If you have an iPhone app in the marketplace or are planning on releasing one, you’re likely astutely aware of the ‘bombshell’ update Apple is about to drop on iOS 14.5 as part of its App Tracking Transparency framework—IDFA deprecation. 😱

To those uninitiated, Identifier for Advertisers (IDFA) is a random device identifier assigned by Apple to a user’s device.

Advertisers, and indeed the entire mobile adtech ecosystem, use it to track the performance of ad campaigns down to the user level.

IDFA gives total visibility over how consumers engage with ads and behave in apps after downloading them—on an individual basis—which makes advertising campaigns highly effective.

Gigantic ‘device graphs’ (databases of IDFAs with behavioural data attached) are maintained by adtech vendors. They record who bought what, how much, and how often.

This has been utilized incredibly lucratively to target ads to the right people.

But, without IDFA, this all disappears. It goes dark. 🌒

There are no cookies to fall back on (like the web).

This is huge.

I can’t overstate how significant IDFA has been for the mobile software industry to date.

It has acted like rocket fuel in terms of enabling the app ecosystem to explode exponentially over the last decade or so.

Even giants like Facebook and Google are getting screwed on this.

But, in the coming weeks, for reasons that are outside the scope of this post, Apple will effectively be ripping out IDFA when iOS 14.5 goes live.

OK, that’s not strictly true. 🤥

It’ll still be possible for iPhone users to ‘opt in’ to sharing their IDFA with app publishers if they so choose.

But, let’s be realistic, the uptake will be tiny.

Particularly when the ‘one size fits all’ language request Apple forces developers to use is pretty scary. 👇

Apple isn’t completely abandoning advertisers, though.

There’s an olive branch. Actually, it’s more like a seedling. 🌱

It’s called ‘SKAdNetwork’—pronounced S-K-Ad-Network by adtech folks.

Indisputably, this is not one of Apple’s finest naming moments. 😂

But, it’s not consumer-facing and the app ecosystem kind of has to use it, so it doesn’t need to be.

I just call it “Skad” which, according to Urban Dictionary, seems far more inline with the app industry’s response to it. 👇

WTF is Skad?

In Apple’s words:

“[An] ad network API [that] helps advertisers measure the success of ad campaigns while maintaining user privacy.”

If diagrams are more your thing, here’s how it works. 👇

Basically, it’s an initiative to help advertisers measure ad campaign performance at an aggregated non-user level, so they’re not stumbling completely in the dark.

But don’t get too excited, the word “help” is doing a lot of heavy lifting there. 🏋️

Compared to what mobile advertisers are used to today, Skad is very much regressive in terms of performance measurement. Plus, it has a bunch of unorthodox privacy-related mechanics that add further friction for advertisers.

The list of limitations compared to the status quo is extensive.

Here’s a summary, paraphrased and expanded upon from AdMonsters, Branch, and Singular:

  1. 🤓 Granularity. Data is presented at the campaign-level only on an aggregated basis, and is limited to 100 campaigns per network, per app. This makes even extremely basic campaign optimization like ad creative testing difficult (though, Apple is making moves to improve this situation). Source App ID is supported, meaning visibility into campaign performance on an individual app level, so advertisers will know which apps their campaigns are performing better in.

  2. Attribution. Skad does not currently support deferred deep linking or view-through attribution (though, the latter is coming) and does not consider anything but the act of downloading and one conversion activity as attributable events. Multi-touch attribution is no longer a deterministic possibility, particularly at the user-level.

  3. 🔎 Post-install measurement. This is severely lacking. Right now app developers can track any and all post-install events they like. With Skad, it drops to one post-install event known as ‘conversion value’, which could be one of many possibilities and works on an unconventional sequential rolling timer logic until the most valuable action has been taken or the timer expires (more on that here).

  4. 😈 Mistrust. Skad passes attribution data to ad networks exclusively, who are essentially “grading their own homework”—this can lead to errors, deliberate or not. Currently, advertisers are accustomed to receiving their data from neutral third-parties (Mobile Measure Platforms like Kochava, Adjust, and Appsflyer) who verify the data and generally act in good standing on their behalf.

