The Algorithm That’s
Eating Your Brand

Meta Advantage+ and Google Performance Max promised to make advertising smarter. What they actually delivered was a generation of marketers who can no longer think without a dashboard — and brands that function only while the drip is on.

The Sales Pitch

Meta Advantage+ and Google Performance Max are, on paper, genuinely impressive. You hand the platform a goal — get purchases, drive installs, maximise conversion value — and the system handles audience discovery, placement allocation, bid adjustments, creative testing, and budget distribution in real time. The pitch is elegant: stop micromanaging the auction and start thinking strategically.

And in the short term? It works. CPAs drop. ROAS numbers go up. The dashboard turns green. Quarterly reviews look good. Everyone is happy.

What Optimisation Optimises Away

Here is what most CMOs are not told during the onboarding call: these systems are not neutral tools that happen to obscure signal as a side effect. They are policy engines. And a policy engine that constantly reroutes spend away from weak pockets — broadening reach, adjusting bids, finding alternate inventory — does not preserve the evidence of why those weak pockets were weak.

Instead of learning that your offer is overpriced, your demand is seasonal, or your product-market fit is weaker than your slide deck implies — you learn that the system found another pocket that still converts. Useful for this week’s dashboard. Catastrophic for understanding your business.

The compounding sequence:

• Constant optimisation removes interpretable failure signals.
• Without failure signals, you stop learning what product-market combinations actually work.
• Without that learning, product development loses its feedback loop. You iterate blind.
• Marketing knowledge atrophies. Your team stops knowing how to read demand, what messaging resonates structurally, what channels build durable awareness.
• Brand and identity decisions get subordinated to whatever the algorithm currently rewards — which changes with every policy update.
• Leadership loses the ability to make independent strategic calls because their data infrastructure is platform-native and optimisation-smoothed.
• The business becomes an execution layer on top of Meta and Google’s logic, with no proprietary understanding of its own customers.

The CAC Trap Nobody Talks About

There is a specific failure mode that Advantage+ is structurally prone to, and almost nobody in the YouTube-marketer industrial complex mentions it: lowest-CAC optimisation and best-cohort optimisation are not the same objective.

If your setup effectively rewards cheap conversions, the system will drift toward cheap, low-quality buyers. It is doing exactly what you told it to do. The problem is that cheap buyers tend to have lower lifetime value, higher return rates, lower repeat purchase frequency, and higher support costs. The ROAS column looks fine. The cohort quality is deteriorating silently.

Meta has value optimisation tools that can partially address this — but only if you feed them accurate purchase value data, seed audiences from your top-decile buyers, and suppress known low-LTV cohorts in your CRM upstream. The machine cannot protect long-term value if you have taught it to count conversions like a lab rat hammering a pellet button.

Most accounts don’t do this. Most accounts run on purchase event counts and celebrate a falling CPA while their customer base quietly degrades in quality.

The Dependency Is Structural

It is worth being precise about what the platforms are actually incentivised to produce. Platform revenue is a function of spend, not advertiser profit. A system that maximises your ROAS is not the same as a system that maximises your business health — and the platforms know this.

The opacity of Performance Max and Advantage+ is not purely an engineering limitation. An advertiser who can measure true incrementality and LTV accurately will spend less — or at minimum allocate differently. The ideal platform customer is a business that is marginally profitable on paid, perpetually scaling CAC to chase growth, with no brand moat and no owned audience. That business will never leave because it cannot afford to stop. It will never get healthy enough to spend less.

What This Does to the Marketer

There’s a quieter form of damage that doesn’t show up in any performance report. The capabilities that make a marketer useful beyond the current campaign — the ones that transfer across platforms, survive algorithm changes, and compound into genuine strategic judgement — are exactly the capabilities that extended automation dependency tends to erode:

• Reading creative performance at a granular level and knowing why something resonated.
• Understanding audience behaviour structurally, not just as a targeting input.
• Making budget allocation decisions from first principles rather than algorithmic suggestion.
• Recognising demand signals the platform doesn’t surface because they don’t fit its reward function.
• Knowing when a channel is genuinely saturated versus when the bid strategy is misbehaving.

A marketer who has run on full automation for two years cannot diagnose a campaign from first principles anymore. The skill has atrophied. This is also why the YouTube-marketer class evangelises these systems so enthusiastically: automation provides a defensible alibi. If results are bad, the algorithm needed more time. If results are good, the setup was correct. Personal accountability for outcomes is quietly offloaded to the black box.

The Remedy

The logical split is to let the platforms own distribution while keeping ownership of understanding. Not vague scepticism about automation — a structural separation of what the algorithm decides from what the business learns.

That separation lives in the analytics layer. A GTM and GA4 implementation with custom behavioural events — scroll depth on product pages, engagement with conversion-relevant content, return visit patterns, session behaviour by traffic source — produces a map of customer intent that no ROAS column approximates. The platform reports what it optimised. The analytics layer reports what actually happened.

The funnel split that holds up under scrutiny:

1. Advantage+ and PMax handle top-of-funnel reach and audience discovery — the area where their scale is genuinely useful and the cost of opacity is lowest.
2. Everything after the click runs on infrastructure the platform has no visibility into: on-site CRO, mid-funnel behavioural analytics, session-level data segmented by source.
3. Post-purchase cohort tracking stays entirely outside the platform. Meta and Google become interchangeable traffic sources measured against the advertiser’s own standards.
4. Brand spend lives in a separate lane with its own budget. Sales-optimised campaigns, left unsupervised, harvest existing demand faster than new demand gets created — and attribution looks fine right up until the pipeline runs dry.

The Longer Game

Becoming proficient at using a platform and becoming a sharper analyst of your market are diverging skill sets. The platforms have an obvious interest in making the first feel like the second — and for a while, they look identical from the inside. The gap only becomes visible when the account hits a wall and nobody in the room can explain why.

The practitioners who tend to hold their value through platform changes and algorithm updates are the ones whose analytical framework doesn’t depend on any single platform’s reporting layer. That’s harder to build than a certification. It’s also considerably harder to commoditise.

The campaigns that look most efficient at any given moment and the campaigns that compound into durable business outcomes are often not the same campaigns. Advantage+ and Performance Max are very good at finding the former. Whether they’re building the latter is a question the platforms have no particular incentive to help you answer.