Rethinking Marketing Measurement: How To Spend Your Budget Like an NFL GM

​​Marketing has always had a measurement problem.

For years, we’ve been searching for the perfect model—the silver bullet that will tell us exactly which ads work, which channels deserve more budget, and which tactics are moving the needle. And for years, we’ve settled for flawed proxies. We’ve over-indexed on attribution models because they’re easy to digest and cleanly presented, even if they’re directionally misleading.

We’ve been optimizing for what’s most visible, not what’s most valuable.

But that’s changing. We’re on the cusp of a new era in marketing measurement—a golden age where the tools we’ve long needed are finally becoming accessible. Measurement methodologies like media mix modeling (MMM), incrementality testing, and attribution modeling have existed for decades, but they’ve always been reserved for the biggest brands with the deepest pockets and armies of analysts.

That’s no longer the case.

Last week, I attended an exclusive event at Google’s New York office called Rethinking ROI in the AI Era. The event was under NDA, so I can’t share specifics. But I can tell you this: we’re about to see a wave of powerful measurement tools roll directly into the platforms we already use every day—tools that bring rigorous data science into the hands of more marketers than ever before.

‍

And at the center of the conversation was a simple triangle of three measurement methods: Attribution. Incrementality. Media Mix Modeling.

This graphic came up in almost every session of the day. Three corners. Three methods. Each pointing to different types of insights—and each informing the others. It was the most important visual of the entire event, because it captured what so many marketers still get wrong: there is no perfect, all-encompassing measurement solution. But used together, these three methodologies offer a full-funnel, probabilistic, and practical framework for understanding performance in today’s complex marketing ecosystem.

While the methodologies aren’t new, many marketers still don’t understand them. Or worse, they use the wrong tool to answer the wrong question.

So we need a better mental model. And I think I’ve got one.

If you know me, you know I’m a Buffalo Bills fan.

An unhealthy obsession with this team is rooted in my blood, and I spend a large chunk of my free time watching "game tape" and carefully tracking offseason roster moves. And as strange as it sounds, that obsession has helped me understand marketing measurement in a much deeper way.

Because managing a football team—especially under the NFL’s salary cap—is a near-perfect analogy for how marketers should think about allocating budgets and evaluating performance. Just like marketing teams, general managers like Brandon Beane of the Buffalo Bills operate within strict constraints. They can’t pay everyone top dollar. They have to make bets. They have to evaluate not just who scored, but who mattered.

And they have to get it right, or the whole thing falls apart. 

So in this post, we’re going to explore the three pillars of modern measurement—attribution, incrementality, and MMM—through the lens of the 2024 Buffalo Bills. Because if you understand how Beane thinks about roster construction, you’ll understand how to think more clearly about your media mix, your measurement strategy, and the true value of your marketing efforts.

‍

Box Scores and Attribution: The Limits of Simplistic Measurement

A box score is a collection of metrics from a football game. It’ll show you stats like passing attempts, completions, and touchdowns, just like how a marketing dashboard might show you impressions, clicks, and conversions.

A box score is a helpful way to get a quick snapshot of a game, but every sports fan knows it never tells the full story. 

In 2024, Bills running back James Cook scored 16 rushing touchdowns—tied for the most in franchise history. It’s the kind of stat that jumps off the page. A quick look at the box score would suggest he was the team’s most valuable offensive weapon. If marketing used this same logic, an attribution model would hand him the biggest slice of the salary cap. Why? Because he finished the drive. He crossed the goal line. That’s the conversion.

But that would be a mistake.

Because anyone who actually watched the Bills play knows that Cook—while talented—is not the engine of the offense. That’s Josh Allen. Allen was the league MVP, and you don’t need a data model to tell you he’s the most valuable player on the roster. It’s obvious.

Now, an advanced attribution model might be able to be tweaked to give Allen partial credit—some kind of “assist” logic, the same way we might apply credit to an ad that appeared earlier in the user journey. That’s better. But even then, it still falls short.

