Data warehouse
A data warehouse is one place in your own cloud where the data from your systems is gathered, cleaned and kept with its history. Your people and your agents ask their questions there, and every figure can be traced to the record it came from. We build it on Google Cloud with BigQuery or on Microsoft Azure with Fabric, and you own it.
This is for you if
- Two reports give two different figures for the same thing, and nobody can say which is right.
- Your systems only show how things stand today, and last year's figures live in old spreadsheets.
- You want your AI tool to answer from your own figures, and each system only knows its own part.
What you leave with
- The result
- One place for your figures, in your own cloud. Your reports, your people and your agents read the same figures, defined once. You decide who may read what, and you can take the data with you.
- History kept
- Data is kept as it was, so you can compare periods and see what a figure was when the report went out.
- Definitions in one place
- Your business rules, such as how revenue is counted or how a payout is matched, are written once in the warehouse and used everywhere.
- Controlled ways in
- Nobody reads the warehouse directly. People and agents come in through MCP servers or a prepared report, and each role sees only its own datasets, within its own limits.
What we do, what you do
| We | You |
|---|---|
| Go through with you which figures matter and how each one is defined. | Tell us which reports you trust and which questions you cannot answer today. |
| Set up the warehouse in your Google Cloud project or your Microsoft tenant, with the tables and their checks. | Create the project or the tenant, through someone with admin rights, and choose where the data is stored. |
| Set up the roles, so each person and each agent reads only what it should, and a log of who read what. | Decide who may read what. |
| Hand over the code and the settings. Gathena's agents build and maintain the warehouse, and every change waits for your approval. | Own all of it, in your own cloud. |
Worth knowing
A data warehouse earns its keep when figures from several systems must be matched. If your figures come from one accounting system whose reports answer your questions, you probably do not need one yet. Then start by cleaning up the records you have, and write down what each key figure means.
Your questions
- BigQuery or Fabric: which should we choose?
- Usually the one that matches the stack you already work in and the place your accounting data lives. Our comparison goes through how data gets in, billing, the built-in AI, the skills each takes and how easy it is to leave.
- How is a data warehouse different from our accounting system?
- The accounting system holds the books and is where entries are made. The warehouse holds copies of the data from all your systems, side by side and with their history, so they can be matched and asked about. Nothing is booked in the warehouse.
- Can our AI tool read it?
- Yes, through an MCP server in your own cloud. Each person signs in as themselves and sees only what their role allows, and every call is written to the logbook.
- Where is the data stored?
- In your own Google Cloud project or Microsoft tenant, in the region you choose; in the DuffBeauty case, the EU. You control access. The model that answers a question runs at the provider you chose.
- Can we leave?
- Yes. The code and the settings are yours, and both platforms can export your tables in open formats.