DWH design
Data Vault, Kimball, and hybrids. From data model to first vertical in prod.
I build a warehouse from scratch or design a replacement for one that has drifted. I start with business requirements: what reports the regulator needs, what questions the product asks, what freshness is expected. Only then do I open the model.
I pick the methodology by task. Data Vault 2.0 where auditability matters and data comes from dozens of sources. Kimball where the reporting is stable and the BI tool is critical. A hybrid where audit is needed on the core, and marts are cleaner in Kimball.
I design the layers: raw, staging, core, mart, semantic. I write ingestion contracts for every source. Together with the team I ship the first vertical up to a report, so that by the end of month two the regulator or the business sees something working.
- For whom
- Banks and fintech building a DWH from scratch or migrating off legacy Oracle / Teradata.
- We solve
- Reporting is stitched from a dozen marts with inconsistent logic. The regulator asks for lineage — no one and nowhere to show it. New sources take two months instead of two weeks. Calculations fall through the cracks between DWH layers.
- Deliverables
- Logical and physical data model, architecture and methodology documentation, ingestion contracts for priority sources, a working first vertical up to a report, and a 6-month rollout plan.
- Format
- Remote. Optional on-site working sessions every 2–3 weeks.
- Duration
- 6–12 weeks to first vertical in prod.