Because it misses the Spencer Browns of the world. Who is Spencer Brown, you ask? That's exactly my point!

If you watch a highlight reel of James Cook’s 2024 season, you’ll notice a recurring theme: right tackle Spencer Brown consistently blowing defenders out of the way, creating the space for Cook to break into the open field. Brown doesn’t score touchdowns. He’s not on anyone’s fantasy team. He doesn’t show up in the box score. And yet, his contribution is critical.

No attribution model—not even the most advanced, AI-powered one—can fully capture that kind of impact. Not in football, and not in marketing.

That’s because NFL teams don’t win games with just one or two stars. They have 53 players on the active roster. And GMs like Brandon Beane have to allocate their salary cap—roughly $225 million—across all of them. Most of those players will never touch the football. Many will never score a touchdown. But all of them matter.

It’s the same with your marketing budget.

You might be tempted to throw more dollars at the last-click channel that looks the most efficient. But that would be like paying James Cook more than Josh Allen because he scored more rushing touchdowns. Or like ignoring Spencer Brown entirely because he doesn’t show up on the stat sheet. That’s not a strategy—it’s box-score thinking.

And box-score thinking doesn’t build winning teams.

Budget Constraints and NFL Salary Caps

Brandon Beane doesn’t get to build a fantasy team. He gets a hard cap—about $225 million in 2024—and has to build a 53-man roster within that budget.

That means every contract is a trade-off. Every dollar committed to one player is a dollar not spent elsewhere. And it’s not just about rewarding the guys who score the most touchdowns. Beane has to consider every position group, future salary demands, injury risk, locker room chemistry, and long-term upside. He’s managing complexity, not just outputs.

Marketers face the same reality.

We have finite budgets. And yet, we’re often tempted to allocate spend based on attribution models that only reward what’s easily measurable—typically, conversions. But that’s like handing out contracts based only on touchdowns. It’s reactive, short-term thinking.

Beane can’t afford to manage that way. Neither can we.

This is where marketing needs to move beyond attribution and start using broader, probabilistic methods like incrementality testing and media mix modeling (MMM)—tools that help evaluate performance in context and allocate budgets the way a smart GM builds a team.

We’ll get into that next.

Incrementality of the Wide Receiver Room

Let’s talk about wide receivers.

In 2024, Mack Hollins led the Bills in receiving touchdowns. He was a fan favorite, a physical presence, and came through in some big moments. But he wasn’t the most targeted receiver, and when the game was on the line, he wasn’t the first read in Josh Allen’s progression.

That was Khalil Shakir.

Shakir led the team in receptions, yards, and targets. His production wasn’t flashy—but it was consistent. It moved the chains; sustained drives. And that’s why, when the season ended, Brandon Beane gave Shakir a four-year extension worth up to $60 million… while letting Hollins walk in free agency.

If you just looked at touchdowns, you’d think that was a mistake.

But Beane isn’t optimizing for box scores. He’s optimizing for wins. And that means looking at the full context of a player’s impact—not just the final stat line.

This is exactly what incrementality testing allows us to do in marketing. Incrementality allows the GM to understand the impact that a given player had on the team's overall performance. For example, How many of James Cook's touchdowns were only possible because Khalil Shakir caught a crucial 3rd down reception early in the drive?

So instead of simply asking, “Which marketing campaign had the most conversions?” we ask, “What actually changed behavior?” We isolate variables. We hold out a control group. We run the kind of experiments that help us understand true lift—not just surface-level performance.

It’s how we uncover that a YouTube campaign didn’t just drive direct conversions, but increased branded search volume and lifted conversion rates across all channels. Or how a paid social campaign might not convert directly, but helps influence higher email marketing open rates.

