Latest / MBA Training Data / Data mesh vs. data lakehouse: what CDOs actually need to decide in 2026
Transcript
- 0:00Welcome to the Leaders Insights podcast.
- 0:08Today's episode, Data Mesh versus Data Lakehouse, what CDOs actually need to decide in 2026.
- 0:15CDOs are still debating Data Mesh versus Data Lakehouse in 2026, and frankly, that surprises
- 0:20me.
- 0:21Haven't we had enough time to figure this out?
- 0:23The debate persists because most organizations made their initial bets in 2022 and 2023,
- 0:30based on vendor pitch decks, and now they're living with the consequences.
- 0:34What's changed is that the consequences are visible.
- 0:37You can actually measure where the architectural choice created value and where it created
- 0:41debt.
- 0:42So what does that debt look like in practice?
- 0:45Take a large European retailer I worked with last year.
- 0:47They went all in on Data Mesh, stood up 12 domain teams, built federated governance on
- 0:52paper.
- 0:5318 months later, seven of those domains had no active data products, and the central
- 0:59team was still resolving schema conflicts manually every sprint.
- 1:03The organizational model outran the organizational readiness.
- 1:06And the lakehouse camp isn't doing better?
- 1:10Some are, some aren't.
- 1:11The lakehouse architecture, Databricks and Snowflake, are the two obvious reference points
- 1:16here, and both publish their own performance benchmarks, so take those with appropriate
- 1:21skepticism.
- 1:23Solves the technical unification problem elegantly.
- 1:26One query engine, one storage layer, reasonable cost structure.
- 1:31Where companies stall is governance.
- 1:33They build a beautiful technical foundation, and then discover they have no decision rights
- 1:38framework for who owns what data, who can certify quality, who gets to deprecate a table.
- 1:44That's not a technology problem.
- 1:46So is this actually an either or decision, or have we moved past that framing?
- 1:50The sharp CDOs in 2026 have moved past it.
- 1:54J.P. Morgan's data organization, for example, runs a lakehouse as the physical infrastructure
- 1:59layer and applies mesh principles.
- 2:02Domain ownership, federated governance with central standards, as the operating model on
- 2:07top.
- 2:08The architecture and the organizational design are separate variables.
- 2:12Conflating them is where the real mistake happens.
- 2:15Where does that leave a CDO who's mid-implementation maybe two years into a mesh rollout that
- 2:20isn't delivering?
- 2:21First, audit what you actually have versus what you planned.
- 2:26Most mesh implementations I've reviewed have one or two high-functioning domains and a long
- 2:31tail of theoretical ones.
- 2:33Rather than burning the whole thing down, consolidate.
- 2:36Make those one or two domains genuinely excellent, documented, SLA-backed, discoverable.
- 2:43Then use that as the template for expansion.
- 2:46Visible success within your organization is more persuasive than any architecture diagram.
- 2:52What about the CDOs who haven't committed yet?
- 2:54Are they actually in a better position?
- 2:56Potentially, yes.
- 2:58But only if they use the time well.
- 3:00The organizations that delayed have the advantage of watching others fail.
- 3:05But I see a lot of them now rushing to catch up and making the same shortcuts.
- 3:09The window where delayed entry is an advantage closes fast.
- 3:13Final word.
- 3:14One thing a CDO should do this quarter?
- 3:16Map your data domains against actual business decisions, not technical systems.
- 3:22If you can't name the business decision each domain supports, you don't have a domain.
- 3:27You have a data warehouse with a new label.
- 3:29To go further, Leaders Insights offers strategic training programs, diagnostic tools, and an
- 3:34executive self-assessment designed for CMOs, CDOs, CFOs, and leaders aiming for board-level
- 3:41performance.
- 3:43For everything at MBA-training.com, thank you for listening to the Leaders Insights Podcast.
- 3:49See you tomorrow for another episode.