The App Tracking Transparency recession

My Federal Reserve driver, Mr. Peña, provided the conclusive evidence of victory. On the way to speak at a big Washington dinner, I spotted a book on the front seat next to Mr. Peña with the title How to Live with Inflation. Shocking that my own driver had no confidence in me! But, Mr. Peña explained, he’d only bought the book because its price had been marked down to $1.98 from $10.95. –Paul Volcker, Keeping At It

In this piece, I’ll argue that Apple’s App Tracking Transparency (ATT) privacy policy, along with consumer preference changes reflecting a reversion to pre-COVID behaviors, has created a recession within the social media advertising economy and certain other advertising-dependent categories. I’ll propose — with analytical support — that macroeconomic frictions mostly have not harmed the broader digital advertising sector, and that, to the contrary and excluding social media advertising, the current digital advertising market is robust, as are other corners of the consumer economy, supported by a resilient American consumer. Note that I focus on the US economy in this piece, as the primary market for the companies being considered is the United States.

While undeniably imprecise, since multiple countervailing forces are at play in the post-lockdown economy, I believe this analytical exercise is important to undertake, for two reasons:

  1. An assessment of ATT’s economic impact provides guidance for estimating the impact of the eventual deprecation of third-party cookies in Google’s Chrome Browser;
  2. If a recession does occur in the United States and elsewhere, it will manifest further pressures on the digital advertising ecosystem, for which advertisers should be prepared.

This piece is comprised of six sections:

  1. Introduction
  2. The Fed’s crusade against inflation
  3. A short primer on digital advertising and Apple’s App Tracking Transparency policy
  4. The resilient American consumer
  5. Divergent paths: Social Media advertising and Everything Else
  6. Everything is an Ad Network


Price inflation on the basis of the consumer price index (CPI) peaked in June 2022 but remains stubbornly high — standing at 6.5% in December, the last month for which it has been reported as I write this, and down from 7.1% in November. Sustained inflation above some target rate is an adverse condition in an economy; Milton Friedman famously quipped that “inflation is always and everywhere a monetary phenomenon,” meaning a function of the money supply. But price inflation can be a sign that an economy is overheating — that is, that demand grows at a pace that can’t be satiated with productive capacity (supply).

The broader Big-T Tech sector has felt a tremendous amount of pain in terms of market capitalization contraction over the past year as the Fed’s longstanding Zero Interest Rate Policy (ZIRP) gave way to an unprecedented regimen of aggressive interest rate hikes. But the social media advertising sector in particular experienced acute revenue shocks over the course of 2022, with revenue growth slowing dramatically and even pitching into negative growth in some cases.

Many commentators point to the deterioration of Big-T Tech P/E multiples as a sign of a downturn in the sector, but many public tech companies featured negative unit economics before COVID that did not improve during the pandemic, even if revenue growth accelerated. It’s entirely logical that these businesses see their market capitalizations recede as interest rates rise.

Many companies across the technology ecosystem — but especially those involved in eCommerce, which benefited propitiously from the COVID pandemic as consumer retail spend shifted online — have addressed pandemic-era employee growth through consequential layoffs. But the broader labor market in the United States remains robust. The Bureau of Labor Statistics revealed last week, on January 6th, that the unemployment rate sat at 3.5% in December — a 53-year low.

But one might assume that the economy has utterly imploded from reading the Q3 earnings call transcripts of various social media platforms. Alphabet, Meta, and Snap, in particular, cited macroeconomic weakness, headwinds, uncertainty, challenges, etc. in their Q3 earnings calls.




But aside from various corners of the economy that are particularly sensitive to interest rate increases, such as Big-T Tech, homebuilding, and finance, much of the consumer economy is robust. Nike reported 17% year-over-year revenue growth in its most recent earnings release last month; Costco reported year-over-year sales growth for December of 7% on January 5th; Walmart’s 3Q 2023 results, reported in November 2022, saw the retailer grow year-over-year sales by 8.2%; and overall US holiday retail spending increased by 7.6% year-over-year in 2022, beating expectations. Of course, these numbers are nominal and not real, but for comparison: holiday retail sales in 2008 were down between 5.5 and 8% on a year-over-year basis, and the unemployment rate in December 2008 stood at 7.3%. And as I’ll unpack later in the piece, many participants in the broader digital advertising ecosystem saw strong revenue growth in 2022 through Q3.

