Models like Bayesian and NBD-Dirichlet are powerful ways to predict human behavior. But they don’t explain why things happen. They can spot patterns. They don’t prove cause. In this recap of Part 2 of The Causal CMO, Mark Stouse explains the shift from pattern-based models to causal inference and why that shift matters now more than ever for GTM teams.
One of the things Mark started with is an age-old question we’ve always tried to answer in business:
“Why things happen is the number one thing that the scientific method seeks to understand.”
Bayesian models were a step in that direction. But the Bayes’ Theorem is 300 years old. And they’re predictive. They can’t tell us what caused what and why.
What they are very good at is telling us the probability that something is happening.
“If you see smoke, a Bayesian model helps update the probability there’s a fire. But it won’t tell you what caused the smoke.”
Same goes for NBD-Dirichlet. It’s great for describing past behavior. It can even predict short-term purchasing patterns.
You can see this in action on many e-commerce platforms. Amazon uses NDB-Dirichlet to model the probability of repeat purchases. If you buy a certain product, for example, you almost always see a “You May Also Like” CTA.
This kind of modeling assumes that buyer behavior doesn’t change much over time.
But human beings are irrational. We choose A today and B tomorrow, depending on how we feel. And in B2B tech, with its layers of procurement bureaucracy, stakeholders, and decision time lag… well, you see where I’m going with this.
Most GTM systems in place today are designed to support correlation-based marketing, not causal decision-making.
They’re pattern matchers. Attribution tools. Regression models.
None of them explain cause and effect. They were built to scale performance efficiency, not understand impact like cost-effectiveness.
Marketing is just as much to blame as anyone. But the rot began with deterministic thinking at the leadership level.
“Most of the bad info came from founders and VCs. They wanted deterministic systems. The idea was simple: put a quarter in, get a lead out.”
That kind of worked for a while, when interest rates were low, uncertainty was minimal, and pipelines moved fast.
But once volatility hit (time lag, market noise, internal complexity), those patterns fall apart.
Causal inference doesn’t just show you a pattern. It tests whether that pattern causes anything meaningful to happen.
And it does it continuously in real-time.
That’s the power behind causal AI.
One of the biggest reasons GTM performance has tanked in the last two decades is that most GTM teams still operate like the environment hasn’t changed.
“70% of the world is stuff you don’t control. And most marketers don’t even account for it.”
The effectiveness of GTM teams has dropped off a cliff, not because marketers suddenly got worse, but because the headwinds got stronger and faster.
It’s a full-blown marketing effectiveness collapse, clearly visible in 2025.
And yet, everyone still keeps trying to optimize for efficiency. Perhaps it’s because we blindly believe we are already effective.
“You can’t be efficient until you’re effective.”
This is an important reminder: Marketing is a non-linear multiplier. Sales is a linear value creator.
For the past 25 years, we’ve been trying to force marketing to abide by Sales’s linear outcomes. That’s no different than forcing a square peg through a round hole.
In a causal model, you can calculate the lift marketing creates while Sales is executing.
If Sales underperforms, Marketing’s lift is zero. If Sales is kicking butt, marketing can multiply Sales efficiency by 5x and Sales effectiveness by 8x.
That’s not for debate. It’s proven math.
Sadly, that multiplier logic doesn’t show up in attribution because attribution is correlation.
It’s only visible in causal inference.
Mark explained the four types of AI in play today. Only one explains “why.”
Most GTM teams are stuck between the first two.
They’re still optimizing a traditional sales and marketing funnel with pattern-matching tools that can’t distinguish signal from noise. In other words, transactional tactics.
Worse still, they’re getting excited about autonomous agents that “do work for you,” without asking if the work being done is even useful.
“Agentic AI without causal logic is just automation with lipstick.”
One of the most overlooked shifts in GTM accountability is that causal models are increasingly being used by finance.
“FP&A and ops teams are going to be the ones evaluating GTM performance. Not marketing itself.”
This shift is already underway. It’s part of a larger response to Delaware’s expanded fiduciary rules.
“All officers now carry personal liability if shareholder risk isn’t managed responsibly.”
Which means random budget cuts, especially in marketing, are going to get harder to justify.
Causal AI gives CFOs the scalpel they’ve needed for years. It helps them decide what to cut, and more importantly, what not to cut. This is how CFOs use causal AI for GTM decisions.
Causal AI is like a CRM for cause and effect that updates continuously and informs real decisions in real time.
One of the analogies Mark uses when explaining Causal AI is that it’s like a GPS. It doesn’t promise precision. But it helps you avoid collisions and reroute when the road changes.
“The route that worked last quarter might not work today. The conditions changed. The terrain shifted. And if you’re not paying attention, you’re going to hit something.”
And it’s not just rerouting.
These systems simulate what’s likely to happen next based on shifting inputs, so you can course-correct before problems hit.
So if you’re running models designed for deterministic systems, you’re essentially driving blind.
The hype around AI isn’t going away. Neither is the pressure to “cut costs,” especially in GTM.
A lot of budget cuts these days are based on correlation, on what appears to be contributing. But contribution isn’t the same as causation. Just because a channel shows up in the report doesn’t mean it drove the outcome.
And if you slash the wrong input, you could kill something that was actually working. That’s the long-term damage most teams don’t see coming (or keep missing).
“AI is going to become a lie detector. It will show where the correlation falls apart.”
That’s the shift GTM leaders need to prepare for.
Fast.
Missed Part 2? Rewatch it on LinkedIn.
If you like this content, here are some more ways I can help:
Cheers!
This article is AC-A and published on LinkedIn. Join the conversation!
