Glossary
What is a data warehouse?
A data warehouse is a centralised repository storing large volumes of structured historical data from multiple sources, optimised for querying, reporting and analysis rather than for transactions.
Why a separate system
Production databases are optimised for transactions: writing and reading individual records very fast, reliably, under load. Running a query that scans two years of history across every table is exactly what they are not built for, and doing it can degrade the service.
A warehouse is optimised for the opposite: large analytical queries over historical data, usually with columnar storage that makes scanning many rows of a few columns efficient.
The separation also means analysis cannot break production, which matters more than the performance argument in most organisations.
Warehouse, lake and lakehouse
- Warehouse. Structured, schema-defined data, cleaned before loading. Reliable and less flexible.
- Lake. Raw data in native formats, structured on read. Flexible, and prone to becoming unusable without governance.
- Lakehouse. An attempt at both — lake storage with warehouse-style structure and guarantees on top.
- The choice matters less than whether anyone maintains the definitions, which is where all three fail identically.
What makes one useful
Not the technology. Warehouses fail for organisational reasons: nobody owns the definitions, so two teams compute revenue differently; transformations accumulate undocumented; and the same metric has four values depending on which table you ask.
The things that actually determine whether a warehouse is trusted are a documented definition for every important metric, tested transformations, and someone accountable for the layer everyone else builds on. Those are unglamorous and they are the whole difference between a warehouse and an expensive archive.
Common questions
Do I need a warehouse for marketing analytics?
Once you need to join data across platforms, yes — reconciling several vendors' exports in spreadsheets stops working quickly. Below that, platform reporting is usually sufficient.
Warehouse or lake?
A warehouse for defined, repeated analysis; a lake where the questions are unknown and the data is genuinely unstructured. Most marketing use cases are the former.
More in analytics and data
Where the numbers are stored, shaped and read.