Amari Cooper is another example. Cooper was a big name who joined the Bills midseason, and his modest stat line left a lot of Bills fans feeling frustrated about his lack of production. But simply evaluating Cooper based on his individual production misses the larger picture: Cooper became the team’s largest receiving threat, which is something opposing defenses need to plan around. He drew coverage. He opened up the field. He created space for others to thrive. When Amari Cooper was on the field, the Bills had a higher EPA (expected points added) per play, and in games where Cooper was active, the offense scored more points—the most ever recorded in franchise history. It’s clear that a player’s value cannot be distilled to the box score metrics, just as the value of a marketing asset might not be incredibly clear even with advanced attribution modeling.

Incrementality testing—a fancy way of describing experiments—lets us detect those kinds of effects in our campaigns.

It helps us see the “on-field” value of a tactic—even when the conversion doesn’t show up in the box score. It reveals the true difference-makers. And it helps marketers avoid the mistake of cutting spend on a campaign just because it isn’t finishing the play.

Just like Beane doesn’t cut a receiver because he isn’t scoring touchdowns, smart marketers don’t pull budget from a campaign that’s indirectly supporting the efficacy of other tactics.

Finally, let’s talk about Stefon Diggs.

Diggs was the Bills’ No. 1 receiver for four seasons and a perennial Pro Bowler—arguably the most high-profile weapon in the offense during that time. So when he was traded to the Houston Texans before the 2024 season, many fans were devastated. Nearly every media outlet predicted a major step back for the Bills' offense.

But much to my personal delight, the opposite happened.

Without Diggs, the Bills became more productive. And while it’s easy to chalk that up to “addition by subtraction,” there’s something deeper going on here. When you have a star like Diggs on the roster, the entire offense tends to bend around him. The coaching staff feels pressure to build plays that feature him. The quarterback might lock in on him as the first read, even when another player is better positioned to make the play.

Very few people expected the offense to improve after losing Diggs—there was no way to know without seeing it happen. And that’s the thing about incrementality: to understand true impact, you need a holdout test.

In marketing, a holdout test might look like turning off YouTube ads in one region while keeping them live in another, then comparing outcomes. In football, the 2024 season gave us a real-world holdout test for Diggs’ presence in the offense.

And the results were pretty clear: Turns out, the offense ran smoother, produced more points, and diversified its targets without him.

Good riddance, Stef!

‍

Dawson Knox and Media Mix Modeling: Forecasting With Imperfect Information

Let’s talk about complexity.

NFL general managers have to build a full roster—offense, defense, special teams—with 22 starting positions, backups at every spot, role players, specialists, and a finite amount of money to work with. That’s what makes the salary cap so difficult to manage. 

This is exactly where Media Mix Modeling (MMM) shines.

MMM is a high-level measurement methodology that helps marketers answer the same question GMs face: How do I distribute my finite budget across a wide range of needs, roles, and opportunities?

It looks at historical data across all your marketing channels—search, social, TV, video, email, display—and uses statistical models to identify which combinations of spend tend to drive the most impact. It's a strategic tool, not a tactical one. It doesn’t tell you what ad copy to test this week; it tells you whether your paid social budget should be closer to 20% or 40% of your total media mix.

Think of it as your salary cap allocation strategy.

Let’s say your MMM tells you that, historically, tight ends have been a high-leverage position in your offense. They convert in the red zone, block in the run game, and create matchup problems for defenses. Based on that insight, you might decide to invest more heavily in that position going forward.

That’s what the Bills did with Dawson Knox.

Heading into the 2023 season, Knox had been productive, especially in high-leverage situations. The Bills restructured his contract and gave him one of the highest cap hits on the team, hoping they’d replicate the magic of the Patrick Mahomes/Travis Kelce relationship. The data supported it, and a smart MMM would’ve told you the same thing.

But performance doesn’t always match forecasted expectations.

Knox’s output dropped significantly over the next two seasons. He caught half the number of passes and a fraction of the total touchdowns. Bills fans love Dawson Knox, but we don't love the fact that his contract still makes him the third highest cap-hit on the team. 

