300+ Data Products Someone Is Responsible For

KB Group, a member of the Société Générale international financial group, runs one of the largest data estates in Czech banking: hundreds of thousands of tables spread across more than thirty data warehouses.

Over the years the bank’s Dawiso environment grew from data models into a picture of the whole data estate: warehouse tables, reports, dashboards and the flows between them. The information was there, but it was spread across the environment. No one could point to one place and say this part belongs together, and this person is responsible for it.

​

The Client

KB Group, a member of the Société Générale international financial group, is a major financial services company based in the Czech Republic. It manages an extensive data ecosystem and develops data and analytics solutions for over thirty financial institutions worldwide.

​

The Problem

With roughly 370,000 tables across more than 30 data warehouses, finding an object was not the hard part. Teams needed to know which datasets are actually meant to be consumed, who is accountable for them, who would be affected if one broke, and whether their personal data had been reviewed. The bank also had a precise idea of what a data product is in its world, including its own classification, roles and confirmation steps, and no standard catalog supported it out of the box.

​

The Solution

Data products are defined and maintained in the bank’s own systems. Dawiso visualizes them and makes them usable: one object per product, with the technical assets it is built from, including the ones that live outside Dawiso, so the record of a product is complete rather than partial. Around that we built the parts the initiative actually depended on: business lineage between products, business terms aggregated from the assets they are attached to, data quality rules, lifecycle and ownership, an interface where owners confirm their consumers, and anonymization of personal data outside production. All of it was customized for the bank rather than configured from defaults. In practice that meant extending the platform’s metamodel with 132 object types, 571 attribute types and 167 relation types built specifically for the bank - about a quarter of the metamodel running in their instance.

​

Dawiso: From Scattered Metadata to Data Products With Owners

Dawiso is a single workspace that links business semantics with technical metadata. At Komerční banka it packages the parts of the data estate that belong together into data products: one screen per product, with a named owner, the assets it is built from, the teams that depend on it and its privacy status.

​

Main features of the solution:

Customized solution

The whole data product application was built to the bank’s own specification: its own classification, its own roles, its own tabs and its own validation steps. Nothing was pushed into a generic catalog template.

​

One Screen per Data Product

Each product opens as a single screen. The description arrives from the bank’s systems and can be enriched in Dawiso. Tabs split the rest so that every tab answers one question instead of one long page answering none.

​

Business Lineage Between Products

Where products depend on each other, each one shows which products feed it and which ones consume it. Source and consumer lists are exportable and carry ownership, so the conversation is about responsible teams rather than table names.

​

Consumers Confirmed, Not Assumed

Owners confirm each consumer of their product with one click. A single button then opens a pre-filled email to the owners of every confirmed consumer, so announcing a change or an outage takes seconds.

​

Anonymization on the Product

Where a product contains personal data, the owner marks it on the product itself and the data is anonymized outside production. Test environments stay usable without exposing customers, in line with GDPR.

​

The Impact

The bank now runs 300+ data products, 300+ of them in operation, holding over 30,000 recorded relations to schemas, data models, reports, business terms and to each other. More than 99 percent have a named owner and 326 are linked to a specific squad, so every product leads to the team that builds it. The 2,327 dependencies between products show who is affected by an outage before it happens, and 15,000+ decisions on personal data have been made by the people responsible for it.