Your MMM & MTA Tools Are Starving for Deterministic Data

Hi there,

Three operators gave me the same objection last month:

"We already use Triple Whale."

"We have Northbeam."

"We're covered -- we have an MMM."

But here's the question I asked back: "What data are you actually feeding it?"

Because your MMM and MTA can only optimize what they can see.

Your MMM is working off modeled data

When I ask what's actually flowing into the model, it's almost always the same answer: e-commerce. DTC transactions. The buyers who came through the website.

Which, for most omnichannel brands, is a sliver. The retail, wholesale, and marketplace purchases (the majority of revenue) just don't show up. They land in what I call the retail black box. The model can't see them, so it infers them.

Most brands are feeding their tools this modeled, probabilistic data, then modeling on top of it. Best guesses stacked on best guesses. When the underlying data is inferred, what you get back is a basic level of reporting dressed up as precision.

The tool isn't the problem. MMM tells you which channels are working. MTA maps the touch path.

The problem is accuracy.

And the industry feels it. In Skai's 2026 State of Retail Media report, incrementality was the single biggest measurement challenge marketers named (75%), and only 15% said they were very or extremely effective at measuring retail media performance at all [1]. The top barrier to getting incrementality right? Concerns about the accuracy or reliability of the results (44%) [2].

So no, you don't need to rip out your MMM or MTA. You need to make it more reliable.

Probabilistic tools get sharper when you feed them something real

You can't fix an accuracy problem by buying a smarter model. You fix it by feeding the model verified, deterministic purchase events instead of inferences.

Deterministic, first-party offline buyer data does what modeled data can't: it confirms a purchase actually happened, tied to a real customer, a real store, a real SKU. Feed that in and the model stops inferring retail and marketplace performance and starts seeing it.

What the brands pulling ahead figured out

The operators who've cracked this went and found the buyers their channels never handed back, and they started feeding those purchases into the same stack everyone else is running.

That's the whole move. Same MMM. Same MTA. A more complete signal flowing in.

That signal essentially proves or disproves whether the purchases your MMM or MTA modeled actually happened. That feedback loop is what makes the model meaningfully smarter over time. You stop optimizing against a best guess and start optimizing against what really occurred.

You already have the model. The real unlock is whether it's seeing your whole business or a quarter of it.

So the next time someone tells me "we already have an MMM," I ask the only question that matters: what are you feeding it?

Talk soon,

Kait

A few extras:

Some of my favorite brands are hiring…

1. Bulletproof 360 - Marketplace Business Manager

Sources:

[1] eMarketer and TransUnion, "MMM, incrementality, and other measurement trends that will define 2026," 2025.

[2] Stacked Marketer, "Meta goes all-in on AI, Demand Gen gets new capabilities," June 26, 2026 (newsletter).