So what’s the source of the pain for the largest social media advertising platforms?

Apple introduced a new privacy policy called App Tracking Transparency (ATT) to iOS in 2021; with iOS version 14.6, that policy reached a majority scale of iOS devices at the end of Q2 2021, in mid-June. ATT fundamentally disrupts what I call the “hub-and-spoke” model of digital advertising, which allows for behavioral profiles of individual users to be developed through a feedback loop of conversion events (eg. eCommerce purchases) between ad platforms and advertisers. In this feedback loop, ad platforms receive conversions data from their advertising clients, they use that data to enrich the behavioral profiles of the users on their platform, and they target ads to those users (and similar users) through those profiles. I’ve written extensively about how ATT disrupts the digital advertising ecosystem, but the disturbance is most pronounced for social media platforms as I’ll describe later in the piece. The shocks of ATT became discernible in Q3 2021 (the quarter after ATT was rolled out to a majority of iOS devices) but were substantially troublesome for Meta in particular in Q4 2021. The disruptive forces of ATT have compounded over time.

My general belief is that the impact of ATT has been underestimated; ascribing the advertising revenue headwinds being felt most profoundly by social media platforms and other consumer tech categories with substantial exposure to ATT to macroeconomic factors is misguided. The general thesis of this article encompasses three points:

  1. That the broader digital advertising market is not in the midst of a cyclical downturn;
  2. That social media advertising platforms and the other consumer technology categories that were acutely exposed to the disruptions of ATT are experiencing revenue frictions that are principally driven by the combination of ATT and changing consumer preferences related to the abatement of the COVID pandemic and not macro factors;
  3. That a potential recession in 2023 would produce additional revenue pressures on those same social media advertising platforms and ATT-disrupted consumer technology categories, as well as the broader advertising market.

Layoffs and stock price contractions create real human pain, and I don’t intend to minimize that with this piece. The purpose of this piece is to attempt to disentangle the impact of various factors on the social media advertising space and to highlight the specific detrimental impact that ATT has imposed there, while also highlighting the relative health of other portions of the digital advertising landscape. My motivation for writing this piece — which puts forth a nuanced, multi-faceted argument — is to draw attention to the various considerations that should be acknowledged by marketers if ATT, and not macro factors, is responsible for the degradation of efficiency in social media advertising:

  • If a recession does take form in 2023, then even more challenges sit on the near-term horizon for advertisers;
  • The acknowledged impact of ATT provides helpful guidance for estimating the effect of third-party cookie deprecation in the Chrome browser and the deprecation of the GAID on Android, in addition to the rollout of the DMA and DSA in Europe and potential Federal privacy legislation in the United States;
  • Since ATT is systemic and permanent, as will be other policy and legislative initiatives, and not a fleeting phenomenon (like a downturn in the economic cycle) that advertisers can simply wait out, then advertisers should shift their attention to resilient methodologies related to measurement and audience targeting that will endure for the long term.

The Fed’s crusade against inflation

Persistently elevated inflation has defined the American — and global — economy since the Spring of 2021. The COVID pandemic and Russia’s invasion of Ukraine (and the resultant energy crisis it caused in Europe) are often cited as the primary and principal factors that caused inflation to rise to levels not seen in over four decades, since Fed Chairman Paul Volcker tamed the runaway inflation of the late 1970s and early 1980s. Apart from these two causes, little consensus exists around how the current inflationary environment was shaped by the various knock-on effects of those causes: a surplus of demand sparked by government stimulus unleashed across the world, a supply chain crisis engendered by COVID-related stay-at-home policies, or some combination of both. For consumers, the underlying agitations and components of inflation are beside the point.

The most recent US consumer price index (CPI) data released by the Bureau of Labor Statistics puts year-over-year price inflation in December at 6.5%, which represents a decrease of 0.1% from November’s value. This is a decrease from 9.1% in June 2022, when inflation seemingly peaked in the United States. The Euro-area inflation rate seems to lag that of the United States, with inflation there having peaked thus far at 10.6% in October 2022 and sitting at 10.1% in November, but with a broad range across constituent countries.

