Can MTA Survive in a Cookieless World?

| Last Updated August 11, 2025 - by brett crawford under data & application engineering

I had the privilege of interviewing Brett Crawford, a marketing attribution expert at Softcrylic, to discuss the evolving role of multi-touch attribution (MTA) as the industry prepares for a cookieless future.

Marketing attribution has always aimed to answer a deceptively simple question: What’s working? For years, MTA felt like the most promising answer. It moved the industry beyond basic last-click models toward something more holistic, where multiple marketing touchpoints could share credit for a single conversion.

But now, that promise is under pressure.

“MTA was the solution to the recognition that traditional attribution was sort of a one-winner-takes-all type of model,” Crawford explains.
“But lately, it’s becoming less popular because of privacy concerns, with the deprecation of cookies.”

What Happened to MTA?

A major force working against MTA is the rise of walled gardens-platforms like Facebook or X (formerly Twitter) that keep their user-level data locked inside.

“A platform like Facebook can track all the user-level interactions within that platform, but there’s no way to get that information out,” Crawford says.
“They’re never going to ship their data out because that’s their secret sauce. It’s the data.”

On top of that, the end of third-party cookies has been discussed for years-yet still hasn’t fully arrived.

“This has been a conversation since I got into the field six years ago, and the can just keeps getting kicked down the road.”

Still, Crawford doesn’t see this as the end of MTA.

“As a technical approach, MTA is never going to be completely irrelevant,” he emphasizes.

“It just depends on whether you have the ability in a post-cookie world to tie people together across touchpoints.”

MMM Is on the Rise

As MTA faces structural barriers, marketing mix modeling (MMM) is gaining traction-often thanks to investments from the very platforms that disrupted attribution in the first place.

“When they recognized that cookies were going to become less relevant, companies like Google and Facebook started investing in open-source MMM platforms,” Crawford notes.
“Facebook has Robyn, Uber developed one, and Google also has one.”

Unlike MTA, MMM doesn’t require user-level data. Instead, it uses aggregated trends, which makes it both privacy-safe and better suited for long-term strategic planning.

“MMM is very aggregated, like the type of model the Federal Reserve would build to measure economic policies,” he says.
“The data’s high-level, but you can still use it to model marketing performance.”

Crawford sees this kind of modeling as essential for forecasting major business shifts.

“What happens if the economy crashes, or I expand into a new market?” he poses.
“MMM is the strategic layer. MTA is more tactical. Ideally, you’d have both.”

lncrementality Brings a Reality Check

Crawford also emphasizes the growing role of incrementality testing-a method for measuring the true lift from a marketing effort by comparing outcomes with and without it.

“You’re creating two hypothetical worlds-one where there’s marketing spend, and one where there’s none-and showing the difference. That’s the incremental effect of your marketing.”

Used in tandem with MMM and MTA, incrementality helps validate whether your attribution models reflect real-world behavior.

Will Al Be the Next Breakthrough?

There’s been a lot of hype around Al in the measurement space, but Crawford approaches it with a healthy dose of realism.

“A few years ago we could’ve said MTA and MMM were Al,” he says.
“So there’s a buzzword element. But in the agentic Al world, there’s no reason to think an LLM couldn’t build an MTA model.”

And with newer models of web interaction on the horizon, Al may offer more than just labor-saving automation.

“The current web is like multiple apps, with the browser acting as the interface,” Crawford explains.
“The cookies lived in the interface, which allowed the different apps to be linked. With agentic Al, it seems likely that it will serve as a new universal interface to the web. In this centralized UI model, I’d bet MTA would come back, marketed by whoever owns the agentic Al interface and can see all of a person’s activities.”

The Real Bottleneck: Culture, Not Code

Despite all the advancements in modeling and Al, Crawford believes the hardest part of measurement isn’t the math-it’s the culture.

“You could build the best model in the world, but if the incentive structure treats losing budget as losing power, people are going to say it’s a bad model-for very personal reasons.”

In his view, organizational readiness is the true make-or-break factor.

“So to me, that’s the biggest challenge with some of these things is: are you in an organization where people will uncritically use this data to make decisions, and there’s no perverse incentives affecting that decision?”

So Where Do We Go From Here?

For all the talk of models and methodologies, Crawford believes success starts with the fundamentals-reliable data, disciplined collection, and internal teams aligned on how to use it.

“Data collection is about putting a pixel on your site or writing the code to collect what you need,” he explains.
“Hygiene is about cleaning the data-digital data is messy, and there’s nothing stopping people from typing whatever they want into a query string.”

But even the best insights don’t mean much if they’re ignored.

“The rubber hits the road at: where do I put my $100,000 to maximize return? The challenge is not the math. The challenge is getting people to use the output.”

Attribution doesn’t need to be reinvented. It needs to be applied with discipline. In the right hands, MTA can still help marketers make smarter, faster, more confident decisions in a post-cookie world.