Term
Data warehouse
A separate store for analytical data from several systems, enabling reporting without loading operational apps.
Updated
What it means
A data warehouse is a separate database into which data is copied on a schedule from operational systems - the books, CRM, the warehouse, the webshop - and reshaped into one form fit for analysis.
The reason is simple: an operational system is built to write individual documents, not to total them across years. Large queries would slow it down, and it overwrites the historical shape of the data anyway.
A data warehouse is neither a data backup nor an archive. It does not keep documents in their original form as evidence but reshaped data for analysis, and if the source system is lost the books cannot be rebuilt from it. It therefore takes none of the company’s duty to retain documents and accounting records.
How it works in practice
A warehouse earns its keep with several sources and a need to compare long periods - several countries, several shops, several companies in a group. Business intelligence tools and reporting then sit on top of it.
A smaller company gets by with a scheduled export and one integration. Building a warehouse pays only once manual consolidation costs more time than running one would.
For a group of companies we deliver each company’s accounting data to the warehouse after the month-end close. To make it add up, we map each company’s sub-accounts onto a shared structure based on the chart of accounts. Without that, the same kind of cost would sit on a different account in every company and the totals would not agree.
What to watch out for
A warehouse does not reconcile data by itself. Until it is agreed what a customer is, what revenue is and which date it belongs to, all you get is faster contradictory reports. Defining the terms comes first, SQL second.
Personal data slips into a warehouse easily - customer names, salaries, contact details. That creates one more place covered by GDPR, with its own access rights and erasure periods. What the analysis does not need should stay out, and the remaining data about people can be pseudonymised as it is loaded.
Need the term explained on your own situation? Consulting from €55 / hour excl. VAT, depending on topic and complexity.