The Marketing Function That's Going to Look Boring - and Win

I want to show you two Monday mornings.

Both are in the same month, the same industry, the same scale of company. Two consumer brands with roughly equal revenue, comparable headcount, comparable ad budgets, comparable AI ambition. One of them began the migration to a calibrated measurement substrate in 2024. The other did not.

What follows is what their marketing functions look like on a routine Monday in 2027.

I have spent the last four issues of this letter describing what is broken in marketing measurement, who the new buyer is, which metric can be trusted, and why the board is now governing the answer. This is the issue where I describe what the destination actually looks like once a company has done the work.

The destination is not what most marketers expect. It is the most counterintuitive part of the whole story. So I want to lead with the pictures, and unpack the principle afterwards.

At 8:17am the Slack notification arrives. Three lines.

Overnight: two reallocations executed within guardrails. One flagged. Open for review.

The CMO opens the flag. The system shows the experiment that earned the recommendation, the holdout result that calibrated it, the projected P&L impact of the proposed move, and the kill criteria she'd want to monitor over the next four weeks. She reads the methodology footnote. She approves the reallocation.

Total time: ninety seconds.

She gets coffee.

At 9:30am she and the CFO open the same dashboard. There is one item on it that neither of them has been able to bound confidently - a brand campaign whose causal impact the most recent geo-lift could not narrow inside an actionable interval. They discuss whether to run a sharper test against a different geo split. They agree on the test design. The CFO returns to his quarterly close.

At 11am the CMO writes the board memo. Three numbers, each defended by a methodology disclosure footnote. The memo is four hundred words. No deck. No reconciliation appendix. She sends it.

At noon she takes her son to the doctor.

That is what the function looks like.

It looks boring.

At 7:42am the analyst sends the first reconciliation email. Five dashboards, three different numbers for the same channel, two different definitions of the conversion event. He has been doing this every Monday morning for four years. He is good at it. He does not enjoy it.

At 9am the CMO opens the team meeting. The first hour goes to deciding which of the five dashboards to trust for the budget conversation. No decision is reached. They schedule a follow-up.

At 10am the CFO emails. Can you walk me through the ROI on the YouTube spend? Board call is Friday.

At 11am the CMO drafts the response. Three slides. Numbers from the platform-reported dashboard. She knows, privately, that the numbers overstate the channel's actual contribution. The CFO will accept the slides because he has no independent way to falsify them. They have been accepting numbers like this, quarterly, for the last six years.

At noon the Meta agent the team deployed three weeks ago helpfully reallocates ten percent more spend toward a campaign whose attributed ROAS is the single most inflated number in the entire stack. The reallocation is logged in the agent's activity feed. No one reads the activity feed.

That is also what the function looks like.

It looks busy.

The first time I saw a Brand A function in the wild, I was not impressed. I had spent two years selling against the kind of theatre Brand B's function produces, and I was expecting the destination to be a louder version of the theatre - more dashboards, more agents talking to each other, more action per hour, more visible AI.

The Brand A function I walked into looked, frankly, slow. The CMO was answering one email an hour. The analyst was reading a textbook on Bayesian inference at his desk. The dashboard was almost empty - three numbers I could see, with a methodology link next to each. I asked the CMO what she did all day.

"Mostly I read what the system surfaces and I ask whether the question it answered is the question I actually had."

That was the day I understood what the destination looks like.

The destination is not more sophisticated on the surface. There is more sophistication underneath the surface, which produces more simplicity on top of it.

This is the part I think Andrej Karpathy has been getting at consistently in his writing on LLM systems for the last two years. The context the model reasons against - the retrieval, the working memory, the substrate - is the rate-limiting input to everything the model produces. A worse substrate cannot be compensated for with a better surface. A better substrate produces a quieter, simpler surface, because the surface has less reconciliation work to do.

The same principle applies to marketing functions. The substrate is the causal measurement layer. The surface is what the CMO does on Monday. A function with a calibrated substrate has very little surface work. The agents handle the boring reallocations. The dashboards show three numbers. The CMO reads what the system surfaces and asks whether the question was the right one. The function looks slow because the work that used to be visible - the reconciliation, the deck-making, the dashboard-trust meetings - has moved underground into a substrate that requires no human time.

Will Thorndike's The Outsiders makes the same point about CEOs. The outliers in his study - Henry Singleton at Teledyne, Tom Murphy at Capital Cities, John Malone at TCI - looked boring to outsiders for most of their tenures. Their boards thought they were too quiet. Their press coverage was thin. Their decision cadence was slow. What they had done was the hard work upstream of the decision, which made the decision itself look trivial.

The marketing function that's going to win the next decade will look the same way. Slow on the surface. Calm in the conversation. Three numbers in the board memo. The work happened upstream, in the substrate, eighteen months earlier.

I want to be honest about something that has become clearer to me over the last six months of customer conversations.

Most companies are going to hear this argument, agree with it, intend to make the migration, and not make it.

The reason is not stupidity. The reason is what Geoffrey Moore named forty years ago in Crossing the Chasm. The pragmatist majority of any market does not adopt a new infrastructure until the cost of not adopting it has become undeniable inside their own peer group. The early adopters are already two cycles ahead. The pragmatists are watching the early adopters and waiting until the gap is so loud that not migrating reads as negligence.

