Your campaigns are optimized. But for whose goals?

How Google and Meta are quietly reducing advertiser control - and why it matters.

Sitting across from an ad manager. Campaign results in front of us. Leads coming in. No deals closed. I suggest a change in configuration.

He looks at me calmly and says:

"But this is how we were taught on the training."

He is not lying. He was genuinely taught this way.

Platform courses attempt to address client business goals. In theory they teach optimization toward conversion, customer acquisition, revenue. In practice they teach how to operate the panel according to platform recommendations - recommendations that frequently reflect the platform's own business interests as much as the advertiser's. These two things are not always the same.

This conversation has happened to me more than once. I found myself explaining why a configuration that follows platform recommendations may not serve the economic goals of a specific business. Explaining LTV, margin, the difference between a lead and a closed deal. Using logic that platforms try to address through their solutions - but which in this particular case required a different approach.

The ad manager eventually changed the configuration.

The platform's recommended setup was working perfectly. Just not for the client.

A panel that keeps changing - not always in the advertiser's favor

I have been watching the Meta panel for over a decade. The changes I observe move consistently in one direction: less precise targeting on the advertiser's side, more decisions on the algorithm's side.

This is not purely subjective. The changes are documented.

• In July 2024, Meta began phasing out detailed targeting exclusions - the ability to exclude specific audience groups from campaigns. From March 2025 this option ceased to exist entirely. Campaigns using it stopped delivering if the advertiser did not update their settings.

  
  

• In June 2025, Meta consolidated many detailed interest categories into broader groups. Specific interests like "EDM fans", "SUVs" and "vegan food" were merged into more general categories. The targeting precision that advertisers had built over years disappeared in a single platform update.
  

• In February 2025, Meta removed the Existing Customer Budget Cap - a feature that allowed advertisers to control what portion of Advantage+ Shopping budget went to existing customers versus new ones. The feature disappeared without announcement.
  

• Jon Loomer, one of the longest-standing Meta Ads analysts documenting platform changes for over a decade, described 2025 as follows: "One of the important trends we have seen continue from 2024 to 2025 is more automation and less advertiser control."
  

• Meta's official communication from March 2025 states that the platform is shifting emphasis from targeting to creativity - the advertiser provides diverse creatives, the algorithm decides who sees them.
  

This can be interpreted as technological progress. One can also ask who benefits from this shift - and whether the answer is always the same.

Performance Max: four years without visibility

In November 2021, Google introduced Performance Max. One campaign covering all channels - Search, YouTube, Display, Gmail, Discovery, Maps - with an algorithm deciding budget allocation.

• For four years, advertisers had no visibility into how budget was distributed across channels. Channel reporting arrived only in May 2025 - after years of industry pressure. The data is available for reading. Budget allocation between channels still cannot be directly controlled by the advertiser.
  

• A study by Optmyzr analyzing 503 accounts found that 91.45% of accounts had keyword overlap between Search and PMax campaigns. In scenarios where both campaigns competed for the same queries, Search campaigns outperformed on conversion efficiency almost twice as often. The conclusion: a significant portion of conversions reported by PMax may not be incremental - they are conversions that Search campaigns would have captured anyway.
  

• Kirk Williams, founder of ZATO Marketing and one of the world's leading Google Shopping experts, said publicly in 2025: "Has PMax jumped the shark? IMO, it has. We have never seen PMax perform consistently worse than it has moving into 2025, with little to no strategy shifts on our end. We have begun transitioning to Search and Standard Shopping more purposefully with obvious revenue and efficiency increase."
  

• I had my own case. I took over a Google Ads account for a real estate client after two previous agencies. Performance Max was running. Leads were coming in - the dashboard showed green. No closed deals.
  

• Switching to pure Search revealed a second problem. Google's broad match was pulling in queries far beyond the client's core intent. Related but commercially useless search terms. Weeks of manually reviewing actual search queries and blocking unwanted phrases - one by one.
  

This is the reality of "control" in modern automated campaigns. The campaign is yours - but making it work according to your intent requires weeks of filtering out what the algorithm keeps adding on its own.

ROAS versus MER - why they are not the same and why the difference is significant

This is the point where most conversations with ad managers stop. I will explain it very concretely, with numbers.

