When Apple’s App Tracking Transparency hit in 2021, iOS conversions on platforms like Meta dropped 70% almost overnight.
Performance marketers panicked, then adapted. According to AppsFlyer, mobile gaming has recovered to 122% of pre-ATT levels by embracing a discipline foreign to most brand teams: signals engineering.
Brand marketers haven’t made the same leap. They’re feeding billions of dollars into algorithms that deliver on proxy KPIs while offering little insight into whether they’re driving the metric that actually matters: incremental profit.
The algorithms aren’t broken. They’re working exactly as designed, but we’re giving them the wrong brief.
Incrementality: The Only Metric That Isn’t Vanity
From a CFO’s perspective, “views,” “completion rates,” and “clicks” are costs, not assets. The only thing that creates value is an incremental outcome, e.g. a purchase or shift in persuasion that wouldn’t have happened without the ad spend.
Traditional digital attribution models measure correlations rather than causal impact, meaning they often credit conversions that would have happened even without the ad exposure. If an algorithm finds 1,000 customers but 900 were already going to buy, you’ve wasted budget on 900 people to acquire 100.
By optimizing for proxies like cost per completed view, brand marketers ask platforms to fill funnels with cheap, non-incremental users—optimizing for efficiency at the expense of business value.
The industry knows this matters: according to the Association of National Advertisers, 71% of advertisers now rank incrementality as their most important KPI in retail media, for example.
The Blueprint: What Brands Can Learn from “Junk” Installs
Years ago, mobile games optimized for app installs. Algorithms found thousands of people willing to install a free app—but those “junk” users never played. Cost per install was low; business value was zero.
So marketers engineered better signals. They stopped sending “install” and fed back “user completed level 4, which is a deep-funnel event that correlated with revenue.
As independent analyst, Eric Seufert, has described, signal engineering is about creating deliberate tests for user intent and feeding those signals back into ad platforms so they can optimize in real time.
The algorithm stopped hunting cheap users and started finding valuable ones. When you change the signal, you change the results.
The Natural Evolution for Brand
Brand marketers couldn’t follow this blueprint because they lacked a real-time signal. You can track “level 4 completion” instantly; historically, you couldn’t track persuasion that fast.
But demand-side platforms (DSPs) have opened their APIs to receive upper-funnel signals, not just purchase events. Brand teams can now use the same pipes performance marketers built, just with a different payload a persuadability score derived from daily brand lift measurement instead of a purchase event.
You’re not tricking the system; you’re finally using it for its intended purpose.
Speed vs. Substance: A False Choice
Brand teams have been trapped by a false choice: either optimize in real time toward fast-but-shallow metrics like completed views, or wait weeks for meaningful brand measurement that arrives too late to influence anything.
This is the peril behind Les Binet’s warning that short-term metrics can distract from long-term growth. Many marketers have ignored the warning, and leaned into short-term signals without worrying if they’re pointing in the right direction. Others heed the warning and reject short-term signals altogether, and content themselves with improving “the next campaign.”
But the problem isn’t real-time optimization per se—it’s optimizing toward the wrong signal. A marketer might discover on Tuesday that Creative A outperforms Creative B, yet that insight sits in a PDF delivered weeks later, never informing the DSP’s bidding logic.
When the “number going up” reflects true persuasion rather than proxy behaviors, short-term signals become accelerants of long-term brand building. The work is ensuring those fast metrics are meaningful enough to predict long-term business value.
The Evidence
Meta reports advertisers making proper use of conversion API infrastructure see 8% CPA improvements on average. Liftoff saw impression-to-install rates improve 590% for gaming clients who engineered real-time signals. MIT Sloan research shows campaigns optimized with incrementality testing achieve 22% stronger budget efficiency.
These numbers come from performance campaigns, but the mechanism is universal. Awareness does translate into measurable business outcomes: advertisers that add upper-funnel discovery campaigns see roughly 35% more new customers on average.
The algorithm doesn’t distinguish between “game engagement” and “lift in awareness.” It only knows the signal you feed it.
The industry is splitting. On one side: marketers who accept vanity metrics. On the other: architects who know that feeding the machine the raw material of truth—actual brand impact—produces the only output that matters: long-term, incremental business growth.