Inflation is a disaster for consumers: it erodes the purchasing power of wages and savings, it puts economic pressure on households because prices increase in real-time while wages are static in the short term, and it confounds household planning (such as buying a home or starting a family). The US Federal Reserve’s mandate as an organization is twofold: to achieve 1) full employment and 2) price stability (ie. to keep inflation relatively consistent). The Federal Open Markets Committee, or FOMC, is a committee within the Fed that is responsible for directing monetary policy through open market operations, which entails buying and selling US government securities. The FOMC also sets targets for the Federal Funds Rate, which is the interest rate that financial institutions charge each other on overnight deposits and which percolates into all corners of the economy, such as credit card and mortgage interest rates. Back in 2012, the FOMC established an explicit year-over-year inflation target of 2%. Inflation is currently much higher than 2%, which has prompted the Fed to take decisive action in reducing it. (For more background on the Federal Reserve, I enthusiastically recommend the book The Lords of Easy Money)

The target for the Fed Funds Rate was reduced dramatically over the course of 2007 and 2008 to roughly 0 to address the Global Financial Crisis; the Fed began raising rates in 2016, but it reversed course in the summer of 2019 when it began a rate-cutting regimen in the face of a slowing economy.

As COVID took root across the globe, the Fed again cut the Fed Funds Rate aggressively to roughly 0, where it remained until March 2022, when, in light of elevated inflation and Russia’s invasion of Ukraine, the Fed raised rates for the first time since December 2018.

Per the chart above, the Fed increased the target range for the Fed Funds rate seven times in 2022:

  • March 17th, 2022: 25 basis points, from 0-0.25 to 0.25-0.5;
  • May 5th, 2022: 50 basis points, from 0.25-0.5 to 0.75-1.0;
  • June 16th, 2022: 75 basis points, from 0.75-1.0 to 1.5-1.75;
  • July 28th, 2022: 75 basis points, from 1.5-1.75 to 2.25-2.5;
  • September 22nd, 2022: 75 basis points, from 2.25-2.5 to 3-3.25;
  • November 3rd, 2022: 75 basis points, from 3-3.25 to 3.75-4.0;
  • December 15th, 2022: 50 basis points, from 4.25-4.5.

A short primer on digital advertising and Apple’s App Tracking Transparency policy

In June 2020, at its annual developer conference, WWDC, Apple announced a new privacy for iOS: App Tracking Transparency, most commonly referred to as ATT. The ATT policy requires apps to present users with an opt-in prompt for what Apple describes as “tracking,” which relates to the practice of sharing user data across contexts and with parties that a user may not have interacted with directly. On mobile, cross-context data sharing is mostly accomplished through Mobile Advertising Identifiers, or MAIDs. The MAID on iOS is called the Identifier for Advertisers, or the IDFA (on Android, the MAID is called the Google Advertising Identifier, or the GAID).

With its ATT policy, Apple states that users must proactively opt into having their data shared with third parties, such as with advertising platforms. Prior to ATT, a user’s IDFA was used as a common identity that allowed their activity in an app to inform advertising decisions made on advertising platforms like Meta’s, Snap’s, TikTok’s, YouTube’s, etc. Note that when a user opts out of the ATT prompt for a specific app, the ATT policy dictates that the app may not share any identifier — not just the IDFA, which is restricted if a user opts out — for that user with a third party for the purposes of advertising targeting. For instance, sharing a user’s email address, real name, phone number, etc. with a third party is strictly prohibited when a user has opted out of the ATT prompt.

An advertising efficiency feedback loop is enabled by unique identifiers: an advertising platform exposes a user to an ad, the user clicks on the ad and lands at a mobile app or website, and that mobile app or website reports that user’s activity on their property back to the advertising platform. This feedback loop allows the advertiser to do two things:

  • To understand what other apps or websites it should promote to that user in the future, based on that user’s interactions with the current app or website;
  • To identify other users to which the advertiser’s product should be promoted, based on the advertising platform’s knowledge of how the current user interacted with the product that was advertised to them.

The diagram below, from this post, describes this feedback loop (Facebook is used as an example in the diagram, but many other ad platforms could be substituted in its place):

It’s important to note or reiterate a few characteristics of the ATT policy:

  • ATT moderates the use of any identifier, including emails, for advertising measurement and targeting, not just the IDFA;
  • ATT exclusively regulates ads displayed in mobile apps, but it applies to ads that lead to app and web destinations. But in-app advertising represents a majority share of the digital advertising market: the vast majority of time spent on mobile is captured by mobile apps (and not the mobile web).

