Google’s PMax, Meta’s Advantage+, and the logic of total advertising automation

Late last year, Google launched its Performance Max (referred to as PMax) campaign strategy out of beta, making it available to all advertisers. PMax is a down-funnel optimization tool: the advertiser supplies creative assets as well as campaign parameters like budget constraints, broad targeting guidance, and bidding targets, and Google automates the delivery of ads across its owned inventory in explicit service of the advertiser’s performance goals.

Note that PMax campaign conversion goals are currently limited to online sales, lead generation, and store visits; PMax doesn’t support app install conversion goals. This is because Google offers a spiritual counterpart to PMax for app install campaigns: Universal App Campaigns, or UAC, which consolidated all mobile app inventory under one automated campaign strategy back in 2017. As I wrote at the time:

The principal benefit of the UAC system is that it allows Google to do the heavy lifting of optimizing at the event level so that budget can be allocated to the combinations of creative / channel / targeting parameters that provide the highest levels of ROI. Given this, it’s also easy to understand why Google would make the decision to unify everything: many advertisers lack the infrastructure (and desire to build the infrastructure) that allows them to calculate granular LTV metrics, and they’d prefer to let Google handle that while they focus on product.

Similar to PMax, UAC campaigns accept assets for ad creative production, budget and bidding parameters, and broad audience targeting settings as inputs from advertisers and automate all aspects of budget allocation and delivery on behalf of the advertiser within the context of whatever conversion standard the advertiser specified. Google’s decision to bundle all of its mobile app install inventory under one campaign strategy, with control over budget allocation taken from the advertiser and tasked to Google’s automation machinery, was met with antipathy from app advertisers at the time. As I note in the above-linked piece, some advertisers, but especially larger advertisers that had built proprietary measurement and optimization infrastructure, bristled at the idea of all optimization agency being stripped of them and placed under the authority of an automated, “black box” platform.

But automation is the defiant trend of digital advertising, and especially app advertising: Meta introduced its App Event Optimization (AEO) and Value Optimization (VO) automated campaign strategies in 2016 and 2017, respectively, followed by Automated App Ads (AAA) in 2020. AAA was the logical evolution of AEO and VO, which automated audience discovery in pursuit of specific conversion goals: particular events such as purchases or cart actions with AEO and one-day and seven-day return-on-ad-spend (ROAS) targets with VO. It seemed inevitable that Meta (then, Facebook) would follow AEO and VO with an automated solution that resembled Google’s UAC and abstracted away most, if not all, optimization decision-making from the campaign management process through automation. As I wrote in 2019:

It’s obvious why the black box, totally automated ad platform paradigm is beneficial to platform owners: it virtually guarantees maximized profits as the platform strives to deliver minimally to a goal, and it also helps small advertisers get onboarded very seamlessly. And it seems almost inevitable that Facebook would evolve its platform in this direction — because where else can it go? The “Facebook UAC” is just a combination of dynamic creative, automatic placements, dynamic budget optimization, and AEO with a cost cap or VO with a min-ROAS setting. It’s fairly easy to make the leap from a campaign with these settings explicitly selected to merely “a campaign.”

I’ll underscore what I identify in the quoted text as the two primary benefits of these types of automated campaign strategies for advertising platforms:

  1. Spend maximization. By bundling inventory across placements and channels into one advertising product, the platform is able to allocate ad spend to destinations that the advertiser otherwise wouldn’t if given the choice, so long as the performance of the campaign meets the advertiser’s standards. Put another way: an automated campaign strategy might allocate budget to a placement that is not profitable for the advertiser in isolation so long as that loss is offset by excessive profits from another placement and the campaign’s performance target is met. See the chart above for a hypothetical example of a bundled, automated campaign that delivers 100% ROAS for the advertiser despite deploying more than half of its budget to a highly unprofitable channel. Bundling inventory and automating budget deployment in this way allows an ad platform to maximize the economic output of all of its inventory;
  2. Ease of use. While sophisticated advertisers might desire granular control of their budget allocation and creative utilization across placements and channels, administering every aspect of a campaign is out of reach for many if not most smaller advertisers. These automated, all-encompassing systems allow many advertisers to scale their ad spend beyond what they otherwise could, and they likely allow some non-trivial number of advertisers to spend their first dollar on digital advertising. So long as these systems are governed by performance objectives, many small advertisers likely are willing to accept that their ad spend is not being managed with the utmost efficiency.