  5. 🚦 Postback problems. There will be a randomized delay of 24-48 hours between when installs and conversion value events occur and when they are reported via postback to the ad network, making it very challenging for app advertisers to optimize in real-time. Additionally, it doesn’t include a date parameter. This makes cohort-based optimization murkier—there’s no deterministic methodology to tie a postback to an install date.

  6. 🤗 Re-engagement. Unless app developers have the user’s permission, re-targeting lapsed customers through third-party services (Facebook, Google, etc) using unique identifiers like email, phone numbers, probabilistic matching, or any other methodology that constitutes personally identifiable information would be a breach of ATT policy.

  7. 😞 Unfairness. Apple will send conversion value and source app IDs in the postback to the ad network only if a certain amount of events have occurred—this is known as the privacy threshold. This means super early-stage startups with small ad budgets could get limited meaningful visibility on their campaign performance from Skad. The last time I checked the exact amount is yet to be disclosed.

  8. 📲 App-to-app only. Skad currently reports app-to-app events only. Again, this is changing. Soon, but likely after ATT goes into effect, it will support app-to-web events. Below is a helpful chart from Branch which clearly communicates the conversion paths with existing or planned support. 👇

For the remainder of this post, I’m going to focus on what Skad can support and I’ll wrap this into a wider narrative of opportunities and vulnerabilities.


First, some context. 🔦

This is not an isolated change. IDFA deprecation is an accelerant of a broader trend that has been happening for years. That is, audiences not sharing their data with publishers, developers, and—by consequence—the entire adtech ecosystem.

Ad blockers and data restriction plug-ins came onto the scene during the 2000s in response to the widespread usage of cookies that track web users across browsers. Not only do they block ads, publishers don’t receive any user-level data from that audience either.

These products went mainstream a decade later as consumers became a lot more consciously aware of how their personal data was being used.

Around that time, governments introduced stricter user-level data control laws [GDPR] and Apple rolled out privacy measures into Safari and iOS that were problematic for the adtech industry—blocking user-level tracking. ⛔

While Safari’s measures (Intelligent Tracking Prevention) were automatically deployed, meaning near full penetration across is usage base, Limit Ad Tracking (which restricts IDFAs being used by Apps) has been opt-in only for years.

Despite that, take-up rates have still managed to hit around 30% of iPhone users. A big piece of the pie—demonstrating consumers care about this.

With the deployment of App Tracking Transparency, IDFA anonymity will suddenly hit 90%+. 📈

And, Google is expected to follow suit with its equivalent mobile identifier, GAID, on Android.

Many app developers are understandably freaking out.

The funny thing is, even though I’m launching an app right now, I actually feel pretty comfortable about it. 😏

For five years, at my last startup, I helped many advertisers successfully market their products and services in an ultra low data environment at the user level.

Why was it ultra low? Because the audience was exclusively ad blocker users.

That’s right, my company built technology that served ads to ad blocker users.

Not only that, but, additionally, we figured out how to optimize ad campaigns so they yielded high RPMs for publishers and high ROI for advertisers.

This was all achieved with severe technical constraints due to the nature of ad blockers and the restrictions they impose.

We simply couldn’t utilize the vast majority of infrastructure the regular adtech ecosystem relied upon. We had to leverage what functionality was possible to maximum effect.

The severity of limitations was significant to the degree it was comparable to a post-IDFA opt-out ecosystem.

And, it worked out pretty well. ✅

So, I’m taking this experience and applying it to my current and future endeavours on mobile.

Next, I’ll share how, and, where I see risk and opportunities for early-stage startups, mixed together with insights from leading practitioners in the mobile user acquisition industry.


Strategies ♟️

With sudden disruptive ecosystem events like this, there are always winners and losers.

Some existing incumbents will suffer whilst others will fare better—either by fortunate happenstance or operational adaptation.

New startups that launch later this year and beyond won’t be encumbered by ‘IDFA process’. They’ll bring fresh new approaches and thinking that’ll enable them to thrive.

What you might call ‘ATT-first startups’.

But, how could this affect your startup? 👀

Let’s break it down with examples.

Specialists vs. Generalists

An absence of user-level data means laser-precision ad targeting is going to be much harder. Right now, app developers largely rely upon Google and Facebook to identify and reach their ideal target audience.

It’s totally automated and easy. But, it relies upon IDFA and it’ll soon be a luxury of the past.

This will have an outsized impact on startups that have a niche product and have to target a small, specific, and potentially context-dependent audience.

