Data warehouse
Also known as: enterprise data warehouse, cloud data warehouse
A data warehouse is a central database designed for analysis rather than day-to-day transactions.
It combines cleaned, organized data from sales, finance, marketing, and other systems, so people can run reports and spot trends across the whole business. While warehouses were once expensive projects for large enterprises, cloud services like Snowflake, BigQuery, and Redshift made them affordable for almost any organization.
Data warehouse, explained at three levels
At home
Imagine collecting receipts, bank statements, and bills from different drawers into one organized binder so you can see your yearly spending. A data warehouse is that binder for a business.
At work
A data warehouse gives a company one trusted place for reporting. Instead of pulling numbers from many systems, analysts query the warehouse, which makes reports faster and more consistent.
At home
Business systems like online stores are built to handle many small transactions quickly. Warehouses are built for big questions, like 'what were sales by province for five years?' They often store data by column to make those questions fast.
At work
Data flows into warehouses through ETL or ELT pipelines. Common designs include star schemas, with fact tables (sales, events) and dimension tables (customers, products), and layered models from raw to cleaned to business-ready. BI tools such as Power BI and Tableau sit on top.
At home
Cloud warehouses separate storage from computing, so you can pay for heavy processing only when you need it. In my opinion, that flexibility is why warehouse bills now surprise people: it's easy to run expensive queries without noticing.
At work
Key decisions include modelling approach, workload isolation (separate compute for dashboards and data science), cost governance, access control for sensitive data, and how the warehouse relates to a lake or lakehouse. Open table formats increasingly blur the line, letting warehouses query data stored outside them.
Real-world examples
At home
Lena runs a small online craft shop. She connected her store, payment, and ad accounts to a cloud data warehouse and finally saw which ads led to repeat customers, something no single app could tell her.
At work
Marcus is the finance director at an Ontario manufacturer whose month-end reporting took a week of spreadsheet merging. A data warehouse pulling from ERP, sales, and inventory systems cut month-end reporting to two days.
How it compares
| feature | Data warehouse | Operational database |
|---|---|---|
| Purpose | Analysis and reporting | Running daily transactions |
| Data | Historical, from many systems | Current, for one application |
| Typical queries | Large summaries and trends | Small reads and writes |
| Design | Optimized for reading | Optimized for writing |
Operational databases run the business. Warehouses help you understand it.
For analysts
A warehouse project succeeds when it answers specific business questions. I'd list the ten reports leadership needs most and build toward those first, rather than loading every table and hoping insight appears.
History & origin
IBM researchers Barry Devlin and Paul Murphy described a 'business data warehouse' in a 1988 paper. Bill Inmon, often called the father of data warehousing, published Building the Data Warehouse in 1992, favouring a central, normalized design, while Ralph Kimball's 1996 The Data Warehouse Toolkit championed dimensional 'star schemas.' Vendors such as Teradata, Oracle, and IBM dominated for two decades. The cloud changed everything: Google BigQuery was previewed in 2010, Amazon Redshift launched in 2012, and Snowflake, generally available in 2015, separated storage from computing. In my view, the Inmon-versus-Kimball debate taught a lesson that still applies: the right model depends on how people will actually ask questions.