On the flight home from Martech World Forum in London, I did what any sci-fi fan waiting for Denis Villeneuve's final Dune film would do. I read Dune Messiah. As if my head didn't have enough to process after two days of Martech talks. 😉
Call it kismet, but Princess Irulan wrote this:
"Empires do not suffer emptiness of purpose at the time of their creation. It is when they have become established that aims are lost and replaced by vague ritual."
Replace "Empires" with "Martech stacks" and read it again.
Is my passion for sci-fi now forever tainted by Martech? Apparently, yes 😐
Last year I shared a smörgåsbord of slide photos and let them speak for themselves. This year the program was about getting past the hype. So, ROI, value, and finding out what your stack actually contains.
And a lot of the speakers found themselves describing the problem Irulan wrote about.
Matthew Niederberger
Last year's recap, if you want the smörgåsbord ☝️
Schrödinger's stack
Juan Mendoza, CEO of The Martech Weekly, opened the event with an analogy. He called it Schrödinger's stack. The people who own a Martech stack can't really tell you what state it's in. Whether a capability is there and working, missing, or failing without anyone noticing, you only find out when someone opens the box.
His own research backs that up.
In The Martech Weekly's Enterprise Martech Outlook Survey 2026, 40.2% of enterprise Martech leaders can't demonstrate a clear, measurable contribution from Martech to financial objectives or accepted proxies. Most of that group, 33.9%, say the contribution is "accepted but not directly measurable."
And the chart files them under "Faith-Based Contribution" (yes, that's the actual label, I couldn't have made it up 😅).
Another of his slides listed what holds enterprise teams back.
Two lines could have come straight out of Dune Messiah.
One was "Good decisions from a decade ago," leaving teams "Stuck with a stack that's holding the business back." The other was "Philosophical ambiguity," where "No one is clear on what the Martech stack is supposed to achieve."
Nobody signs for a CDP without a purpose. A few reorganizations later, the people who had it have moved on. But the renewal still comes around every year.
One caveat.
It's TMW's own research, presented at TMW's own event. And some of the cohorts behind the splits are small, so I'd call them directional.
Shoehorning
On day two, Lucy Spooner, Head of Customer Intelligence at Fortnum & Mason, shared the stage with Sonal Goyal, founder of Zingg AI. Lucy walked through how Fortnum replaced, or rather rebuilt, its single customer view.
The old one was held by a third party. Lucy described it as a black box, where "whenever numbers were jumping around we didn't have the ability to go and deep dive." When Fortnum changed its till systems, the team "ended up just shoehorning our new data into our old formats." And the customer IDs were rebuilt every night, "which meant that everybody would get a new URN overnight."
Try running a control group on that. 🤦🏼♂️
Each of those was a sensible call when someone made it. Years later, it was simply how things were done.
The fix started with discovery. "I can't stress enough that phase one was so important," Lucy said, meaning the quality of the data, not just the fields and the volumes. They built the single customer view in-house, moved to a persistent ID, and connected it to their email tool so everyone worked from the same numbers. She mentioned presenting their data to the board the day before, then added:
We wouldn't have done that a few years ago because we wouldn't have felt confident in it.
Shruti Ahuja, Head of Data & Digital Transformation at Virgin Media O2, covered how rituals survive a migration. One of her lessons:
If old tools remain unofficially available, teams will revert during stress, maintaining double run-costs.
I mean, that's paying twice with better intentions 😉
And this one:
Elegance in enterprise architecture is measured by what you remove.
Matthew Niederberger
I wrote about the double spend two weeks ago ☝️
Upkeep
Irulan covers lost purpose. Entropy covers the other side of it, which is upkeep.
A stack that nobody maintains decays even while the dashboards still load and the invoices still get paid. You rarely get one big failure. You get a slow pile-up, one workaround and one extra tool at a time.
Matthew Niederberger
More on entropy and Martech's Law
Fortnum's rebuild was as much maintenance as it was a build.
The data and analytics teams met twice a week, and the Director of Customer and the Director of IT sat in a weekly stakeholder meeting. Lucy called the single customer view "a continual journey," and they're still migrating it and adding data sources.
Keeping track of a stack starts with knowing what's in it, what each part costs, and when each contract comes up for renewal. I built Martech Stack Builder for exactly that. You map the stack, its vendors and its costs in one place, and contract and SLA management is now in open beta. End plug 😉
Stepping aside
Maintenance is also a people job. Shruti put it plainly: "Transformation is Fundamentally a People Challenge," because "human behavior, trust, and habits dictate final outcomes."
Pete Markey, CMO at the University of Nottingham, made a related point in his conversation with Juan. He was talking about speeding up internal maturity and building skills. As a leader, he said, you sometimes just need to step aside and let the people who work for you do the job. His test for a use case was simple too, whether it adds value and makes the team faster and smarter.