The unit economics just won’t work out as well, or at all, particularly if revenue per transaction is low. 📉

Examples? As a rough guideline, app experiences that appeal to like 1 in 1,000 people. Things like niche D2C (rock climbing brand), life-event services (wedding app), and food specialists (wheat-free vegan recipe app).

With the collective rollback of niche ad campaign ad budgets, startups offering products and services that have more of a general appeal will fill the void and thrive due to access to more inventory at cheaper rates.

Examples? As a rough guideline, app experiences that appeal to 1 in 100 people or less. Lowest common denominator products like dieting, mass-market games, or ecommerce.

And, they’ll do extremely well. 💥

With the absence of niche advertisers bidding high amounts on platforms like Facebook and Google to reach their target audience, it’ll be cheaper for generalists to acquire customers at scale and grow exponentially.

But, competition amongst this group will be fierce. 🔥

To gain an advantage in capturing consumer attention more cost-effectively, workable strategies for generalists include:

  1. ✍️ IP licensing. This generates higher click-through and conversion rates, making advertising more cost-effective. Example: signing a celebrity for an exercise app. Signing a movie franchise for a mobile game.

  2. ⚙️ Ad creative process innovation. Doubling-down on ad creative ideation, production, and testing in a more scientific way. This can improve marketing funnel efficiency, and, deliver higher click-through and conversion rates if app developers lean into the data more heavily and remove idealistic creative restrictions. This is something Wish has executed flawlessly.

  3. 🔀 Channel diversification. This involves broadening customer acquisition spend beyond the silo of traditional attribution marketing channels—social, app, and search ads—into TV, print, PR, OOH, etc (basically marketing mix modelling). This is realistically the domain of larger marketing teams who have the appropriate analytical processes and infrastructure in place. That being said, super early stage startups can experiment with alternative channels quite meaningfully since there’s less inbound customer noise, making attribution and ROI approximation feasible.

Conversely, ‘app first’ startups with niche target audiences have limited options if their go-to-market strategy is predicated on laser-targeted paid mobile user acquisition.

Similar to the above, they could try channel diversification.

But, this could be problematic depending upon the market and team size. Finding relevant and effective channels can be difficult, particularly when scaling.

And, it may simply not be cost-effective. 💵

Experimentation and investment will be needed in other areas—search engine optimization, app store optimization, communities, partnerships, word-of-mouth, guerilla marketing, etc.

Another option is to experiment with Apple’s own ad products like Search Ads.

As reported by Eric Seufert, earlier this year Apple “published documentation for a new attribution API for Apple Search Ads that contains more functional granularity than” Skad.

This includes creative ID and datestamp, meaning advertisers can perform better cohort and ad creative analysis and optimization. These are huge levers, and a massive advantage for using Apple’s own promotion services that could make or break the cost-effectiveness of running a paid acquisition campaign.

Returning back to Skad, a temptingly easy option is to change up the ad creatives so they attract the interest of a broader audience. But, due to product-ad misalignment, this will inevitably lead to low conversion rates, low customer lifetime revenue, and a lot of wasted advertising budget.

Therefore, an effective strategy could be 'product marketing’. 🧪

That is, instead of relying upon ad campaign testing to find the right audience for a product that’s developed independent of that process, app developers iterate and test the product alongside ad campaigns in symbiosis with one another.

It’s more of a holistic approach to customer acquisition where marketing and product people work together as an integrated process. The learnings of ad campaigns feedback into product development decisions, and vice versa. 🔁

In this instance, a product marketing strategy can be utilized to navigate to an app experience that has a broader reach (so less targeting is required). IP licensing can form part of that approach, too.

For example, a hardcore mobile strategy game could shift to a more accessible mid-core format. And, then license third-party IP to increase its appeal (click-through rates) in advertising creatives. 🎥

Business Models

Furthermore, IDFA deprecation will impact certain business models more than others.

App developers that rely upon advertising revenue for the bulk of earnings may find their unit economics turned upside down.

Why? The value of ad inventory is expected to decrease, at least in the short term, as demand is removed from the ecosystem in response to reduced targeting capabilities.

This puts a greater emphasis on in-app purchases and credit card transactions as more viable monetization models, which app developers can experiment with to plug unit economic profitability gaps.