The first presumed benefit outlined here is cynical; the second is functional. But another benefit exists, which is speed of campaign optimization: an algorithm can edit campaign settings and alter audience targeting faster than a human team can.

Meta recently expanded its Advantage+ automated campaign product to retail and ecommerce advertisers for web campaigns (per the documentation, Advantage+ is a re-branded version of Automated App Ads, which was previously only available to app advertisers). With Advantage+, retail and ecommerce advertisers can utilize the Advantage+ campaign strategy to automate various aspects of campaign creation, targeting, and creative pairing, with the added functionality that Advantage+ can route users to a retailer’s web shop or native Meta Shop (Content Fortress!), depending on which destination is predicted to deliver the most expected value.

Automation takes on added value for ad platforms, but especially for Meta and Google, in the new privacy environment. PMax and Advantage+ allow these platforms to optimize on-site ad engagement while being blind to the precise consequences of any given ad interaction: because the connection between ad engagement and off-site conversions (eg. purchases made on a website or in a mobile app) is broken, platforms can’t target individuals on the basis of behavioral history. And, at least with SKAdNetwork, Apple’s measurement framework for use within the restrictions of its App Tracking Transparency (ATT) privacy policy, the limitation of available campaign parameters dramatically constrains experimentation with ad creative.

So the optimization calculus changes in the new privacy setting, where very many combinations of ad creative and targeting parameters cannot be joined to off-site conversions such as purchases for measurement. Automated solutions like PMax and Advantage+ allow ad platforms to optimize dynamic ad creative to the engagements that can be observed, such as clicks, while ingesting either campaign-level conversion signals through privacy-safe frameworks like Meta’s Aggregated Event Measurement (AEM) or SKAdNetwork or through other methods (eg. Meta’s CAPI, Google’s conversion tags). These two data sets — on-site ad engagements and off-site conversions — remain disjointed but can be compared in aggregate to inform campaign spend using probabilistic modeling. An ad platform can optimize creative delivery, for instance, to minimize cost per click while modeling advertiser outcomes using campaign-level or anonymized conversions.

Meta struggled with conversion reporting following the introduction of ATT given the limitations of SKAdNetwork and AEM, but a few months ago, it revealed that it has reduced undercounting of iOS web conversions by about half — from 15% of conversions being lost to 8%. This was likely achieved through an expanded conversion modeling program, which is to say that Meta estimates conversions based on some combination of inputs:

  • Observed conversions from opt-in users;
  • Anonymized conversions transmitted through the Conversions API (CAPI);
  • Android conversions;
  • On-site conversions through native Meta Shops.

If this type of conversion modeling is the key to clawing back performance from the impairment of ATT and other privacy initiatives, then a shift into end-to-end campaign automation is the perfect vehicle for implementing it, given that campaign decisions are abstracted and placement- and creative-level reporting is obfuscated through bundling and dynamic ad units.

But there’s another advantage of automation: it incentivizes advertisers to concentrate budget across fewer ad platforms, but especially those that are filling owned-and-operated inventory. An advertiser would seek to minimize the number of channels in its traffic portfolio if some of them, but notably the largest, are modeling performance using anonymized conversion signals that can’t be directly attributed. Per this guest post: the more budget allocated per channel, the more genuine signal each channel receives from which to model (also: there are only so many organic conversions that can be unjustly claimed by overzealous conversion models). If the largest platforms are using end-to-end automation to drive immediate campaign reactivity based on modeled conversions, then it’s natural for an advertiser to allocate a disproportionate share of budget to those channels.