Digital advertising platforms engage in three separate activities:

  1. Targeting: determining the appropriate group of people, or audience, to which an advertiser’s ads should be exposed;
  2. Ad serving: filling an impression on a property with an ad;
  3. Measurement: determining the value (in observed or expected terms) of a served ad.

ATT fundamentally prevents ad targeting at the level of an individual user: since user profiles cannot be developed through aggregated, user-level conversion data — like purchases, registrations, ‘add to cart’ actions, etc. — then the user-level targeting mechanics previously employed by large platforms are rendered inoperable. These platforms must revert to group-level targeting, which almost by definition is less precise and efficient than user-level targeting, especially since the broad demographic data — such as age, gender, and geography — available to display ad platforms generally carry very low predictive power for purchase intent. I unpack the mechanics of this dynamic in Why did CPMs increase following App Tracking Transparency?.

As I discuss in IDFA deprecation: winners and losers, ATT is most disruptive to companies that fit a specific profile. The primary ad platforms and ad networks of focus in this piece are Meta, Snap, and YouTube, since they are public companies and they facilitate the hub-and-spoke feedback model described above; Google Search is largely exempt from ATT, as are desktop and browser-specific ad networks. The principal advertiser categories of focus are eCommerce and mobile apps (but especially mobile gaming).

For ad platforms and ad networks, the magnitude of exposure to ATT is determined by whether they:

  • Serve in-app display ads that rely on user-level behavioral profiles and not any other type of context (eg. search queries, site content) for targeting;
  • Employ a pixel, in-app SDK, or server-to-server API to collect user-level conversion signals from advertising clients;
  • Allow user-level and “lookalike” targeting facilitated by user identifiers such as MAIDs, email addresses, etc.
  • Aggregate “device graphs” of MAIDs for the purpose of targeting specific users in the open programmatic environment.

For advertisers that buy in-app inventory, the magnitude of exposure to ATT is determined by whether they:

  • Rely heavily on “lookalike” audience targeting on large advertising platforms, facilitated by the provision of user-level identifiers to those platforms;
  • Rely heavily on user acquisition or “retargeting” through the provision of user-level identifiers to ad networks;
  • Operate niche products that require specific behavioral knowledge of potential users for profitable advertising unit economics;
  • Operate products that see a long-tail distribution of Lifetime Value, or LTV, where a very small minority of highly-engaged users generates the vast majority of revenue.

Note that the impact of ATT is not just expressed within advertising targeting: it also disrupts measurement, or profitability accounting. For mobile app advertising, the tool that Apple provided to replace the user-level conversion observation that was previously facilitated with the IDFA is called SKAdNetwork. SKAdNetwork was profoundly inadequate for proper advertising measurement at the time that ATT was launched; Apple released an improvement to SKAdNetwork late last year, but the degree to which it will benefit measurement for app advertising campaigns is as yet unclear. And notably, Apple’s Private Click Measurement (PCM) specification, which has influenced web advertising campaign measurement for ad platforms for compliance with ATT, is considerably more restrictive than the pre-ATT measurement status quo.

The social media advertising sector is currently under substantial duress with respect to revenue growth — and, in some cases, is experiencing year-over-year revenue contraction. I’ve covered these revenue hardships on a quarterly basis for Meta (Q4 2021, Q2 2022, Q3 2022), Snap (Q2 2022, Q3 2022), and Alphabet (Youtube) (Q1 2022, Q3 2022). Note from the previous section that the Fed only began hiking rates in March 2022 (late Q1); Snap first reported that its revenue was impacted by ATT in its Q3 2021 earnings report, which caused its stock price to drop by 25%. Snap missed revenue estimates in Q3 2022, one year later, causing its stock to drop 30%. Meta’s stock dropped by 26% after reporting Q4 2021 results, which showcased weaker-than-expected revenue growth. Meta’s stock dropped by 25% after it reported its Q3 2022 results.

Some might look to anguish in the social media advertising space and consider it an isolated problem. But slowing revenue growth, or revenue declines, on these platforms is, by definition, the consequence of a reduction in spending from advertisers. Advertisers are spending less money on these platforms because the economics of their advertising budgets has degraded. Direct response advertisers spend against return-on-ad-spend (ROAS) targets, and budgets and ROAS tend to exhibit an inversely correlated relationship: as budgets increase, ROAS decreases, and vice versa. As a result of a decline in efficiency (ROAS), advertisers will decrease spend. And if no other platform exists to absorb that spend, then that spend can simply evaporate, meaning those advertisers buy less revenue.

