The Turf Wars Are Over. Time to Cross-Train
Follow into
Save into
There was a time you could win as a pure stylist, and plenty of people still can. A great boxer is still a marvel in a boxing ring. A great wrestler still dominates on the mat. Those crafts never stopped advancing, and the specialists who give a lifetime to them keep getting better at exactly what they do. None of that has changed1.
What changed is that a new game showed up. When a pure boxer stepped into a cage under mixed martial arts rules, his hands were suddenly not enough. He suddenly had to think about defending takedowns and getting kicked. The lifelong wrestler who’d never thrown a punch got picked apart on the feet. Neither of them was bad at his craft. They were now playing a different game from the one they’d trained for, and it required a different blend of skills and techniques.
Practical Data Modeling is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
Say that part plainly, because it’s the spine of everything that follows: different games require different styles. The boxer isn’t wrong to box. He just can’t box his way through a sport that rewards boxing plus much more.
Data has been running its own version of the style war for thirty years. Relational versus everything else (see Chris Date’s books for some hilarious commentary). Kimball versus Inmon. Data Vault has its infighting drama. One Big Table cuz joins are bad. The “knowledge” crowd has lately squared off against everyone over AI. I see many of these arguments online, shake my head, and ask in the spirit of Rodney King, “Why can’t we all get along?” We’ve spent an enormous amount of energy arguing about which style is supreme, and right now, with the ground shifting under all of us, the arguments still continue. That kind of noise usually signals fear more than confidence.
I’ve been in this game for a while, and I have deep respect for what various people have built and for the practitioners and innovators still carrying it forward. So I’m saying this from inside the tent (or cage?), not throwing rocks from outside. The turf wars are boring, childish, and petty, and they are eating up time we do not have while we’re staring at a massive tsunami coming at us.
We Were Arguing on the Wrong Axis
Most of these methodology fights were never really about which approach was correct. They were about which approach was correct for a particular game: a world built entirely around human data consumption. Dimensional models, relational models, normalized warehouses, ontologies, taxonomies, etc. All of them answer the same underlying question: how do we organize information so people can find it, trust it, and reason about it?
They’re still good answers to that question. Dimensional modeling still wins the analytics game in terms of mindshare. A well-built data warehouse still does exactly what Bill Inmon designed it to do (“A subject-oriented, integrated, time-variant, and non-volatile collection of data in support of management’s decision-making process”). None of that craft got worse, the same way boxing didn’t get worse the day the cage opened. The specialists are right to keep sharpening it.
What’s different is that a new game showed up, and it has its own rules. The world we’re building for now includes machines as first-class consumers. Agents are querying your data. Models are being grounded in it. Pipelines are increasingly written, monitored, and repaired by systems that don’t read your wiki and don’t care which 1990s methodology won the war. When your most demanding consumer is a probabilistic system stitching together context on the fly, “which dimensional pattern is purest” stops being the only thing worth asking. The new questions sit right next to the old ones: Is this data legible to a machine? Is it trustworthy enough to act on? Can a system reason over it without a human holding its hand?
Those are questions about fundamentals. No single methodology owns the answers.
Fundamentals Are the Foundation Everybody Shares
Every camp got something right. Inmon and Kimball both still matter, and it has nothing to do with which of them “won.” They matter because they were wrestling with real, durable problems such as grain, entities, relationships, how meaning attaches to data, how you make information trustworthy at scale. Data Vault gave us a way to absorb relentless change without losing our minds. The knowledge folks are right that semantics and meaning matter more than ever in a world of machine reasoning. Every one of these traditions found something true, and each is still the best tool for some game you’ll actually have to play. Those problems didn’t disappear when the cloud arrived, and they’re not disappearing now. If anything, machines are about to stress-test our fundamentals harder than any human ever did.
That shared foundation is the thing to stand on. It’s also exactly what I mean by Mixed Model Arts.
