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
Think of gathering receipts, bank statements, and bills from different drawers into one organized binder, sorted by month and category, so you can see your yearly spending at a glance. A data warehouse is that binder for a business. It collects data from many separate systems, cleans it up, and organizes it so people can answer questions about the whole organization.
At work
A data warehouse is a central database designed for analysis rather than day-to-day transactions. It combines cleaned, consistent data from sales, finance, marketing, and other systems, keeping history over years. Businesses care because analysts and dashboards can query one trusted source instead of stitching numbers together from many systems, making reports faster, more consistent, and easier to audit.
At home
Apps like online stores use databases built for many tiny, fast transactions: adding one order, updating one address. A warehouse is built for the opposite: huge questions across millions of rows, like sales by province over five years. To answer those quickly, warehouses usually store data by column, compress it heavily, and split each query across many processors working in parallel, a design called massively parallel processing.
At work
At work, data reaches the warehouse through ETL or ELT pipelines, usually in layers: raw copies of source data, cleaned staging tables, and business-ready models. A common design is the star schema, with fact tables recording events, such as sales, surrounded by dimension tables describing customers, products, and dates. BI tools such as Power BI and Tableau query the warehouse, and platforms include Snowflake, BigQuery, Redshift, and Microsoft Fabric.
At home
Modern cloud warehouses separate storage from computing. Data sits in cheap storage, and compute clusters start only when queries need them, billed by the second or by the data scanned. That's flexible, but it also means one careless query over a huge table can cost real money. In my opinion, that's why warehouse bills now surprise people: expensive queries are easy to run without noticing.
At work
Key architectural decisions include the modelling approach, Kimball dimensional models or Inmon-style normalized layers; workload isolation, giving dashboards and data science separate compute; and cost governance through quotas and monitoring. Access control needs column masking and row-level security for sensitive data. Open table formats such as Iceberg increasingly blur the line with lakehouses, letting warehouses query data stored outside them, which shifts lock-in toward catalogues.
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.