This has demonstrably happened. Newzoo estimates that the mobile gaming category, which is as large as the PC and console gaming categories combined, contracted by 6.4% in 2022; (formerly, AppAnnie) estimates roughly a 5% contraction for 2022, with a 3% decline in 2023. While it’s difficult to parse apart the effects of consumer preference changes related to COVID and ATT on mobile gaming, it’s interesting to note that 2022 is the first year for which the mobile gaming market has ever declined; rather than seeing a pronounced jump in revenue in 2020 or 2021 from the COVID pandemic, mobile gaming mostly grew at a predictable clip for those years, per the chart above.

Direct-to-Consumer (DTC) retailers, which are generally dependent on social media advertising for sales, have likewise seen a diminution in advertising efficiency since ATT was introduced. I discuss the erosion of advertising performance and efficiency in the eCommerce category with the owner of Common Thread, a large DTC advertising agency, in this podcast. According to the Wall Street Journal, nearly 50% of eCommerce advertisers polled had decreased spend on Facebook by more than 25% as a result of ATT.

It should be noted that, while ATT only impacts the iOS ecosystem, iOS captures both the majority of smartphone share in the United States (but a minority worldwide) as well as the majority of in-app revenue, which seems reasonable to extrapolate to mobile eCommerce retail purchases. Critically, the share of advertising revenue generated on iOS devices shrank from 60% in 2020, before ATT was introduced, to 46% in Q2 2022. Again, it’s reasonable to deduce that this reduction wasn’t the result of a share shift — that is, advertisers shifting budgets sustainably to Android — but rather of a dissipation of ad spend on iOS.

The resilient American consumer

The consumer is the bedrock of the American economy, accounting for 2/3rds of GDP. If consumers capitulate and stop spending in light of high inflation or other economic pressures, the economy will falter. An advertising recession seems predicated on that eventuality, as I point out in my Advertising Strategy in a Recession presentation. Is the American consumer healthy?

On January 6th, the Bureau of Labor Statistics released its Employment Situation Summary report for December. The BLS reports that the US economy added 223,000 non-farm payroll jobs in December, down from a revised 256,000 in November, which results in a three-month moving average of roughly 247,000 job additions per month. This represents a substantial decline from the three-month average for July through September, which stood at 366,000 new jobs per month. While the job growth for December was higher than consensus expectations of 200,000 new jobs, the decline in job growth over time is reflective of a moderating labor market, which is likely what the Federal Reserve hopes to achieve with its interest rate policy: slowing job growth creates slack in the labor market, which ultimately could lead to decreases in inflation. Per the section above on inflation, this seems to be taking shape.

To that point: as noted earlier, the unemployment rate in December decreased to 3.5%, which is a 50-year low. While this could be considered at odds with the Federal Reserve’s policies, given the focus on reducing inflation, the unemployment rate inched lower for reasons that can be considered optimistic (with respect to decreasing inflation). Labor force participation increased for the first time since August, and employment increased by roughly 717,000 in the household employment survey that had exhibited weakness in recent months: growth from September to October was -257,000, and growth from October to November was -66,000. The payroll and household unemployment surveys use different methodologies, and when they conflict, economists tend to put more weight on the establishment survey (which in December reflected +223,000 jobs), since it draws from a larger sample. But the household survey can be more sensitive to changes in the economic environment, and a consistently reported decrease in employment might have signaled a downturn in the economy.

The jobs report revealed specific points of weakness at the industry level: declines in Information and Professional Business Services are related to the layoffs in tech that took place in November (Meta: 11,000 employees; Stripe: 1,100 employees; Twitter: 3,700 employees), and declines in Temporary Help Services relates to a generally slowing hiring environment.

Of course, employment metrics, especially unemployment, are lagging indicators, but various forward-looking indicators also seem to portend a slowdown in labor markets that the Fed likely views as an endorsement of its hawkish policies to date. Hourly wages increased by 0.3% in December on a month-over-month basis, below the 0.4% expectation and last month’s metric of 0.5%, and 4.6% on a year-over-year basis, below last month’s metric of 5.1%. Hours worked also decreased sequentially to 34.3 in December from 34.4 in November and 34.5 in October. Again, these are positive signals that the Fed’s delicate balancing act of attenuating inflation through creating slack in the labor market while also staving off a sharp increase in unemployment — which would lead to a recession — is working. Another forward-looking indicator, the ISM Manufacturing PMI, which captures changes in manufacturing production levels, declined in December; the New Orders component of the PMI declined to its lowest level since 2008 (excluding the early stages of the COVID pandemic).