How ROAS works

ROAS - Return on Ad Spend - is the revenue attributed to a campaign by the platform, divided by the campaign cost. If you spent 10,000 USD on Meta Ads and Meta reports that your ads generated 40,000 USD in revenue, your ROAS is 4x.

The structural problem is this: Meta measures the effectiveness of Meta ads using Meta data. Google measures the effectiveness of Google ads using Google data. Each platform is simultaneously a media seller and the provider of measurement for its own results.

In practice it looks like this. You spend 10,000 USD on Meta and 10,000 USD on Google - 20,000 USD total. Meta reports ROAS 4x and attributes 40,000 USD in revenue to itself. Google reports ROAS 3x and attributes 30,000 USD in revenue to itself. Total reported by platforms: 70,000 USD.

Actual company revenue that month: 38,000 USD.

Both platforms attributed parts of the same transactions to themselves. Neither technically lied - each looked through its own attribution lens. But together they gave you a picture that does not exist.

This is not a hypothetical scenario. According to eMarketer research from 2026, only 27.6% of marketers consider their measurement methodologies reliable. Marketing directors from Nestle, Haleon and Molson Coors acknowledged publicly at the Possible conference in Miami in April 2026 that measuring incrementality remains an unsolved problem - despite teams of analysts and budgets most companies cannot imagine.

How MER works

MER - Marketing Efficiency Ratio - is total company revenue divided by total marketing spend. Calculated not from a platform dashboard but from your own financial data.

In the same example: 38,000 USD revenue divided by 20,000 USD spend = MER 1.9x.

Not 4x. Not 3x. 1.9x.

This is a number no platform can falsify - because it comes from your bank account, not from a pixel. No attribution algorithm touches it. No conversion window changes it. MER is a return to basics. Something every CMO should have on their dashboard at all times. The formula is deceptively simple - yet in companies with complex spending structures, multiple channels and CMOs raised on ROAS as the primary truth, it is far from universally applied. The more sophisticated the marketing stack, the easier it is to lose sight of the one number that actually tells you whether the whole system is working.

Why this difference has a cost

Consider two scenarios for the same company over 12 months.

• Scenario A: the company manages budget based on platform ROAS. Meta and Google report positively - budgets grow. Revenue grows moderately because some conversions are the same transactions counted by both platforms. The company believes it is scaling effectively.
  

• Scenario B: the company manages budget based on MER. In a month when MER drops from 2.1x to 1.6x - despite platforms reporting green ROAS - the company pauses spend growth and looks for the cause. It finds it: a PMax configuration change that began cannibalizing branded search. It fixes it. MER recovers.
  

In Scenario A, the company burned budget for a quarter without knowing it. In Scenario B, the problem was visible in the first month.

MER will not tell you which specific campaign is working. It will tell you whether your marketing system as a whole is generating value. That is a different - and often more important - piece of information.

What happens to traffic quality

Automation and reduced targeting control have a direct impact on the quality of traffic you are buying.

According to the Pixalate Q1 2025 benchmark report, 22% of clicks from mobile applications in ad networks are invalid traffic. In high-CPC industries like real estate, legal services and finance, this rate reaches 30%. Global losses from ad fraud reached 84 billion USD in 2025 according to Juniper Research.

Display Network and Performance Max campaigns without proper exclusions regularly reach children's apps - toddler games, educational applications, drawing programs. Users click ads accidentally while trying to close a window or navigate forward. Budget drops, clicks grow, conversions do not exist. Kirk Williams at ZATO has been updating his guide on excluding mobile apps from PMax since 2017. The problem has existed for a decade without being structurally resolved.

In 2023, analytics firm Adalytics published a report based on analysis of campaigns from over 1,100 brands covering billions of impressions between 2020 and 2023. The Wall Street Journal reported the findings: approximately 80% of Google video placements on third-party sites in the Google Video Partners program violated the platform's own standards. Ads were muted, auto-played and displayed in small players outside the main page content - rather than as skippable in-stream video with audio as the purchase terms defined. Brands whose campaigns were covered by the research included Samsung, Disney+ and Johnson and Johnson. Joshua Lowcock, Global Chief Media Officer at agency UM Worldwide, told the WSJ: "This is an unacceptable breach of trust by YouTube. Google must fix this and fully refund clients for any fraud and impressions that failed to meet Google's own policies." Google disputed the report's methodology and denied the violations.