Deplatforming and Replatforming

Another approach I see working, for some, is a shift in platform strategy.

App developers that exclusively publish on the Google and Apple App Stores now have more of an incentive to expand into web-based apps.

In that environment, the controls and restrictions of either of the tech giant’s app terms and conditions  do not apply— In-App Purchase fees have certainly been a point of contention.

Doing this has become a lot more feasible recently. The technology used to build web-based applications has come a long way in the last few years. In some examples, I barely notice a difference between a web-based app and a native iOS app.

Plus, incredible opportunities will likely emerge with other existing big platforms.

Tracking only becomes difficult when a user switches between apps, meaning if you keep the user in the same app performance measurement becomes a whole lot easier.

This provides a strong incentive for app developers to build and distribute products (apps — although you can’t call it that due to App Store T&Cs) within big established apps that already have huge audiences. This already happens to a certain degree, but IDFA deprecation will surely only accelerate it.

For example, social networks. Playing games, making purchases, consuming content, all within the confines of a big social network like Facebook, Instagram, or TikTok.

App developers could advertise their offering within that app, acquire customers, and measure performance pretty accurately.

In the past, companies like Zynga and Playtika built huge audiences operating games on Facebook’s desktop product.


Leveraging SKAdnetwork 💪

So what can Skad do, and how can startups utilize it to gain an advantage?

First of all, it’s important to recognize I’m ignoring other approaches app developers could take.

Once ATT drops, some companies in the app industry will experiment with workarounds that seek to probabilistically track app activity at a user-level. This is basically a shady workaround to Apple’s actions. 😬

I consider this a short term ‘cat and mouse’ exercise since it’s against the spirit of ATT and Apple will eventually shut it down. In the long run, it’s not maximally beneficial—I’ve been down this road before and don’t intend to return anytime soon.

I’m taking a much longer term view with iOS user acquisition frameworks, which, at this point, Skad is the measurement foundation of.

Below are key themes and actionable methodologies app developers can test and remain compliant with ATT.

Know Your Customer 🔎

First-party data is going to be super important—a material way to gain an edge.

That’s because iPhone users are not required to ‘opt in’ to share their personal information with the app developer itself, only with third parties.

This data sharing can be used for commercial gain. 📈

And, it could deliver outsized returns for companies with multiple apps—in part due to a mechanism called Identifier for Vendors (IDFV).

WTF is IDFV?

It’s basically a ‘company specific IDFA’ for publishers that have multiple apps in the App Store. When a user downloads one app, they can be tracked in the other apps owned and operated by the same company through IDFV.

This has numerous synergistic benefits, such as cross promotion and data-rich ad sales. But, IDFV as a mechanism generally favors large corporations over startups, since the former tend to have more apps live on the App Store.

That being said, there are examples of small, young startups that outmanuevered large competitors—with only a handful of app properties—by flooding the App Store with many more. It can become a competetive advantage in terms of carving out a market share through increased discoverability.

There are lots of ways knowing your customer better can make a positive impact to your business, here’s some examples:

  1. 📧 Collect email. When a user downloads an app, prompting them to sign up and input their email or phone number could be a big advantage. That’s not with a Single-Sign On partner like Facebook or Google. I mean, signing up directly with the app developer. Traditionally, this option has been widely avoided—particularly by content publishers—since it creates a friction point. But, IDFA deprecation changes things. For example, with IDFAs gone ad retargeting, and thus re-engagement campaigns, become insanely hard to execute. Emails or phone numbers can be used to re-engage lapsed customers through first-party communication.

  2. 🗄️ Rich First-Party Data. Collecting further personal information like address, age, interests, and tieing this back to in-app activity will further boost competitive advantage—particularly if the app is ad supported. App publishers can utilize this first-party data to sell targetted ad campaigns. It can also be useful to identify patterns amongst the most valuable users, the learnings of which can be plugged back into user acquisition and product development efforts.

  3. 👥 Two-way dialogue. With the luxury of utilizing Facebook and Google to find perfectly suited ‘golden cohort’ adopters for a prototype or early-stage product eroded, a greater emphasis is placed on better dialogue with ‘less than perfect’ users—those that need more hand-holding. Requesting feedback and communicating a contextual narrative around the mission and development of the app helps to fill in knowledge and experiential gaps that could otherwise cause users to leave and never return. This is something I have been experimenting with recently and is relatively easy to conduct to a basic degree.