The chart above, from Pantheon Macroeconomics, suggests that the gap between current (Q4 2022) levels of wages (Average Hourly Earnings, or AHE) and those from the 2010-2020 period consistent with low inflation and full employment is shrinking.

Aside from unemployment rates sitting at 50-year lows, other economic indicators support the notion that the American consumer’s finances are not currently in a state of distress that would comport with claims that macroeconomic headwinds are suppressing advertiser spend on social media. Critically, American consumers increased their household savings meaningfully during the COVID pandemic, and “excess savings,” or the absolute dollar value of savings above the counterfactual case where COVID had not occurred, stood at a substantial level of $1.7TN in mid-2022, although the savings rate has declined precipitously and the stock of excess savings is being depleted. While excess savings are being used by consumers to cushion the impact of inflation now, many economists expect that the stock of excess savings will be depleted toward the end of 2023.

Note that while the majority of this stock of excess savings has been contributed to by households in the Top Quartile of income, it is exactly that group that represents the majority of consumption:

According to Moody’s Analytics, nearly 60% of the $1.7TN of excess savings is held by the top 20% of households by income. And although the personal savings rate has fallen dramatically from mid-COVID highs, as of November, it was still positive.

And another signal of the general financial health of the American consumer is relatively low debt delinquency rates, especially for credit cards (note that Student Loan and Mortgage delinquency rates are likely low as a result of government forbearance and relief programs). The Q3 2022 rate of credit card delinquencies sits below both the Global Financial Crisis peak of 13.7% as well as the pre-pandemic level of 8.4%.

My interpretation of these data points is that the US consumer is in a relatively comfortable position currently, and that the encouraging jobs report suggests that any potential recession might be faint, although this is by no means a consensus view. Note that while consumer debt does sit at an all-time high as a percentage of disposable income, most of that debt has a fixed rate and is not necessarily sensitive to changes in interest rates.

One indicator that does suggest that a recession will manifest is the yield curve, which is currently inverted, meaning short-term, often 2-year, treasury notes pay higher yields than 10-year treasury notes. An inverted yield curve has preceded each of the last 10 recessions; as such, it is considered a reliable indicator. But the average lag time between yield curve inversion and recession has averaged 22 months, although for the last six recessions, the range of lag times has widened.

The two-year/10-year yield curve inverted briefly on April 1st of last year, but it reverted again and much more deeply in July, with yields on two-year treasuries sitting 80 basis points above yields on 10-year treasuries. In October, the yield on three-month treasuries eclipsed that of 10-year treasuries; the lag time on this particular yield curve inversion to recession ranges between six to 15 months.

These signals can be interpreted to mean that a recession will happen at some point in 2023, depending on the curve being considered. But according to other meaningful measures of economic health, the American consumer is financially vigorous: while wage growth is slowing, unemployment is at historic lows, credit card default rates sit below pre-pandemic levels, and excess savings are being used to absorb higher prices as inflation abates. The conditions don’t seem to exist that would explain a broad withdrawal by the consumer to result in systemic weakness in the digital advertising market.

Divergent paths: Social Media advertising and Everything Else

Perhaps the clearest sign that the social media advertising space is uniquely impacted by ATT and not macroeconomic factors is simply that it has seen an exceptional and almost singular contraction in revenue within the broader digital advertising ecosystem.

Recall the criteria I identified in the introduction that determine whether an advertising platform is exposed or not to ATT. It’s important to note that brand advertising is mostly if not entirely immune from ATT because brand advertisers, as opposed to direct response advertisers, generally target very broad audiences and aren’t optimizing advertising spend for conversions. Brand advertising representation differs by platform; while platforms don’t release these metrics, I crowd-sourced estimates of brand spend from the Mobile Dev Memo community and have charted them above (the Twitter data is sourced from Twitter’s 2020 Analyst Day event, at which it revealed that 85% of the advertising spend on its platform is sourced from brands. Recent publicly-circulated estimates are lower but not officially sanctioned).

