Do you need a data warehouse?

You have three or more systems that hold related numbers — say billing, inventory and a CRM — and someone combines them by hand every week or month.

Two departments regularly bring different figures for the same measure. Revenue, active customers and stock on hand are the usual suspects.

Reports slow down or lock the operational system when they run, because they query it directly during working hours.

You want history that your source systems do not keep, such as what a customer’s credit limit or a product’s price was last March.

A warehouse is not a dashboard, and buying one does not produce insight by itself. It is the foundation the reports and forecasts sit on. It is also not necessarily expensive: for many mid-sized businesses, a well-designed Postgres or BigQuery warehouse costs less per month than a single additional software licence.

The technology is the easy part. The real work is agreeing definitions — when an order counts as revenue, what makes a customer “active”, how returns are treated — and writing them down so every table follows them.

Then the data has to keep arriving. A warehouse fed by a manual upload once a month quickly becomes another stale spreadsheet. It needs scheduled, monitored pipelines from each source system.

If nearly all of your data lives in one system with good built-in reporting, and nobody is stitching exports together, a warehouse can wait. Start with clear reports from that system, and revisit when a second or third source appears.