Execution 🔧

How app developers execute on user acquisition and product development initiatives will have a big impact on degrees of success.

Similar to my experience building and optimizing infrastructure for ad blocking audiences, resourcefulness and creativity are key.

Leveraging constraints to the max will become a competitive advantage—especially for small startup teams that are unencumbered by the challenges of scale and inertia and can consequently be much more nimble.

Here’s some examples:

  1. 💬 Audience vs. context. Depending upon the product, it may make sense for app developers to weight their advertising efforts in favor of audience or context. For example, advertising a non-context dependent product (match-3 casual game) that’s popular with a certain demographic (middle-aged women) on publisher apps where that audience is the majority (Reader’s Digest). Or, advertising a context-dependent product (car finance) on publisher apps (AutoTrader) that attract consumers in the right state of mind to consider it at that moment.

  2. 🧪 Ad creative testing. This is where early-stage startups with a new product have an advantage over large incumbents. Companies with multiple ideal customer profiles and a global presence on the App Store have a huge amount of scenarios to optimize ad creatives for. With a limit of 100 campaigns IDs per ad network and no creative ID visibility, this is a significant challenge. Conversely, a startup with one app, one ideal customer profile, and one target launch market has a lot of wiggle room. Depending upon how ad networks chose to orchestrate campaign IDs internally, there should be material bandwidth to obtain a sense of visibility into ad create performance by testing one ad creative per campaign, across many campaigns, and treating all campaigns collectively as a traditional ad group.

  3. 💳 Monetization adjustment. For apps that generate revenue via purchases, there’s an advantage for monetization models that generate fairly consistent sales value per customer. For example, an app that charges a subscription has a maximum yield of $X per annum per customer, and the average $X revenue per paying user will be somewhere below that. Revenue is fairly evenly distributed across its customer base, a strong purchase signal (subscribing) can occur pretty quickly after install, and churn rates are predictable. This makes customer acquisition investment relatively reliable. Conversely, an app that charges per transaction and illicits most sales from a small set of customers that appreciate over what can be long periods of time, there’s significantly less predictability and a dependence on ‘whale hunting’ for more top 1-5% spenders.

  4. 🧰 Conversion value engineering. Since post-install activity is the best indicator of ad campaign performance and Skad only supports the communication of one conversion value out of a possibility of many (up to 63)—the significance of which are all determined by the app developer and subjected to the aforementioned timer logic—structuring conversion value logic optimally will a huge part of iOS campaign optimization. The way it’s designed generates an advantage for apps that illicit a monetization event quickly. The sooner the better, so the time logic ends and the postback containing the monetization event gets sent back to the ad network. This creates a quick feedback loop for advertisers to optimize campaigns. It also permits a crude form of cohort-based analysis, since an app can be designed in a way that encourages the majority of users to trigger the primary event within a short time period, allowing for date of install approximation.

  5. 🔮 Value prediction. With ad campaign costs a known and visible expense in app developers’ ad network and mobile measurement platform dashboards, predicting expected return on ad spend (ROAS) within the constraints of Skad functionality becomes a critical objective. The strategy on this can go deep and get complex quickly , but, to summarize, the core mechanism with which app developers have to build a predictive customer lifetime value (LTV) framework around is the conversion value. Industry expert Eric Seufert explains how: “There are a number of ways that a developer might impute an LTV to some moment or user state in an app. The first is aggregating up average values based on conditional probabilities with respect to combinations of “events” completed by the user in the app. An example might be: what’s the average monetization level of a user for a shoe-buying app at the point of adding a pair of shoes to the cart, given that they have viewed 10 pairs of shoes prior to doing that?”

This last example completely differs to the status quo, in which developers send a stream of in-app events back to the ad networks, who utilize sophisticated internal machinery to optimize targeting around. 

As Eric suggests, going forward, it will be more productive to structure events so they correlate to a behavioural user state (pooling them into an aggregated group that has performed X, Y, and Z actions) than attempting to mirror existing event apparatus in Skad.

It’s a fluid situation. At some point I’ll write a follow-up post with tried and tested approaches and the latest insights from practitioners in the industry.

Until next time!

Martin 👋


Not subscribed to this newsletter? Let’s fix that right now 😉👇