The degree to which a company 1) meets the requirements that I proposed above for exposure to ATT and 2) is reliant on direct response advertising (that is, sits at the right-hand side of the graph above) should determine the magnitude of disruption it experiences from ATT. In attempting to parse apart the impact of ATT versus the impact of inflation on the advertising sector, it’s possible to simply compare the companies that would presumably be most impacted by ATT to the companies that would have very little exposure to ATT.

The first such example is straightforward: Google Search vs. YouTube. As I noted previously, Google Search is observably exempt from the restrictions of ATT because Google Search is primarily conducted in a browser, and Google Search results are primarily delivered as a result of search queries and first-party, on-site data (while the number of Google searches conducted on Google’s iOS app is not public information, my sense is that it is a vanishingly small proportion of the total). What’s evident in the chart above is that both YouTube and Search see year-over-year growth decline from a peak in Q2 2021. Each subsequent quarter after Q2 2021 faces comparables to mid-COVID quarters, in which engagement in consumer technology broadly had increased precipitously. But YouTube’s year-over-year growth rate slows more dramatically than Search’s after ATT reaches majority scale in Q2 2021. And for the first time in the product’s history, YouTube’s revenue declines on a year-over-year basis in Q3 2022, the last quarter for which data is available.

It’s important to keep in mind that YouTube’s business is not entirely mobile: YouTube is also accessible from the browser, and that use case is not moderated at all by ATT. Also note the estimated 70% proportion of brand spend on YouTube from the chart above, which likewise is unaffected by ATT. If a reversion to post-COVID behavioral norms explains the decline in Search revenue starting in Q2 2021 (because that quarter compares to Q2 2020, which was early in the pandemic), then my assertion is that ATT explains the more aggravated decline in growth in YouTube relative to Search. And ATT doesn’t impact the entirety of YouTube’s iOS business (given the browser product).

The vast majority of Meta’s advertising revenue is generated from its mobile apps; in Meta’s (then, Facebook’s) Q2 2019 10-Q SEC filing, the company states that, “we estimate that mobile advertising revenue represented approximately 94% of total advertising revenue.” Meta doesn’t break out engagement metrics across mobile web and its apps, but my sense is that nearly all of the company’s mobile advertising revenue is delivered by its apps. With that context in mind, the above chart is stark: as with YouTube and Search, Meta’s growth peaks in Q2 2021, the same quarter in which ATT reached majority scale on iOS devices, and subsequently declines progressively, reaching negative growth (that is, year-over-year revenue declines) in both Q2 and Q3 2022. Again, my sense here is that Search is a reasonable comparison to Meta when considering the impact of a reversion to pre-COVID norms on growth. Given Meta’s appreciable exposure to ATT in light of its advertiser mix favoring direct response, as well as the fact that its tools for utilizing conversion data lead the industry, the sharp decline in growth and indeed revenue contraction seems rationally attributed to ATT.

Snap’s revenue profile tells a similar story, albeit from a higher pre-COVID and pre-ATT growth baseline: year-over-year revenue growth peaks at 116% in Q2 2021, the same quarter in which ATT reaches majority scale on iOS devices, and declines sharply thereafter, registering at 6% in Q3 2022. Note that Snap reduced its guidance for Q2 2022 in May of that year — in the middle of the quarter — citing deteriorating macroeconomic conditions. But this reduction in guidance came on May 23rd, roughly one month after Snap announced Q1 2022 results and provided its initial guidance for Q2.

An interesting comparison for these companies when evaluating the impact of COVID, ATT, and theoretical macroeconomic weakness is any advertising business with very little to no exposure to ATT. The Trade Desk is one such company: it is a publicly-traded demand-side platform (DSP) that primarily empowers browser-based digital advertising, although a significant component of its recent growth has been delivered by CTV. In the company’s Q3 2021 earnings call, The Trade Desk’s CEO, Jeff Green, responded to a question about the potential impact of ATT:

So about 10% of our spend uses IDFA. And because we’ve had limited targeting on that 10% for quite a long time, continuing to limit it or limited in a new way, doesn’t have a material impact to our business. Because we’re looking at roughly 12 million ads every single second, when you take 1 million-ish of those and say, we’re going to allow less data to be used on those. We just look more carefully for gems in the other 11 million.

So I don’t expect it to have a material impact on our business the way that it will others. So when you hear Facebook, talk about having a big impact, just remember, they’re 70-ish percent mobile, not 10% IDFA. So very different impact to us. That said, I do believe this is Apple trying to mess with Facebook’s business and Google’s business, they’re much more committed, I think, to payments than they are to the advertising ecosystem.