Regardless of the methodological dispute between Adalytics and Google, the existence of this discussion at this scale illustrates the difficulty of independently verifying where advertising budgets actually go.

Three levels of interpretation

I am not here to deliver a verdict. I want to present three possible interpretations of what I have described - and leave the choice to you.

• The first level. Automation genuinely outperforms human decision-making in optimization at sufficient data scale. Platforms reduce manual control because algorithms achieve better results with more freedom. Losing precise control is the price of higher system efficiency. This is the official platform narrative and many practitioners share it.
  

• The second level. Every platform is simultaneously a media seller and the primary source of measurement for its own effectiveness. Its revenue grows when your spend grows. Automation that "optimizes for conversion" while also expanding reach serves both goals at the same time. This is not an accusation of bad intent. It is an observation that the incentive structure of platforms and the incentive structure of advertisers are not identical.
  

• The third level - a question I leave open. The history of advertising platform relationships with advertisers includes documented cases where standards promised in agreements were not met in practice. These cases were disputed by the platforms and resolved in various ways. Automation and reduced visibility make independent verification more difficult. In this context, the question of whether "AI optimizes in your interest" is a legitimate question - not as an accusation, but as an element of due diligence for any serious advertiser.

What to do about it

I am not writing this to tell you to stop using platforms. I am writing this so you use them as a vendor with its own business model - not as a neutral partner without interests of its own.

Three concrete changes.

• First: measure MER as the primary metric for your marketing system's health. Total revenue divided by total spend. Calculated from your financials, not from a dashboard. If MER falls while platforms report green ROAS - something is wrong and it is worth finding before you spend more.
  

• Second: keep manual Search campaigns as a reference point. This is the channel where you have the most visibility into user intent and can see exactly what you are spending on. PMax and Advantage+ may work - but you need a clean comparison to evaluate whether they are actually working for your specific business.
  

• Third: find someone who evaluates campaign effectiveness in terms of the client's business economics - not in terms of platform metrics. Someone who asks about LTV, margin and closed deals rather than just leads. Someone who can identify when a platform recommendation conflicts with the client's interest.
  

I come back to the conversation with the ad manager from the beginning of this article.

He changed the configuration. Against the platform's recommendation. Results improved.

Platforms genuinely try to address their clients' business goals. But the platform's goal and your business's goal are not always the same thing. Knowing when they diverge - that is a skill no platform course teaches explicitly.

Robert Sosnowski is the founder of Growth Architects Studio. He has worked with 200+ brands across Europe and the US over 22 years in brand and growth strategy.

Sources and further reading

On Meta targeting changes Jon Loomer Digital - 83 Meta Advertising Changes 2025 https://www.jonloomer.com/meta-advertising-changes-2025/

On Performance Max Optmyzr / smarter-ecommerce - State of Performance Max 2025 https://smarter-ecommerce.com/blog/en/google-ads/state-of-performance-max-campaigns-2025

On Google video ad placements Adalytics - Google TrueView Research Report 2023 https://adalytics.io/blog/invalid-google-video-partner-trueview-ads

On ad fraud and invalid traffic Pixalate - Click Fraud Benchmark Reports https://www.pixalate.com/blog/q3-2024-click-fraud-benchmarks-reports

On measurement and incrementality eMarketer - FAQ on Incrementality 2026 https://www.emarketer.com

Adweek - Brands Like Nestle, Haleon and Molson Coors Wrestle With Broken Measurement, April 2026 https://www.adweek.com/brand-marketing/brands-like-nestle-haleon-and-molson-coors-wrestle-with-broken-measurement/ Wall Street Journal - Google Violated Its Standards in Ad Deals, June 2023 https://www.wsj.com/articles/google-youtube-ads-media-buyers-11687960409  Juniper Research - Ad Fraud https://www.juniperresearch.com  Kirk Williams / ZATO Marketing - How to exclude mobile apps from PMax https://zatomarketing.com/blog/exclude-mobile-apps-google-adwords-easy-way