Lesson learned implementing a large-scale data mesh transformation
summary
Three-year Data Mesh at PAYBACK reorients organization (domain-driven design, PO/BO/TO trinity) while migrating on‑prem Oracle workloads to GCP and BigQuery; three core crews (cloud infra, cloud data & AI platform, cloud migration) build a vendor‑agnostic DSL, use Dataplex and DataHub for enforcement and cataloging, and deliver ~22 true data products plus under 1,000 public cloud assets. Key specifics: lightweight lineage extraction, migration team lifts strategic datasets (not full one‑to‑one ETL), cost and naming policies, staged decentralization models, and formal learning paths to drive adoption.
AI-written summary · about
chapters (17)
- 0:00Intro
- 0:14What is Payback
- 2:37Why Data Mesh
- 14:38Choosing the Tamesh
- 15:15Preparatory work
- 15:44Cloud infrastructure
- 18:56CNU Team
- 20:10Data Governance Council
- 22:15Lightweight DSL
- 23:58Naming
- 25:29Problems arise
- 26:11Data duplication
- 28:05Data lineage
- 31:56Case by case
- 32:28Feedback
- 33:50Data product ownership
- 36:32Lesson learned
related talks
44mData Mesh: Lessons from the trenches
50mAmy Raygada | Data Track | Compass AI & Data Summit 2025
48mBuilding a data mesh
31mData Mesh In Practice – How To Set Up A Data Driven Organization
01h 02mShift Data Quality Left with Data Mesh Principles
23mHow To “Data Mesh”: What You Need To Know To Make It Real
43mBoF: Data Mesh Community of Practice: What It Is, Why It's Hard, and How to Help...
36mIntroducing Data Mesh A Paradigm Shift In Analytical Data Management Architecture