What Mixed Model Arts Actually Is
Mixed Model Arts is not a new camp built to beat the old ones. If it were, I’d be guilty of the exact thing I’m complaining about, planting one more flag and daring everyone to come fight over it. That’s not what this is about. It’s about accepting that the world is dynamic, complex, and you need evolve with the times.
It’s the practice of a good mixed martial artist. You study every discipline. You become well-rounded and agnostic to any particular approach or dogma. You take what works. You invent your own approaches. You respect the lineage of each technique you borrow, and you stop pretending any single style wins every game.
None of this means abandoning a specialty. Every great mixed martial artist has a base (wrestling, striking, jiu-jitsu) that they lead with. They don’t dissolve into something formless. They build outward from their strength so they can hold their own wherever the fight goes. A modeler works the same way: keep your base, then round out your game for the problem in front of you. Dimensional thinking where dimensional thinking wins. Normalization where it wins. Vault patterns where change is constant. Graph and semantic approaches, where meaning and relationships carry the load. ML/AI, where you need to process unstructured data for classification, prediction, and generative data. And increasingly, designs that serve machine consumers right alongside human ones, because that’s the game we’re actually stepping into.
Nobody is excluded. Everyone is invited to participate. Take everything the field has earned over sixty-plus years, call it good, and build the next thing on top of it.
Stop the Stupid Comparisons
The comparison game of “my camp versus your camp” is a status game dressed up as a technical debate. The identity politics of data needs to end. Asking whether Kimball beats Inmon is like asking whether a boxer beats a kickboxer: the only honest answer is “in which game?” I just saw someone in the knowledge community bash data warehouses, calling them old hat and saying graphs are far superior for agents. Maybe, maybe not. The argument seemed more like a sales pitch for the person’s knowledge graph courses than something a sensible practitioner would implement. Are knowledge graphs worth learning about and using? Certainly, add them to your toolkit and use them where they make sense. Are they the end-all, be-all of the practice of data? No. Nothing is. Strip away the tribalism, and the argument barely means anything, just like similar arguments. Nobody building something that matters in 2026 is focused on these tribalistic comparisons. They’re figuring out how to make data trustworthy and usable for both humans and machines, and they’re pulling every good idea they can find, regardless of whose flag it flew under. That’s exactly how a serious mixed martial artist trains. Borrow relentlessly, stay loyal to none of it, get good.
The old world is genuinely crumbling. People are clamoring for a piece of a world being reshaped in real time, and the camps that respond by digging trenches around their methodologies will be left behind by those that figure out what comes next. The reckoning is real, and it does as much clearing as destroying. The dead weight burns off, and the people willing to build finally get room to work.
Stand on the Foundation. Build What’s Next
We are now designing for humans and machines. That’s it. That’s the whole headline.
So put your energy where it counts. How do you model data so an agent can reason over it without hallucinating its way into a bad decision? How do you encode enough semantics and context that a machine understands not just the shape of your data but its meaning? How do you build trust and lineage into systems where the consumer can’t pick up the phone and ask what a column means? How do you keep the fundamentals (grain, entities, relationships, attributes, time, etc) intact when half your pipeline is being authored by something that never worked in your business as an employee, has no tacit knowledge of your operations, and will eagerly work on any task?
Those are hard, open, genuinely interesting problems, and they’re worth the talent in this field. They beat relitigating a 2003 blog argument every single time.
So here’s the invitation, and I mean it. Keep your craft. Keep sharpening it. But know which game you’re in, and bring the right blend for it. Honor every tradition that got us here, thank the people who built it, and turn to face what’s actually in front of us. The field has had enough holy wars. What it needs now is people willing to cross-train, stand on the foundation we all share, and build for a world of humans and machines.
That’s Mixed Model Arts. The turf wars only ever made sense when there was a single game to win. And even then, it was silly fixed-mindset theatrics. There are many games now, more than any one style can cover, and the whole sport is wide open for the people willing to build the future.
Even Jake Paul is becoming an OK boxer when he's not committing elder abuse.