Amazon unBoxed 2024 is getting the trade press this week, but the more instructive earnings story is Meta’s Q3 results, which were published alongside the broader adtech conference circuit. Meta reported strong revenue growth, and the company’s narrative around ATT recovery has shifted from defensive to confident: the signal loss from Apple’s App Tracking Transparency framework is, for practical advertising purposes, largely recovered.

Understanding how Meta rebuilt what ATT destroyed — and what advertisers who haven’t fully implemented the rebuild are leaving on the table — is one of the more important technical and strategic stories in performance advertising right now.

What ATT Actually Broke

Apple’s App Tracking Transparency framework, which became mandatory with iOS 14.5 in April 2021, required apps to ask users for permission to track them across other apps and websites. The opt-in rate for ATT has consistently been around 25-30% — meaning that roughly 70% of iOS users are not trackable across apps under the traditional IDFA-based measurement architecture.

For Meta specifically, ATT was devastating in the immediate aftermath. The company reported a $10 billion revenue impact in 2022, attributed primarily to the inability to measure campaign performance, attribute conversions, and build lookalike audiences at the scale and fidelity that had previously driven its advertising efficiency.

The damage wasn’t that Meta couldn’t serve ads — it could. The damage was that Meta couldn’t prove its ads worked. When Facebook can’t close the loop between an ad impression and a purchase, performance advertisers lose confidence in spend. When advertisers lose confidence, they reduce budgets. The revenue impact followed directly from measurement degradation.

The Three-Layer Rebuild

Meta’s recovery from ATT has happened through a combination of three technical systems working in coordination: Aggregated Event Measurement, the Conversions API, and Advantage+ automated campaigns.

Aggregated Event Measurement (AEM) is Meta’s on-device attribution framework for iOS, operating within Apple’s SKAdNetwork architecture. Rather than tracking individual users across apps, AEM uses aggregated, privacy-preserving conversion signals that Apple permits. The aggregation means some signal precision is lost — Meta can see aggregate conversion volume but not individual user conversion paths. The framework has matured significantly since its 2021 introduction, and Meta’s modeled attribution built on top of AEM has improved consistently.

The Conversions API (CAPI) is a server-side event reporting system that sends conversion data directly from an advertiser’s server to Meta’s systems, bypassing the browser and mobile OS layers that ATT affects. When a customer completes a purchase on a website, the event — enriched with hashed customer identifiers like email and phone — is sent from the advertiser’s backend to Meta’s API rather than through a browser pixel. CAPI operates independently of ATT, cookie restrictions, and browser privacy settings.

CAPI is the single most important under-implemented Meta advertising tool in the industry. Advertisers with full CAPI implementation — covering web purchases, form submissions, app events, and offline conversion uploads — can recover 20-40% of signal that ATT and pixel blocking would otherwise eliminate. Advertisers without CAPI are running on degraded signal that produces inefficient attribution and suboptimal bidding.

Advantage+ Shopping Campaigns are Meta’s automated campaign architecture, launched in 2022 and significantly expanded since. Advantage+ removes most manual campaign controls — audience targeting, ad set structure, budget allocation across creatives — and replaces them with Meta’s AI optimization. The AI works on both on-platform behavioral signals and the server-side conversion signals from CAPI, finding patterns that human campaign managers can’t access because they operate in privacy-preserving aggregated spaces.

What the Q3 Numbers Show About On-Device Modeling

Meta’s Q3 revenue growth demonstrates that its on-device modeling infrastructure has reached the point where advertiser returns are sufficient to sustain and grow budgets. The company’s description of its technical recovery centers on building AI systems that can model audience quality and conversion likelihood from on-platform engagement signals alone — watch time, engagement depth, profile characteristics — without requiring cross-app tracking.

The implication is that Meta’s advertising system now works primarily from its own first-party signals, supplemented by advertiser-provided CAPI data, rather than from the IDFA-linked cross-app profiles that ATT disrupted. That’s a fundamentally different architecture than pre-ATT Meta, and it’s one that’s more resilient to future platform privacy changes.

