Do You Actually Need a Data Warehouse?

The symptom that justifies it

Not data volume — most businesses never have a volume problem. The trigger is questions that span systems: revenue by marketing channel, margin by customer including delivery cost, lifetime value by acquisition source.

Each of those requires joining data that lives in different places, and if someone is doing that manually every month, that is the case for a warehouse.

What it actually is

A separate database that holds copies of data from your operational systems, structured for asking questions rather than for running transactions. Reporting queries hit it rather than your live systems.

The immediate practical benefit is often overlooked: heavy reports stop slowing down the systems people are trying to work in.

Cheaper steps first

1. Better reporting from the source system. Many businesses commission a warehouse to produce reports their existing system could already produce.
2. Scheduled exports into one shared database. A small step that answers most cross-system questions without a formal warehouse.
3. A single spreadsheet assembled automatically. Unfashionable, effective, and enough for many businesses.

Work through those before commissioning anything, because each is cheaper and one of them usually suffices.

What it costs to run

• Build: £20,000–£60,000 for a first warehouse over a handful of sources
• Hosting: modest for typical business volumes, often tens of pounds a month
• Maintenance: pipelines break when source systems change — budget for it
• Ownership: someone must own definitions, or every number becomes disputed

Definitions are the hard part

The technical work is routine. The difficult part is agreeing what “customer”, “order” and “active” mean across departments that have quietly used different definitions for years.

Expect that conversation to take longer than the build, and treat it as the valuable part rather than as an obstacle.

Build it incrementally

Start with the two systems that answer your most-asked question. Prove it, use it, then add. Warehouses built comprehensively before anyone asks a question tend to be comprehensive and unused.

Frequently asked questions