Definition
A data warehouse is a database designed for analysis rather than for running an application. It holds data from many sources, with history, and is queried on demand, usually in SQL. In marketing the typical sources are the analytics export, daily spend and conversions from each ad platform, leads and deals from the CRM, and orders and refunds from the commerce platform.
Two properties separate it from the tools that feed it. It is general: it has no built-in idea of a session, a campaign or a customer, only tables that the team designs. And it is owned: what was loaded stays as it was loaded, regardless of what a platform later changes in its interface, retention or counting.
In practice: loading, modelling and the reports built on it
Work in a warehouse runs in three layers.
Loading. Data arrives through connectors, scheduled exports or a platform's own export, as raw tables that are not overwritten without a record. The guide to loading ad platform data into a warehouse shows why this is harder than it looks: conversions keep arriving for days after the click they are credited to, each platform counts under its own attribution settings, and accounts report in their own currencies and time zones.
Modelling. Raw tables are turned into modelled tables with a defined grain, such as one row per day, channel and campaign, and with keys that let spend find the leads and orders it produced. Definitions of each metric are written here once.
Reporting. Dashboards and analyses read the modelled tables. The comparison of reporting and BI treats the warehouse as the dividing line: a dashboard connected to platform connectors is reporting, and the same tool connected to a joined, defined warehouse table is the presentation layer of BI. The tutorial on moving a Looker Studio report onto a warehouse shows the switch for a report that has outgrown its blended connectors.
Warehouse, CDP and reporting tool compared
The three are often confused because each holds marketing data.
| Tool | Built for | What it is not |
|---|---|---|
| Data warehouse | History, joins across the whole business and shared definitions | A segment builder or a real-time activation tool |
| Customer data platform | Resolving identity, building segments and pushing them to channels | A record of everything the business does |
| Reporting tool | Presenting data as charts and tables | A place to store, join or define data |
The guide to data flow and integrations describes the division that works for most teams: the warehouse as the system of record, the customer data platform or a reverse-sync tool as the system of action. The guide to designing a marketing stack places the choice between them in the wider build-or-buy decision.
A home insurer replaces four blended connectors with one table
Starlingmede Home Insurance reported cost per policy by channel from a dashboard that blended four live connectors: analytics, two ad platforms and the CRM. On Monday mornings, with 25 people opening it, charts timed out, and the blended figure never matched finance's count of policies. The analytics lead loaded daily spend from both ad platforms, the analytics event export and CRM policy records into a warehouse, then built one modelled table at the grain of day and channel, joining policies to spend on campaign keys. After a fortnight of running old and new side by side, the table's 1,140 policies for the month matched finance exactly, against 1,212 conversions in the old blend, and the dashboard loaded in seconds.
Related terms
The GA4Google's event-based analytics platform. export is often the first raw table a marketing team lands, because it holds event-level data the interface summarises. The CRMThe system that stores an organisation's contacts, companies, deals and interactions, and the processes built around it. supplies the outcomes, leads, deals and customers, that spend is judged against. A warehouse is one of the main stores of First-Party DataData a business collects directly from its own customers, visitors and users., which brings the same consent and minimisation duties as any other system holding it. AttributionAssigning conversion credit to touchpoints. becomes a modelling choice once paths and spend sit in the same place, though platform-reported conversions still carry each platform's own settings.
Common confusions
- A warehouse is a dashboard. It stores and joins data; a reporting tool presents it.
- A warehouse makes numbers right. It holds what was loaded; definitions, refresh windows and reconciliation decide whether the numbers are right.
- Every team needs one now. The triggers are joined questions, history beyond platform retention and figures that must reconcile with finance; without them, a well-owned dashboard serves better.
- Warehouse and CDP are rival purchases. They do different jobs; the arrangement to avoid is running both with separate identity logic and segment definitions.
- Loading the data is the hard part. Keeping it complete and consistent as platforms change is the ongoing work.
