Star schema
Fact table in the middle, dimensions around it. Surrogate IDs via IDENTITY (or SERIAL / AUTO_INCREMENT on other databases).
Data warehousing without a designer
Star schema, slowly changing dimensions, incremental facts — as code you can test. Not a second BI canvas.
A warehouse is dimensions around facts, history you can explain, and loads that run again without doubling the grain. SSIS and ADF will draw that on a canvas. ETLBox writes it in C#: CreateTableTask for the schema, lookups for the surrogate keys, DbMerge for Type 1, new rows for Type 2, incremental facts from the last loaded date.
The star schema does not care about the tool. The expensive part is changing a dimension six months later when the only artifact is a package nobody wants to reopen. In ETLBox the load is a network you can unit-test and put in CI.
.dtsx or ADF canvas as the source of truthNot a methodology slide. A SQL Server star schema — customers, dates, orders — loaded from OLTP with ETLBox. The full C# is in the article and on GitHub.
Fact table in the middle, dimensions around it. Surrogate IDs via IDENTITY (or SERIAL / AUTO_INCREMENT on other databases).
Overwrite when history does not matter. Insert a new version with valid-from / valid-to when it does.
A consistent time grain for every fact — days, weeks, fiscal periods — not a GETDATE() in the query.
Read the last loaded date, pull what changed, land it. A full reload is a last resort.
The complete load — including error handling — is Building a Data Warehouse with ETLBox. Demo repo: StarSchema in etlbox.demo.
Commonwealth Bank of Australia, Deloitte, Health Catalyst, SEW Eurodrive, Calpine, FASTEC. Warehouses that have to survive the next dimension change, not the next designer.
Build the star in C#
A trial key unlocks the full library. NuGet without a key is limited to 5,000 rows per data flow. Coming off SSIS for the warehouse? That path is documented too.
Get a free trial key Read the walkthrough A C# replacement for SSIS