This sentiment seems to materialize in the company’s revenue growth; there is no discernible impact from ATT, which reached majority scale on iOS devices Q2 2021. Yet there’s no sign of macroeconomic pressures on the company’s revenue growth, either. In fact, The Trade Desk’s revenue grew to a greater extent in Q1 2022 than it did in Q1 2021.

Similarly, Amazon’s advertising revenue is mostly immune from the restrictions of ATT, since it utilizes first-party data for targeting. And while Amazon’s advertising revenue growth slowed over the course of 2021 and 2022, it remained in the double digits, and growth increased in Q3 2022. Note that Amazon’s advertising platform is inextricably tied to eCommerce, given the nature of Amazon’s retail platform.

And finally, a company with no conceivable risk from ATT: Lamar Advertising, which operates outdoor advertising assets like billboards, transit displays, and airport advertising formats, among others, and has a market capitalization of nearly $10BN. The company’s revenue growth increased dramatically in Q2 2021, as would be expected, and growth has decelerated moderately since then.

There is a clear divergence of outcomes across the digital advertising and broader advertising marketplace related to post-COVID behavioral reversion and ATT. The shape of the growth curves for social media advertising platforms is distinctive, with a spike in Q2 2021 and a dramatic slope downward — in some cases, into negative growth. Outcomes for advertising platforms that are less impacted by ATT, or not impacted by ATT at all, are observably different.

Various media agencies, which are reliant in large part on social media platforms for delivering media spend to clients, have similarly faced divergent outcomes: for instance, Publicis and Omnicom, two of the largest advertising agencies in the world, have substantially different outlooks, with Omnicom’s growth mostly flat against 2021 whereas Publicis continues to grow:

It’s difficult if not impossible to tease the constituent revenue streams for these agencies apart to determine exposure to ATT, which is why I focused on companies with clearer and more distinct operating scopes. Surely individual examples of companies can be surfaced that defy the distinction that I’ve drawn here: companies that aren’t exposed to ATT that have seen growth collapse since peaking in Q2 2021. But I think one dynamic above all supports the notion that the broader advertising ecosystem is healthy and the ATT recession is confined primarily to the social media market: Everything is an Ad Network.

Everything is an Ad Network

I’ve written extensively on the concept that Everything is an Ad Network: I contend that ATT deprives Google and Meta of their jointly exclusive claim to growth in the digital advertising space, which has catalyzed an outbreak of retail media networks. From Why is Everything an Ad Network:

Social platforms can no longer ingest the totality of digital artifacts emitted by users as they traverse across apps and websites. In this new privacy landscape, the best place for some advertisers to deploy budget may be contextually-relevant retail or eCommerce properties that possess first-party, user-level purchase history data within the product categories most relevant to their own.

Indeed, the Duopoly — a term that had been used to describe Google and Meta, given their erstwhile combined ownership of almost all growth in the digital advertising landscape — now captures less than 50% of digital advertising revenues, according to a recent report from Insider Intelligence. Amazon has grown its advertising business materially over the past several years, as is evidenced by the chart above. And so has Apple — somewhat controversially, given that its ATT policy impaired the digital advertising incumbents.

And a diverse assortment of companies has recently launched wholly new advertising initiatives, some in just the past few months: Netflix and Disney+ both launched ad-supported product tiers at the end of last year, and Uber launched a mid-ride ad product. And while TikTok is certainly a competitive threat within the social media space, it appears to be grappling with the same challenges as Meta, YouTube, and Snap, with potentially negative revenue growth in the latter part of 2022, according to Morgan Stanley. The FT reported recently that TikTok revised its growth forecast down for 2022 by roughly 20%.

The rapid growth of retail media networks and the curtailment of the Duopoly’s share of digital advertising revenues is simply inconsistent with the notion that macroeconomic factors are broadly constricting advertising spend. While one might materialize, the belief that an advertising recession is currently and comprehensively depressing advertising spend is difficult to support with analytical rigor.

Edit, January 12th, 2023: This piece has been updated to reflect the latest CPI metrics released by the Bureau of Labor Statistics. The CPI decreased by 0.1% from November to December, and year-over-year price inflation stands at 6.5% in December vs. 7.1% in November.