CMOs today have more performance data than any generation of marketing leaders before them, and yet it is still one of the shortest tenures in the C-suite. More of them fell short on acquisition, retention, and ROI goals than a year ago, on budgets that have stayed flat since 2022.
Many reasons for the short tenure have been suggested over the years. A poorly defined role, an inherent conflict between performance and strategy, or unrealistic expectations and ill-conceived metrics have all been debated. There is a consensus among industry commentators that CMOs are given broad and rapidly changing expectations, but their authority, resources and success metrics often do not match those expectations.
Another explanation is worth putting forward though, along with a possible remedy.
No One Knows What Works
A major claim in evolutionary science, known as the social-intelligence hypothesis, argues that human intelligence evolved first and foremost to understand and master group communication. Marketers are tasked with by far the most complicated task of their organization – understanding people, at scale.
A CMO is asked which message will land, which creative deserves the budget, which campaign will hit the quarter’s budget, and is expected to answer before the money is spent. Performance marketing is supposed to make this measurable, but measurement is retrospective. It records what happened but cannot say what will.
The evidence behind human foresight is not flattering. Back in 2022, in a study of more than 750 marketing professionals, participants correctly picked the winning ad creative 52% of the time, the equivalent of a coin flip. And the payoff distribution is unforgiving. Most of the emails, ads and variations a team creates will do close to nothing, while a small handful carries the entire result, and nobody knows which is which until the results come in.
This is almost impossible to solve in isolation.
Consider a promotional email. It consists of a subject line, a preview text, a hero image, a layout, a CTA verb, and a send time. The space of possibilities is infinite. No matter what a team produces, it will almost certainly be far from the best version available, even for a single segment. This problem is compounded even more by the chaotic nature of trends and a market that never holds still.
Nonetheless, companies still expect marketers to solve an unsolvable problem. No wonder that 58.1% of marketers feel overwhelmed, 56.1% undervalued and 50.8% emotionally exhausted.
The Data-Creative Disconnect
Creative is the most important lever in performance marketing. It drives about 49% of the sales lift, more than targeting, reach and recency put together. In most organizations, two divisions control that lever, and different metrics reward each of them.
Performance is measured on weekly data, such as Customer Acquisition Cost (CAC), Return on Ad Spend (ROAS), and conversion. Brand is measured on consistency, affinity, and recall, over many years. Both sets of metrics are important for any business, but having two pilots controlling the same plane creates innate conflict, and is a common source of friction in many organizations.
Nearly half of organizations operate separate brand and performance teams and 65% keep separate budgets for them, while only 44% say the two sides share a common language for the audiences that drive growth. Every performance marketer knows the feeling of waiting— for assets, for approvals, or for a creative cycle that moves in weeks while the channel moves in hours. Every brand creative knows the other side of it: a test comes back, B beat A by 4%, and the number lands as a verdict that cannot say which part worked, or what to make next.
Somewhere in that infinite space of options there is content that serves both by being on-brand and high-converting at once. It exists, and the problem for CMOs has always been finding it.
The New Marketing Flow
For the first time, creative is not the bottleneck— thanks to generative AI.
Creative was expensive, so teams could only make a few things. Generative AI removes that constraint, leaving only the question of what to produce.
Generative AI uses statistical models to generate content. Using the right tools, marketers can incorporate their own performance statistics into the generation process and allow the model to explore what works on its own, while keeping brand intact. This process is known as reinforcement learning and was in use for many years in the development of frontier models such as the LLMs we use today.
Connected to a company’s performance data, reinforcement learning algorithms can guide generative models to explore intents and messages on their own, balancing exploration, trying options whose value is still unknown, against exploitation, generating options that have worked before. This approach is the scalable evolution of A/B testing. Given the right brand guardrails, teams can constantly test new angles and executions, learning from every campaign.
Putting reinforcement learning into production takes a new type of marketing flow – briefing a model instead of a designer, applying human judgment on the results, and sending multiple variations to be tested. Once results arrive, they’re fed automatically into the same model that generated them, and this closes the feedback loop, enabling each message to learn. This new flow does not expect marketers to know what works in advance, it expects them to guide the system to optimal, continuous testing.
Extending the Tenure
The change runs deeper than tooling. While the old job asked marketers to know what would work, the new one asks them to keep this machine running at full pace, constantly balancing exploration and exploitation.
That is a more honest job than the one CMOs have been given thus far. Nobody has to forecast the unforecastable. The work becomes setting the guardrails, judging what the system surfaces, and getting results back quickly enough to act on them. The performance data for this already sits in every marketing stack, unread by the systems that produce the creative. The two have simply never been wired together. Creative drives half the sales lift in the first place, so improving the odds on it, campaign after campaign, compounds into a number no CFO can ignore.
For the CMO, that is the line between a cost center and a defensible growth function. A leader running a system that learns from every campaign can account for why the winners won and can point to the process that will find the next ones.
CMOs have spent years being judged on a question nobody can answer in advance. A leader who demonstrates the system that answers it continuously is no longer the easiest person in the room to let go.
Nir Weingarten is the Co-Founder and CEO of Eikona, a start-up using GenAI and reinforcement learning to transform lifecycle marketing. A published AI researcher with a master’s degree in machine learning from Reichman University, he has spent over a decade leading multi-disciplinary technical and product teams across the fields of AI, data and performance.