Stacks get defended by the people who built them.
To be fair, that's human. Whoever chose the current CDP has every reason to keep it looking like a good decision.
Retiring a few rituals
Tony Byrne, Founder & President of Real Story Group, drew on his new book, MarTech 2030, for his keynote.
He started by admitting he's wary of future-gazing altogether. He still remembers the keynote that sold Second Life as the future of marketing, and later the NFT ones. His words: "people don't always hold futurists accountable".
Then he named a few.
He called the big analyst firms and their quadrants "typically not your friend," which, having just written about the Forrester Wave halo, I didn't mind hearing. His core advice was to "lighten at the edge".
Cap or cut spend on engagement platforms and "allocate those resources towards building your core", the way HSBC did. Large suites, in his view, "want to start at the edge and vacuum into their infrastructure". Nail, head... just spot on!
Jasmin Guthmann, Community Chair at MACH Alliance, put the capability problem in two sentences:
AI capability is what your tools can do. AI operability is what your enterprise can absorb, govern, scale and turn into outcomes.
Her boardroom checklist for any AI or Martech decision starts with "Can it interoperate?" and asks whether a new tool "Connects cleanly, or creates another island?"
The next empire
This is where I got a bit uneasy. Almost every prescription at the event pointed toward a new center of gravity.
Matt Pilgrim, EMEA Partner Development Lead for Adtech and Martech at Snowflake, explained data gravity. Applications, CDPs and AI models "increasingly come to the data not the other way around".
His example was Maya, a valuable customer who just bought skis and keeps looking at boots. The AI sends her 20% off. It doesn't know she has an open claim with the service team.
In Matt's words:
Every fact that AI used was accurate, but a good prediction is not automatically a good business decision.
Matthew Niederberger
What happens when two systems disagree about Maya ☝️
Data gravity was also the topic of the roundtable I hosted, and honestly, the discussion mostly repeated what Matt had shown on stage.
James Taylor, Head of Data & MarTech Platforms at HSBC, made the ownership case with a slide that simply said:
Own your data, or your vendors will.
His list of outcomes for HSBC's own platform included "No vendor lock-in" and, my favorite, "Happy CFO". 😎
Another slide made the long-term case: building your own tooling needs sustained investment, and the payoff in total cost of ownership comes eventually.
Fortnum is doing a smaller version of the same thing. The core is built in-house, and Hightouch is now being implemented for activation, which is pretty much Tony's lighten-the-edge advice in practice.
Everyone on stage worried about somebody else's lock-in.
Tony pointed at suites and HSBC at vendors in general. Snowflake flagged "Model family lock-in" on a slide where the recommended home for your context was a governed enterprise foundation, which is what Snowflake sells. It was also the event's global ecosystem partner, for what it's worth.
Look, I don't think any of them are wrong.
Jasmin made the point that groundwork laid today will most likely still be around in ten years, and I agree with her. Martech moves fast, but implementations have long shelf lives. Some of the warehouse builds going live this year will still be running in 2036.
Where I'm unsure
Is the warehouse the next empire?
The data is centralized, access is decentralized, and costs scale with growth. That makes it useful, and also a candidate for the next single point of failure.
I don't know yet which of those wins out. Yet, I will remain positive.
Forward deployed engineers came up in the session by Nate Wardwell, Manager Analyst Relations and Partner Marketing at Hightouch. He described them as a service Hightouch provides, and one that's helping companies.
My question is a different one.
Isn't that what independent consultants do?
And if a vendor has to embed its own engineers before customers get value, what does that say?
To me it points at, and strong reaffirms, the distance between what the tools can do and what the companies buying them can absorb. That distance is Martech's Law.
Full disclosure, I'm an independent consultant, so I have a stake in the question. Understaffed, underskilled, or tools that are still too complex? No idea yet, maybe all three.
And then there's that HSBC slide.
Last year I argued that composable stacks look cheaper in year one and cost more by year three, once the operational work moves in-house. HSBC is describing the opposite curve for building your own tooling, expensive first and better later.
To be honest, the two might describe different things.
Or it really is a waiting game, and three years is too short to judge. When I wrote about vibe coding, I said building hands you the pager. HSBC's betting that carrying it pays off.
Matthew Niederberger
The year-three argument
Anyway.
Irulan's warning was about empires once they're established, not while they're being built. Most of the stacks discussed in London are still being built, or rebuilt. So in five years, who checks whether anyone still remembers what the new one was for?
Thanks to Juan and the whole Martech Weekly team for having me again.
Start building your Martech Stack designs today 👇🏻
Discussion