012: Embedded, Centralized or Hybrid: Structuring Your Data Team to Be Incredibly Effective w/ Lydia Monnington

May 13, 2025

Lydia Monnington, Data Science Lead at Google DeepMind with a rich background at Meta and Ocado, dives into the optimal structuring of data teams. She discusses the evolution from centralized to embedded teams and the perks of a hybrid model. Lydia emphasizes building strong relationships between data professionals and business stakeholders, plus shares practical tips for managing diverse roles like data engineers and analysts. Her insights aim to transform data organizations into effective decision-making engines.

Ask episode

AI Snips

Chapters

Transcript

Episode notes

ANECDOTE

Lydia's Data Journey Begins

Lydia first fell in love with data at Citigroup by modeling future profitability for care homes to evaluate investment risks.
She combined domain knowledge and data to answer important business questions clearly and effectively.

INSIGHT

Centralized Team Pitfalls

Centralized data teams build efficiency and best practices but risk becoming bureaucratic data factories.
Without deep stakeholder relationships, they struggle to prioritize and deliver impactful insights effectively.

INSIGHT

Embedded Teams' Double-Edged Sword

Fully embedded data teams gain domain expertise and trust but risk inconsistent tooling and duplicated efforts.
Shadow data teams can fragment data sources, causing inconsistent business answers and inefficiency.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

As a data leader, one of your most crucial decisions is how to structure your team. Should you centralize for consistency, embed for domain expertise, or find a middle ground?

In this episode of More than Numbers Live, Ollie Hughes sits down with Lydia Monnington, Data Science Lead at Google DeepMind, to explore this perennial challenge facing data organizations of all sizes.

With experience leading data teams at Meta, Ocado, Stuart, and now Google DeepMind, Lydia brings practical insights on what works (and what doesn't) when organizing data professionals. She breaks down the evolution of data team structures and provides actionable advice on making the most of your talent—regardless of your current setup.

In our conversation, Lydia shares:

The natural evolution from centralized to embedded teams and the pitfalls to avoid in each model
Why a matrix/hybrid model often delivers the best results and how to successfully implement it
Practical approaches to managing specialized roles like data engineers, analysts and scientists
The critical importance of establishing strong relationships between your data team and business stakeholders

Whether you're managing a small centralized team, leading embedded analysts, or navigating the complexities of a hybrid structure, this episode offers valuable frameworks to make your data organization more effective.

What you'll learn

[01:10] Lydia's journey from financial modeling at Citigroup to data science at Google DeepMind.
[03:45] Inside DeepMind's data challenges: from fusion to forecasting typhoons.
[05:29] How a 15-person central team supports DeepMind's data-intensive organization.
[07:32] Common misconceptions about "perfect" data in big tech companies.
[10:56] The evolution of data team structures: from centralized beginnings to embedded specialists.
[12:08] Why centralized teams often become bureaucratic "data factories" as organizations scale.
[15:00] The potential pitfalls of fully embedded teams: inconsistent tooling and duplicated work.
[18:30] Matrix/hybrid structures: getting the best of both worlds with the right relationships.
[22:20] Practical steps to move toward a hybrid model from either extreme.
[25:40] How to approach structuring specialized roles like data engineers vs. analysts.
[29:24] Lydia's career-defining advice: focus on outcomes and the "why" behind your analysis.