The Stacked Data Podcast

Roman leads analytics for marketing and media at CarWow, and he's not interested in a team that just ships things. His argument is that most analytics teams get stuck answering questions instead of shaping decisions, and fixing that has almost nothing to do with better tooling. He walks through why he'd rather kill the ticketing system than keep it, why mandating AI adoption across a team never works, and why using AI on ninety five percent of his team's code still hasn't tripled their output. We cover: * Why shipping a dashboard or a pipeline isn't the same as actually driving business impact * Why he'd rather scrap the ticketing system than let his team just take requests * His risk-based framework for when a system needs to be bulletproof, and when moving fast matters more * Why mandating AI adoption across a team doesn't work, and what he did instead * Why using AI on ninety five percent of his team's code still hasn't tripled their output, and what that reveals about where the real bottleneck is

Rich leads the data practice at AND Digital, and he's not buying into the "AI's coming for your job" panic. If anything, he believes everyone else in the business is about to need the exact skills data leaders already have. His argument is that data's moved through three distinct eras: warehousing, then impact and value, and now the agentic era, where the gap between having an insight and acting on it has basically closed. And this, he says, is the time for data leaders to step up and own that space. We cover: * Why Rich thinks data leaders are best placed to own the agentic era * His three-eras model of data, and what's actually different this time round * The paradox of modern data teams: everything's changed, but also nothing has * Why AI's real bottleneck is context, not intelligence, and what that means for semantic data layers * His framework for building a business case a CFO will sign off on, including why joint sponsorship matters

Phil Goddard leads a 70-person data team on FDJ United's Sportsbook platform — one of the most technically demanding platforms out there, processing prices, bets and risk across thousands of markets in sub-second timeframes. Rather than buying a vendor platform, Phil built FDJ's in-house, and we dig into why: what actually makes something a "data product," where data mesh goes wrong, and how hub-and-spoke fits as the transition stage most teams are really in. We cover: * Why a real-time sportsbook platform is one of the most technically demanding things you can build * Why Phil built FDJ's data platform in-house instead of buying one * Phil's four-part test for what actually makes something a "data product" * The three failure modes he's seen most often in data mesh adoptions * Why AI is "fundamentally an amplifier," and what "governance debt" means for teams trying to get ahead of it

Jose Garcia didn't just grow Skyscanner's data organisation from 13 to around 60 people — he made sure impact grew with it, across the data science, marketing and ad tech teams he leads. We unpack the challenges of prioritising the right problems instead of the most interesting ones, building leadership layers that don't bottleneck through him, and why he thinks great data science has never really been about the models. We also get into how AI tools like Claude are beginning to change the skills, hiring decisions and day-to-day ways of working inside modern data teams. We cover: * How Skyscanner's data org grew from 13 to 60 without losing focus * Why business impact is harder to prove than most data teams expect * Jose's method for prioritising the right problems, not the interesting ones * What actually changes about leadership once a team crosses a certain size * How AI is reshaping who data teams hire and what they're hired for A practical conversation about data science, leadership, scaling teams and the impact of AI.

Monzo went from zero rules to one of the most opinionated data architectures in fintech — and cut warehouse costs by 40%+ doing it. Bruno Campos, Analytics Engineer at Monzo, joins Harry to unpack the rebuild of Monzo's entire data platform — tens of thousands of dbt models, hundreds of teams, one BigQuery warehouse, and a migration that's still only 30-40% done. Bruno breaks down Monzo's new "OOM" architecture: four fixed layers, data contracts between teams, and an internally-built tool (ModelGen) that turns a YAML file into standardised SQL. The result — faster builds, easier onboarding, and serious cost savings that show up on the GCP bill. They also get into why "data" as a discipline is still, in Bruno's words, "incredibly immature" — and what it'll take to fix that. In this episode: * Why Monzo moved from a free-form warehouse to a heavily opinionated one * The four-layer OOM model: landing, normalized, logical, presentation * Interface models — Monzo's version of data contracts between teams * How ModelGen automates standardisation across 100+ teams * The real cost and time savings from re-architecting at scale * Bruno's take on AI's role in analytics engineering — and why the job isn't going anywhere