For advertisers, the practical consequence is that Meta’s Advantage+ products work best when fed with clean, complete conversion signals from CAPI. The AI optimization is only as good as the conversion data it trains on. Advertisers still using pixel-only measurement, without CAPI augmentation, are handing the optimization system degraded inputs and then wondering why performance is inconsistent.

Amazon unBoxed 2024: The Other Side of the Q3 Story

Amazon’s unBoxed conference this week announced several retail media and measurement products that deserve attention alongside Meta’s earnings.

Amazon DSP’s expanded measurement capabilities — including Amazon Marketing Cloud’s incrementality testing features and enhanced audience builder tools — reflect the company’s push to position itself as a full-funnel advertising platform rather than a bottom-funnel conversion machine. The additions are meaningful but incremental rather than architectural.

The more interesting Amazon announcement from a competitive dynamics standpoint is the expansion of Amazon Publisher Services to non-Amazon properties at scale. Amazon is growing its ability to monetize publisher inventory beyond its own ecosystem, bringing its first-party shopper data to the open programmatic market. This positions Amazon as a direct competitor to data enrichment companies like LiveRamp and Oracle’s late advertising segment — which is relevant context for the open programmatic market’s trajectory.

What Advertisers Are Leaving on the Table

The CAPI implementation gap in the advertiser market is larger than Meta’s public messaging suggests. Independent audits of advertiser Meta account setups routinely find that 40-60% of advertisers lack complete CAPI implementation — missing key event types, sending duplicate events without deduplication logic, or failing to enrich server events with customer identifiers that would allow Meta to match them to user profiles.

The consequences are specific and measurable. Incomplete event matching means Meta’s bidding system is optimizing toward conversions it can partially observe rather than conversions it can fully measure. Duplicate events without deduplication create artificially inflated conversion counts that throw off bidding strategy. Missing customer identifiers (hashed email, phone) on server events reduce Meta’s ability to tie server-sent conversions to specific user accounts.

A competent Meta CAPI audit — reviewing event match quality scores in Events Manager, deduplication configuration, and conversion signal coverage across transaction types — is a higher-return investment than most advertiser teams realize. Meta’s Events Manager provides an event match quality score that identifies specific signal gaps. Teams that haven’t reviewed this in the past six months should.

FAQ

Q: Is the Conversions API only for e-commerce, or does it apply to lead generation and B2B campaigns as well? CAPI applies to any conversion event an advertiser wants to report to Meta. For lead generation, that means sending form submission events, qualified lead status, and pipeline stage data from the advertiser’s CRM directly to Meta’s API. For B2B, uploading closed deal data and enriched lead quality signals from Salesforce or HubSpot to CAPI is one of the highest-leverage optimizations available.

Q: If Meta has rebuilt signal through on-device modeling, why does CAPI still matter? Meta’s on-platform modeling is trained on Meta’s own behavioral signals. CAPI adds conversion truth — actual purchase events, revenue values, lead quality — that Meta’s behavioral signals can only approximate. Training Advantage+ AI on actual business outcomes rather than modeled proxies consistently produces better optimization. CAPI is the mechanism for feeding business outcome data directly to Meta’s systems.

Q: How does Advantage+ Shopping compare to standard ad set-based campaign structures for performance advertisers? For advertisers with sufficient conversion volume (roughly 50+ purchases per week as a practical minimum), Advantage+ Shopping consistently outperforms manually structured campaigns in head-to-head tests, particularly when combined with complete CAPI implementation. For lower-volume advertisers, the AI optimization needs sufficient data to learn effectively and may underperform manual structures in the short term.

Q: What should advertisers do if their Meta performance degraded after iOS 14.5 and hasn’t recovered? The most likely cause is incomplete CAPI implementation combined with continued reliance on manual audience targeting rather than Advantage+ optimization. The diagnostic path: check Event Match Quality scores in Events Manager, confirm server events are deduplicating correctly, audit which conversion event types are being sent server-side vs. pixel-only, and run an A/B test of Advantage+ Shopping against the existing campaign structure with full CAPI enabled.