Something fundamental changed in what drives growth, and most brands haven't caught up.
For nearly a decade, the performance playbook was straightforward. You bought cheap attention on platforms that offered extreme targeting precision. You chose your audiences — lookalikes, interests, custom lists. You measured everything with cookies and last-click attribution.
When performance dipped, you refreshed creative, restructured campaigns, or hired a better agency. The levers were clear, and they lived inside the ad platforms.
Well, that playbook is dead!
The brands still operating on it aren't collapsing overnight, but experiencing something far more dangerous: a slow deterioration in unit economics that dashboards don't fully explain.
The shift nobody talks about enough
Meta, Google, and TikTok have moved to fully algorithmic buying. Advantage+, Performance Max, Smart Bidding. The names differ, but the principle remains the same:
You don’t choose the audience, the algorithm does.
All in a few milliseconds, based on the signals you send it.
This is not an optimization. It's an inversion. Platforms used to be tools you operated. Now they're machines you feed. What you feed them determines your CPMs, your prospecting efficiency, the quality of customers you acquire, and whether you can scale without watching unit economics collapse.
Most growth teams understand this intellectually. Very few have internalized and implemented what it means operationally.
Because if signal quality is the lever, then media performance is no longer primarily a media buying problem. It's a data infrastructure problem. And that changes where your attention should be.
Meanwhile, the inputs are degrading from every direction
At the very moment platforms began demanding better inputs, those inputs got worse than ever.
- iOS introduced opt-outs
- Privacy legislation expanded state by state and country by country.
- Browsers restricted cookies. Ad blockers grew.
Brands lost visibility into the anonymous acquisition journey — the part of the funnel where ad dollars do most of their work and where algorithms need the strongest feedback.
The industry's response has largely been patchwork. Server-side tagging moved data off the browser, which was necessary, but it's still plumbing. It forwards what your commerce platform sends. It doesn't shape what the algorithm learns from. Traditional CDPs unify known customers well after an email is captured, but the anonymous, pre-identification journey is where many of them go dark — because their identity layer was built on cookies.
So brands are feeding degraded signals into increasingly autonomous systems, then asking media teams to fix the outcome.
The gap between what dashboards show and what's actually happening
This is where it gets painful. Because dashboards often look fine. ROAS is stable. Revenue is growing. Nothing appears obviously broken. But underneath, predictable distortions are compounding.
The ROAS illusion
Your "prospecting" campaigns serve ads to existing customers — not because you misconfigured them, but because without durable identity, exclusion lists match far less than you think.
The algorithm fills the gap with the easiest conversions available, i.e. people who already know you. Blended numbers look healthy while true new customer acquisition costs quietly run 20–30% higher than reported. Prospecting is partly retargeting in disguise, and your dashboards can't tell the difference.
Attribution bias
Channels that close demand take credit for conversions that other channels initiated. Last-click logic structurally over-rewards the bottom of the funnel and under-rewards discovery. Without signal-level visibility into incrementality, you optimize toward what looks efficient in-platform, not what actually drives growth.
Short-term ROAS improves. Long-term demand weakens. And budget flows to the wrong places for the wrong reasons.
Value blindness
If you sell across categories or price points, most algorithms care only about one thing: "conversion." A low-value accessory and a high-margin hero product look identical unless your event structure tells the system otherwise. So optimization drifts toward the easiest purchases. Revenue can grow while margin quietly erodes — and the algorithm has no idea it's making the wrong tradeoff because nobody gave it the information to know better.
None of this shows up cleanly in standard reporting. It shows up later.
And that’s when the finance dept. asks why is CAC rising, LTV cohorts softening, & how come revenue increased but profit didn't. Marketing points to the creative and Finance to the channel mix but the real constraint lies in infrastructure no one explicitly owns.
What this means for the next phase of growth
The brands pulling ahead have recognized that growth moved down the stack. They started treating data infrastructure not as a back-office concern but as a primary growth lever.
Not in the abstract "we need better data" sense that every brand pays lip service to. In a specific, operational sense, they:
- Control what signals each platform sees.
- Build an identity that persists beyond a cookie window.
- Suppress existing customers at the signal layer rather than relying on lossy lists.
- Filter junk traffic before it reaches the algorithm.
- Structure conversion events so platforms can distinguish between customer types and product value.
- Treat consent as a signal strategy, not just a compliance requirement.
- Manage how signals are routed and prioritized across platforms.
The result is that the algorithm works harder and helps acquire an even higher share of truly incremental customers. Campaigns optimize toward the right products, not just the easiest ones. Economics remain more stable as they scale because the system is learning from cleaner, more representative data.
This creates a compounding dynamic. Cleaner signals lead to better optimization, leading to better customers, which generates even better data, enabling cleaner signals. That flywheel does not start with media buying. All of it starts with infrastructure!
The uncomfortable question
The real question is not whether you should try out more creative, hire a new agency, or restructure campaigns. Those things matter, but they are second-order levers as of today.
The first-order questions are more uncomfortable:
What does the algorithm actually see?
What identity layer is your measurement built on?
How much of the anonymous journey is visible?
What happens to your signals when users switch devices, browsers, or opt out?
If you can't answer those with specificity, you're optimizing a machine you're not properly feeding. The gap between brands with strong signal infrastructure and those without will keep widening, as platforms become more algorithmic everyday.
The game didn't get harder but just moved layers.
The real question you need to ask yourself is whether your infrastructure is built for how brands grow today, or how they did three years ago?
