What a data warehouse is, in one paragraph
It is a database designed for analysis, not for day-to-day operations. It holds historical data from many systems in one agreed format so it can be compared. The difference from an ordinary store database is that the store database has to handle an order quickly, while the warehouse has to answer a question about a thousand orders quickly. It is also a tool for running the business on data rather than on impressions from ad platforms.
Why platform totals do not add up to revenue
Each ad platform attributes sales to itself according to its own model and its own time window, so the same transaction can be counted two or three times. It is not a system error but a result of none of them knowing about the others. Resolving such conflicts is one of the main reasons for building a warehouse.
Contribution margin as the decision metric
Revenue minus product cost, returns, shipping and payment costs, and ad cost gives a number you can move budget on. Platform ROAS cannot replace it, because of those costs it only knows the ad cost. Agreeing this definition is the first step of implementation and usually the most contentious, because it touches accounting.
An ecommerce data warehouse and one for a service business
A store has an order with a price and a cost, so the account closes inside the warehouse. A service business has an enquiry whose value appears only after a sales conversation, so the warehouse has to reach into the CRM, and some figures stay estimates until the contract is signed. That changes the scope of the project and we price it differently.
Data warehouse from Kraków, done remotely
Our office is in Kraków, and the project is remote by nature, because the work happens in the cloud and in the client's systems, so we deliver it for companies across Poland and abroad. It is worth meeting live once, when agreeing the definition of an order and of margin, because marketing, sales and accounting all need to be at that table at the same time.
How data gets into the warehouse: ETL
A data warehouse gathers data from different sources in one consistent format. The ETL process pulls it from operational systems such as the store, the ERP and the warehouse management system, then it has to be processed and structured, and finally it lands in an analytical database. That way integrated data from ads, stock levels and orders can be analysed together instead of being kept in separate spreadsheets. Real-time refreshing is rarely needed: for budget decisions a daily rhythm is enough, and that makes it possible to optimise query cost and performance with large volumes of data.