Centralized vs. Federated data teams is the wrong debate.
Julia Bardmesser
Helping Companies Maximize the Business Value of Data and AI | ex-CDO advising CDOs at Data4Real | Keynote Speaker & Bestselling Author | Drove Data at Citi, Deutsche Bank, Voya and FINRA
I have a confession to make.
I started my data career managing a team that did its best to stay outside of enterprise data management prescriptions.
We even called our database a “datamart”—not because it technically was one, but because it let us bypass the enterprise rules for data warehouses.
Our excuse?
We had to deliver for two trading desks + a quant team, under intense pressure, and the enterprise group just didn’t understand that kind of urgency.
Of course, karma being what it is, I spent the rest of my career firmly on the enterprise side—one of those people who “didn’t understand.”
That experience gave me a front-row seat to one of the oldest debates in data:
Should data and AI functions be centralized or federated?
Each model comes with pros, cons, and plenty of passionate opinions.
I’ve learned the best answer often isn’t one or the other—it’s a smart blend of both. Here’s why.
Centralized Model
Pros
Strong governance and consistency—easier to enforce standards, policies, and privacy compliance
Efficient use of resources—one team, one set of tools, better economies of scale
Easier to build shared capabilities like MDM, metadata, observability, and analytics platforms
Cons
Bottlenecks when demand is high
Harder to get business buy-in—can feel like an external service
Perceived lack of business knowledge or alignment
Federated Model:
Pros
Closer alignment to business priorities and nuances
Faster iterations—no waiting in a central queue
Clear ownership and accountability at the domain level
Cons
Hard to solve for cross-domain challenges, including AI governance
Inconsistent standards make reuse and cross leverage challenging
Duplication of tools and effort
My Take: The Hub and Spoke Model
When done right, hub and spoke blends the strengths of both worlds.
The Hub
Builds shared infrastructure (MDM, lakehouse, lineage tools, observability)
Defines and enforces enterprise standards
Leads cross-domain initiatives and AI platform strategy
Drives data literacy and education
The Spokes
Sit in business lines/functions, owning use cases tied to KPIs
Maintain local domain models and dashboards
Bridge domain experts with centralized capabilities
Feed lessons learned back into enterprise strategy
What matters most is clarity:
Who owns what?
Who’s accountable?
How will you collaborate without creating silos?
If you’ve been on either side of the fence like I have, you know there’s no perfect org chart—only structures that fit your culture and maturity level.
The right model isn’t about where the data sits—it’s about how people work together to turn it into value.
So, what model are you using today? Has it evolved? What’s working (or not) for you?
From Debate to Delivery. Clarity Wins.
I help data and AI leaders move beyond centralized vs. federated ideology and design operating models that actually deliver business value.
With 25+ years leading data, AI, and risk transformation at firms like Voya Financial, Deutsche Bank, Citigroup, and Freddie Mac, I know how to blend governance with agility—turning data strategy into measurable impact.
→ Planning an event or need a practical, straight-talking speaker on data operating models, monetization, AI, or risk? Book me here.
→ Want to see how a hub-and-spoke approach could unlock hidden value in your data? Schedule a discovery chat.
Great illustration. What about when you have multiple external parties involved? This hub and spokes/federated model is on trend in health care data analytics where organizational partnerships allow researchers to access sensitive/hard to find data locked locally while protecting patient privacy.
Global Sales Development Leadership | GTM and AI agent architect | GTM and AI Ops | Servant Leader
1moJulia Bardmesser - Excellent articulation of the trade-offs in data organization. The "Hub and Spoke" model is exactly the approach I've seen deliver the most sustainable value. The key is enabling the "Spokes" without creating chaos. I've found that balance is best achieved when the "Hub" provides a common platform that facilitates a federated governance model—setting enterprise ground rules while empowering domains to execute. This requires a shift to decentralized, domain-driven ownership. When domains like sales or finance truly own their data as a product, you see a significant uplift in quality and relevance. The central team's role rightly evolves to become an enabler of this federated model. This is exactly what we are trying to solve for here at Semarchy. Great post. It all comes back to the clarity you mentioned on ownership and collaboration.
Chief Information & AI Officer | Scaling Enterprise AI with Strong Data & AI Governance | Data Products • Decision Intelligence • Generative AI (RAG/Agents) | Board & PE/VC Advisor | NACD | Healthcare
1moGreat topic for Friday. I see hub-and-spoke as federated by default. The real question is not what model you pick—we’ve all learned the hard way you cannot centralize what is not natural to centralization. The question is how you deploy hub-and-spoke. I’ve built this model four times in large, globally distributed companies. When done well, it unlocks both governance and innovation. When done poorly, the spokes turn into data-copying machines, multiplying inconsistencies at the speed of light. The secret sauce isn’t in choosing central vs. federated—it’s in designing the flows, incentives, and accountability that make the hub and spokes operate as one system.
Build it and they will come! This concept is never true when centrally managing data.
Human Consultant Oasis Blue, EX Test Consultant at Fujitsu, EX Business Consultant at B&Q, WH Smith and SwissAir
1moThank you for framing the problem. This can be applied to leadership models, operations and organisations. The key imo is to let the knowledge and information drive the operation and from that the organisation. When the hierarchical command led organisation drives the operation then trouble will most probably be ahead. 🙏✅♥️