Black-Box Marketing

It’s a scenario that nearly every platform keynote is touting these days: The customer journey has become «predictably unpredictable»—a chaotic web of search queries, videos, shorts, and recommendations that no marketing team can manage manually anymore. So we need artificial intelligence to bring it all back together—automated, data-driven, and scalable. Google’s presentation at OMR is just one variation of a pattern that can be found in nearly identical form at Meta and TikTok. Google aggregates intent. Meta, the social graph. TikTok, the culture. And all three do so within a system that simultaneously controls the process and measures its own success. Even at the 2026 Cannes Lions, «black box» suddenly became a buzzword—the first providers there explicitly positioned themselves as a «glass box» alternative to the lack of transparency on the major platforms. For the pattern behind this, I propose a term of my own: black-box marketing.

Three «new items,» one sample

With «Performance Max,» the Google algorithm distributes the budget across Search, YouTube, Shopping, and Display, without the advertiser being able to determine exactly what the optimization is based on. At Meta, «Advantage+» has been the default setting for new campaigns since December 2025, rather than an option; Mark Zuckerberg described the vision as early as 2025 as a redefinition of the advertising category: A company simply specifies its goal and budget, links its bank account, and the machine takes care of the rest. On TikTok’s For You page, no one outside the company knows the algorithm that determines—based on dwell time, likes, shares, and even the brief hesitation before swiping away—whose message is displayed and how often. New to TikTok is the «Symphony Agent.» This is an intelligent, «agent-based» AI layer that is directly integrated into TikTok’s existing advertising ecosystem and «helps» marketers create campaigns in record time.

From an economic perspective, this is a classic «principal-agent problem» (Jensen/Meckling, 1976), exacerbated by information asymmetry, as described by George Akerlof (1970) for markets with unevenly distributed knowledge: The agent—the platform—has a knowledge advantage that the principal—the advertiser—can neither see nor independently verify. «Black-box marketing» thus refers to advertising whose management, optimization, and performance measurement take place within a system controlled by one party, which simultaneously acts as the “judge” of its own success.

Brand Vector vs. Platform Vector

A simple illustration can help explain what this concept means beyond individual campaigns. Every brand sends a message with a specific intention, a specific tone, and a specific target audience—let’s call this the «brand vector.» However, each platform doesn’t optimize for this vector, but rather for its own: dwell time, engagement, and ad revenue. That is its «platform vector.» The challenge is that these two vectors are rarely aligned. How far apart they diverge depends on the logic of the respective platform. The distortion is likely to be the least pronounced on Google Search: an explicit user intent meets an ad that responds precisely to it, because the business model itself is based on this alignment. However, the «Gemini layer» is leading users further and further away from the actual search results. Searching and finding content are becoming less transparent here as well. And YouTube, in particular, alters the picture even more significantly—a fact that can now be substantiated. A study by Webrepublic and GfK on Swiss YouTube usage (2026) shows that reach alone does not guarantee impact. What matters is whether users are actively and intentionally engaging with a piece of content or simply letting the platform run in the background for casual entertainment.

Here are the facts: On average, Swiss people spend 41 minutes on the platform on weekdays and 50 minutes on weekends, with the quality of their attention varying greatly. «»YouTube isn’t always YouTube,’ conclude the study’s authors—the same ad can be played in an attentive or a casual context. Which context it ends up in isn’t decided by the brand, but by the algorithm. On Instagram and TikTok, finally, the distortion is amplified: Assets are remixed, relabelled with new audio, removed from their original context, and distributed according to cultural rather than communicative logic. A campaign intended for a specific context can end up as an audio snippet alongside completely unrelated content, without the brand even knowing it.

"Black-box marketing," then, doesn't just mean not seeing the mechanism at work. It means not even knowing how much your own message differs from what actually gets across at the other end.

«Vendor Lock-in» as the Price of Passivity

Just enter the goal and budget—the machine does the rest. This convenience has a strategic downside that is rarely discussed: vendor lock-in. The more functions are outsourced to the black box, the more expensive it becomes to switch providers. This is not a side effect, but part of the business logic, and it doesn’t happen overnight—rather, it unfolds across four mutually reinforcing levels: A: At the data level, conversion signals, target audiences, and learning curves are stored within the platform and cannot be exported; what the AI has learned about a company’s own customers remains with the provider—anyone who switches has to start from scratch. B: At the level of expertise, the organization loses the ability to manage the process; as soon as the machine takes over the rest, in-house knowledge of campaign management, media planning, and creative work erodes until the organization simply lacks the ability to do things differently (or perhaps even better), even if it wanted to. This is the most dangerous level because it remains invisible. C: At the measurement level, success is gauged by metrics that the platform itself defines, collects, and reports—without an independent benchmark and without cross-provider comparison. D: Ultimately, even at the level of creation, assets are increasingly generated by AI and optimized directly for the algorithm that delivers them. As a result, brand communication becomes platform-driven, tailored to a system over which one has no control.

Together, these four levels result in at least an information asymmetry—if not a power asymmetry—that is structural rather than accidental. Meta itself provides evidence of this. In January 2024, the company removed several thousand detailed targeting options from Ads Manager without warning—categories on which many advertisers had built their audience strategies for years. Those who still wanted to manage their campaigns themselves were effectively pushed toward «Advantage+ Audience»—the automated system that has been the standard since December 2025 anyway.

The same applies to attribution: Meta has repeatedly shortened the standard time window for attributing success, most recently from 28 to 7 days after a click—overnight, identical campaigns began to look less effective, even though their actual impact hadn’t changed. The benchmark shifted, not reality. Anyone who builds their marketing on a third-party platform accepts that its owner also writes the rules by which their own success is measured.

What CMOs Should Do Now

None of this means we should demonize the platform. The efficiency gains from Performance Max and Advantage+ are real, and no marketing strategy today can ignore this reach. Something else has shifted: the question of what really matters. The following consequences can be directly inferred: Independent measurement: When the algorithm both controls the campaign and measures its success, optimization is based on a reality defined by the platform itself. Understanding impact in one’s own market requires benchmarks that go beyond the platform’s own metrics. Transparency as a value in its own right: The relevant demand on platforms is not «less AI,» but transparent rules—and the willingness to evaluate providers based on them, not just on reach. Sovereignty instead of dependence: Keep first-party data in-house, retain internal control, and practice multi-homing instead of monogamy. Find the «Big Idea» (again)!: · Develop your own clear idea against which the platforms« »math” must be measured—instead of leaving it up to them from the start.

These reflections are part of a broader examination of what I call the fragmentation paradox: platforms that market as a solution something they themselves have helped create. Black-box marketing is not a marginal phenomenon here, but rather the core of a business model. A black box remains a black box, no matter how efficiently it calculates.

The author

Prof. Dr. Marcus Schögel, Institute for Marketing and Customer Insight at the University of St. Gallen (HSG), President of the Swiss Marketing Association (gfm).