That’s the reality of modeling. You’re working with imperfect information. You’re trying to predict the future based on the past. And sometimes, the model misses.

But that doesn’t make MMM useless—it just means it needs to be updated.

Also, two things can be true: Knox can be a net positive for the Bills organization as a whole, but his contract can be a suboptimal use of the salary cap budget. The same is true in marketing: CTV ads might generate incremental revenue, but if the CPMs are too high, it might not be profitable.

Marketers need to continuously refine their models based on what’s actually happening in the market. MMM is a starting point—a strategic guide for where to place your bets. It helps you plan. But like all models, it should evolve over time.

The Measurement Triangle: Three Methods, One System

Attribution. Incrementality. Media Mix Modeling.

Each method has a distinct role to play—and each helps answer a different type of question.

Media Mix Modeling (MMM) helps answer the big, strategic question:
“How should I allocate my budget across a wide range of channels and campaign types?”
It looks at historical performance across your full media mix and uses statistical modeling to forecast where your dollars will have the most impact over time. It’s your salary cap strategy. Your allocation map. It doesn’t operate in real time, but it gives you a high-level view of how to invest smartly across the full roster of tactics and platforms.

Incrementality testing (aka experiments) helps answer the performance question:
“Which campaigns or tactics are having the most impact on my ROI?”
By holding out control groups or testing different treatments, incrementality isolates what actually drives behavior—separating correlation from causation. It’s your way of understanding which players are making a difference when they’re on the field, and which ones might be getting credit without creating real value.

Attribution modeling helps answer the optimization question:
“How can we make sure our budget—once allocated—is working as hard as possible?”
Attribution gives you real-time signals about which channels or touchpoints are converting most efficiently. It’s not built for strategic planning, but once your budget is set, it’s incredibly useful for tactical execution. Think of it as your in-game analytics—helping you call better plays with the players you’ve already chosen to fund.

No one model can do it all. But together, they create a complete, layered measurement system.

That’s what that triangle at the Google event captured so well: these models aren’t competing—they’re complementary. When used in tandem, they make each other smarter. Incrementality testing can refine your MMM forecasts. MMM can establish the strategic guardrails that attribution works within. Attribution can provide directional signals that highlight where deeper testing is needed.

It’s not about choosing a favorite. It’s about building a system where each tool reinforces the others—just like a football team is built with complementary roles across the roster.

That’s how you win. On the field, and in the market.

Final Thoughts: Moving Beyond the Box Score

We’re entering a new era of marketing measurement.

The tools are finally catching up to the complexity of the job. Media mix modeling, incrementality testing, and attribution are no longer reserved for the top 1% of advertisers. Thanks to AI, platform integration, and smarter software, these methodologies are becoming more accessible to more teams. That’s exciting—and it’s long overdue.

But access alone isn’t enough.

Marketers need to adopt the right mindset. That means letting go of the myth of perfect, deterministic tracking. It means understanding that no single model will tell you everything. And it means building a system—just like a great GM does—that uses multiple perspectives to guide smarter decisions.

Measurement isn’t about declaring winners and losers. It’s about creating a more honest view of performance. It’s about making sure your budget is working as hard as it can, and being able to explain why you’re investing where you are.

So here’s the takeaway: Don’t build your media strategy around box scores. Build it like Brandon Beane builds a roster—with context, trade-offs, long-term thinking, and the humility to know that not every bet will pan out.

This analogy comes full circle with the fact that the Bills 2024 season ended one game short of the Super Bowl. Beane built a great roster—good enough to be within arms reach of the title game. So, like every offseason before, he’ll go back to the data, optimize the approach, and look for every edge that brings him closer to the ultimate goal.

Marketing is no different. Build the right infrastructure, then make disciplined, ongoing optimizations that slightly improve your odds of success over time.

And as always: Go Bills.

Finally, a special thank you to Joe Marino of Locked On Bills, whose podcast and Substack content helped shape this piece. If you're a Bills fan, I highly recommend Marino's Substack community.

‍