We are right at that inflection. The brands that completed this migration in 2024 and 2025 are now visibly lapping their categories. Their CAC curves look different. Their CFO conversations look different. Their AI deployments are not embarrassments. By late 2027 the gap will be loud enough that the pragmatist majority will move. By 2028 it will be the default. By 2029 the question will be whether you survived the four years in between.

The deeper structural reason most companies will not make the migration on time is the one Clay Christensen named in The Innovator's Dilemma. The existing function - attribution dashboards, attribution-based budgeting, attribution-based vendor relationships, attribution-based agency contracts, attribution-coded promotion criteria for marketing analysts - is profitable enough, in the short term, that questioning it feels premature. Until it is too late.

I keep watching marketing teams discover, six months after the migration window has closed, that the cost of moving has just crossed the cost of waiting. Most of them are going to migrate in 2027 or 2028, not because they figured it out earlier than their peers, but because the gap finally became too loud to ignore.

The companies that move now will be running an unmistakably different function eighteen months from now. The companies that wait will spend 2027 and 2028 catching up to where Brand A was in 2026. That is the asymmetry the next two years are going to make legible.

I want to be specific about what the compounding gap looks like, because I think most CEOs are still underweighting it.

The brand that completed the migration in 2024 has, by 2027, accumulated three things their non-migrated peer has not. First, a calibrated record of which channels actually drove growth - five quarters of incrementality-anchored evidence that lets the team allocate the next quarter with confidence intervals their CFO can read. Second, AI agents trained against a substrate that is not biased - agents whose recommendations the team has learned to trust because the recommendations have, demonstrably, been right. Third, an operating cadence in which the CMO and CFO debate one or two genuinely-uncertain questions per quarter instead of relitigating attribution every Monday.

The non-migrated peer has none of these. They have, instead, six more years of accreted attribution debt - historical numbers that everyone privately knows are inflated but that the entire budget process is reverse-engineered from. They have agents reasoning against a biased substrate, which means the agents are confidently optimizing in the wrong direction. They have a CMO who spends most of her week reconciling dashboards instead of running her function.

The compounding part is the agent layer. An agent making decisions against a biased substrate gets worse over time, not better, because the substrate's bias compounds across every reallocation the agent makes. By 2028 the non-migrated function is running on a substrate that has been further corrupted by the agents that were deployed on top of it. The substrate cannot be unwound from the agents. The unwinding is, in practice, a complete rebuild.

That is what waiting costs. Not a delay. A debt that compounds inside the agent layer until the entire function needs to be rebuilt from the substrate up.

Three things I am watching as the destination becomes more visible.

One. Whether the first major brand publishes its measurement methodology as a public disclosure. The next step is brands publishing their own methodology disclosure footnotes - three numbers a quarter, each defended by an experiment, in a format an investor can read. The first major DTC brand to do this publicly is going to set a pattern that becomes a competitive expectation within twelve months. I expect this in the second half of 2026.

Two. Whether the CMO–CFO joint dashboard becomes a board-reporting standard. I wrote in Issue #4 that audit committees would start asking about marketing measurement methodology. The natural artifact of that pressure is a dashboard the CMO and CFO co-sign and the audit committee reviews quarterly. I expect at least one Fortune 500 company to publicly disclose that they have adopted this format by Q4 of next year.

Three. Whether AI-native marketing job descriptions appear publicly. The earliest signal that the new function exists will be a CMO posting at a high-growth brand that explicitly says: we operate on a unified causal measurement substrate. We do not run our marketing function on platform-reported attribution. You will read what the system surfaces and ask whether the question it answered is the question you actually had. Three years ago that would have been an absurd job description. By 2027 I expect to see the first one. Send it to me if you see it before I do.

I want to close with the part I owe the reader.

The Brand A picture I painted at the top is real. I have walked into it. But it is not yet common. The functions I have seen approach it are the ones that started the migration eighteen to twenty-four months ago, and even there, the Monday morning I described is the steady-state, not the migration period. The migration itself is messy. It takes twelve to eighteen months. The political work inside the company is real. The change-management work on the agency side is real. Some functions that have started the migration are still in the awkward middle of it, where the new substrate is partly built and the old dashboards are still being maintained.

I also want to be honest about a misread of this piece I want to head off.

The point is not that every marketing function should look identical. There are categories - early-stage brands, hits-driven businesses, businesses where the creative judgment is genuinely the rate limiter - where less rigor and more storytelling is correctly the operating model. Cassie Kozyrkov's writing on decision intelligence makes this distinction better than I can. The rigor is calibrated to the magnitude of the decision and the volatility of the underlying business. A small DTC brand spending two hundred thousand dollars a quarter does not need the Brand A function. A consumer brand spending a hundred million dollars a quarter cannot afford not to have it.

The version of this argument I am making is for the second category. The mid-market and enterprise consumer brands whose marketing budgets are material capital allocation lines. Those functions have the most to gain from the migration, and the most to lose from waiting.

If you sit inside a marketing function that has begun this migration - or one that hasn't - I would like to hear what your Monday morning actually looks like. The cleanest way to know whether the destination is real is to compare notes with the operators who are in the middle of building it.

Thanks for reading the fifth one.

Tobin Co-Founder & CEO, Lifesight June 2026