Why I Started Uncommon Move and What Move 37 Has to Do With It

In March 2016, the world's best Go player sat across from a machine and lost.

Lee Sedol had won 18 world titles. AlphaGo was software. Experts had given AI a decade before it could crack Go, a game so strategically complex that brute-force computation couldn't touch it. Then AlphaGo played Move 37.

It was the second game of the match. AlphaGo placed a stone so far from expected territory that the commentators assumed it was a mistake. No human player would make that move. It violated every conventional heuristic the game had developed over centuries.

Lee Sedol took nearly fifteen minutes to respond. He walked away from the board.

AlphaGo won the game. Move 37 was the turning point.

What struck me watching that footage wasn't the AI. It was the reaction. The commentators, some of the most knowledgeable Go players on the planet, saw a mistake because they were pattern-matching against everything they had ever learned. Their expertise was real. But it was expertise built inside a particular frame. AlphaGo didn't have that frame. It saw the whole board differently. And that different view produced a move that looked wrong but was exactly right.

The problem I kept running into

I've spent years working inside the Google advertising ecosystem. Not one channel but all of it. Search, YouTube, Demand Gen, Performance Max, Display and Video 360. The measurement layer underneath: GA4, Meridian Marketing Mix Modeling, incrementality testing. The AI infrastructure that now sits on top of everything and changes what's possible every few months — which is exactly why I argue you should never build for a single LLM.

The work taught me a lot. But the most important thing wasn't how any single product worked. It was how rarely organizations see the whole board at once.

A hands-on campaign manager knows their campaigns deeply. A VP knows their channel budgets and aggregate ROAS. A CFO knows their media spend as a line item. Almost nobody in the room has a clear view of how the pieces interact. How YouTube signals feed Search efficiency. How Demand Gen builds the audience pools that make Performance Max work harder. How Meridian MMM reveals that the channel with the best attribution report is often not the channel doing the most incremental work.

That's not a failure of intelligence. It's a failure of frame. Everyone is optimizing the piece they can see. Nobody's watching what happens at the intersections, and that's where the compounding value lives.

What the fragmented view actually costs

Here's what silo management costs in practice. A media team runs Google Ads, YouTube, and DV360 as separate channels with separate budgets, separate agencies, and separate measurement logic. Each channel reports its own performance. Each looks defensible in isolation.

Meanwhile, the upper-funnel investment that's priming purchase intent and quietly making the bottom-funnel campaigns more efficient shows up as a cost center with soft metrics. It gets cut in the next budget cycle because it can't prove its value in the attribution model everyone's using. Search CPAs start creeping up over the next two quarters. Nobody connects the dots.

This isn't a hypothetical. It's a pattern that keeps repeating because the measurement infrastructure most organizations rely on is built to report channel performance, not ecosystem value.

The fix isn't complicated in principle. You need a measurement layer that sees across channels: Marketing Mix Modeling, holdout testing, GA4 configured to capture cross-channel signals. You need campaign architecture designed around how the channels interact, not just how each performs alone. And you need a strategic frame that connects media decisions to business outcomes, so investment conversations happen at the right level.

Where this view came from

I didn't arrive at the whole-board frame by theorizing about it. I arrived at it the slow way — by working inside the Google ecosystem since 2008, across nearly two decades, through every reorganization of how the platform learns and what it rewards. Long enough to watch Google Ads go from a keyword auction you could read line by line to a set of machine-learning systems that decide most of the allocation for you. Long enough to see Performance Max arrive and split the industry into people who fought the black box and people who learned to feed it. Long enough to sit through the measurement wars — last-click, data-driven attribution, then Marketing Mix Modeling — and notice that each new model mostly relocated the blind spot instead of removing it.

That tenure matters here for one reason only: it's why I trust the pattern. A single bad quarter can be a fluke. The same failure repeating across accounts, sectors, and a decade and a half of platform changes is structural. When I say upper-funnel investment gets cut because it can't defend itself in the attribution model everyone's using, I'm not describing a thought experiment. I'm describing something I've watched happen, predictably, to good teams with smart people, over and over.

It's also a view Google itself has put me in the room to share. In 2026 I was selected for Google's Master of Leadership Summit — one of only 34 people in the country chosen — which is as direct a signal as I can offer that the systems-level read on this ecosystem isn't a fringe take. It's the conversation the platform's own leadership is having.

The specifics changed every couple of years. The shape never did. A team optimizes the piece it can measure, the measurement system rewards the channel that reports well rather than the one doing the incremental work, and the compounding value that lives at the intersections goes uncounted until it shows up as a slow, unexplained rise in blended cost. The tools got more sophisticated. The frame stayed narrow.

Why I work at the systems level

Most marketing content operates at the campaign level — settings, tactics, the lever you pull this week. That work is real and I've done plenty of it. But it's not where the largest, most durable mistakes get made. The expensive errors are architectural: how the channels are wired together, what the measurement layer is actually capable of seeing, and whether media decisions are connected to business outcomes or just to channel dashboards.

So that's the level I write at. Not because tactics don't matter, but because the tactics only compound when the architecture underneath them is sound. It's the same reason I argue you should design for the system rather than the tool — whether that's refusing to build for a single LLM, configuring Meridian MMM to see across channels instead of crowning the best-reporting one, or treating Performance Max as a signal problem rather than a campaign problem. The throughline is always the same: optimize the whole board, not the square you happen to be standing on. And when the conversation reaches the budget meeting, that means translating the architecture into the only language that survives there — how the decision reads on the P&L.

Why this space exists

I started Uncommon Move as the space I didn't have. Somewhere to think out loud about the full system. Not vendor talking points, not surface-level certification content, but the framework-level thinking that connects execution to business strategy.

The audience isn't one role. It's the campaign manager who wants to understand why the ecosystem works the way it does. It's the marketing director trying to make the case for measurement investment to a CFO who wants cleaner attribution. It's the business leader trying to figure out whether their Google investment is structurally sound or just reporting well.

Those three people need different entry points into the same picture. That's what I'm building here.

The name comes from Move 37. Not because AI is the whole point, though it's increasingly unavoidable, but because the most valuable moves in complex systems are often the ones that look wrong from inside the conventional frame. The ones that require seeing the whole board.

Move 37 looked like a mistake. It won the game. The question isn't whether the Google ecosystem is getting more complex. It is. The question is whether you're building the capability to see the whole board, or optimizing one piece at a time and calling it strategy.