Product

Data warehouse

The dependable option for structured finance and operations data. SQL-first, mature governance, predictable cost, and a dimensional model that still reproduces last year's numbers.

Fig. 01 - the warehouse, end to endPick a stage →
The three modelled layers

Stage 01

Sources - most estates have more than they think

The inventory usually turns up systems nobody listed: a legacy instance still feeding a report, a supplier file drop, a departmental database holding something month-end depends on.

ERPRelational sourcesFile feedsAPIs
  • ERP finance, supply chain and HCM modules
  • Relational databases across the estate
  • Extracts from reporting and consolidation tools
  • Supplier and partner file feeds
  • APIs where a direct database route does not exist
  • Statutory reporting feeds where they touch the same data

The idea

Same discipline, expressed in schemas

A warehouse and a lakehouse separate their layers for the same reason. The names differ and the technology differs; the argument for keeping them apart does not.

Staging answers what arrived

A one-to-one landing of each source with load metadata attached, so downstream can always be rebuilt without another extract.

Conformed answers what is true

Typed, deduplicated and agreed, with exceptions visible in a review table and reconciliation counts produced on every run.

Dimensional answers what it means

Facts and conformed dimensions with proper history, built for the questions the business asks rather than the way the source stored them.

Cost is predictable

Sizing is known in advance and scales on a schedule you control, which matters more than peak performance for most finance reporting.

Fit

When this is right, and when it is not

A good fit when

  • Your reporting runs on structured finance and operations data
  • The team is strong in SQL and wants to stay there
  • Predictable cost matters more than elastic scale
  • Governance, auditability and stable definitions are the priority
  • You need last year's report to reproduce exactly, years later

Probably not when

  • You have significant streaming, document or unstructured data
  • Machine learning at scale is central to the plan
  • Volumes are large and spiky, making fixed sizing wasteful
  • A lakehouse would serve you better - we will say so

Tools

Dimensional modellingELTChange data captureExternal tablesPL/SQLSQLPartitioningMaterialised viewsNatural-language queryIn-database ML

Platforms we work on: Oracle Autonomous Data Warehouse, SQL Server, Snowflake and the major cloud warehouses.

Ready to build a warehouse?

We can size and build one from nothing, or take on a warehouse that is already running and bring it up to standard.

Start a conversation