Sievert SE A data foundation instead of data silos

The need for action was clear, but the approach and the specific technology still had to be explored. Since the existing data management architecture had to be fundamentally redesigned, Sievert did not want to do this without a requirements analysis.

That is why Sievert started the project with a two-day workshop, in which we analysed the existing challenges and requirements together with the customer. The initial focus was not on selecting tools, but on determining the actual needs for the data strategy and understanding fundamental architecture concepts such as data lake, data warehouse, data lakehouse and data hub – always in the context of the company’s specific processes.

After identifying the use cases from controlling and sales that offered the greatest potential for implementation in a proof of concept, the suitable architecture approaches were discussed. It quickly became clear that the most urgent need was an analytical data platform in the form of a scalable cloud data warehouse. To connect the source systems, Sievert relied on flexible data modelling with Data Vault on the one hand and on a service infrastructure with data pipelines on the other.

Taking into account requirements such as extensibility, operating costs, functionality and innovation potential, a combination of the Talend Data Integration Platform, Snowflake and Power BI as the visualisation tool emerged. With this setup, we started the pilot implementation and delivered a data mart including the associated data loading and modelling within five weeks. This confirmed both the technical feasibility and the immediately usable business value of the solution.

The technical foundation built in the process was already so complete that it could be used almost without extension for the full implementation of the controlling and sales use cases. In addition to the automated and timely delivery of operational data, the data pipelines brought the long-awaited quality assurance and standardisation of KPI calculation.

Another focus was the creation of interactive dashboards for detailed contribution margin analyses and commission calculations. Creating individual reports based on self-calculated KPIs became a thing of the past – another step towards more transparency and trust in the data foundation.

To handle the variety of data, a centralised, maintainable “golden dataset” was built that maps planned and actual data at different levels of granularity. The aim was to provide consistent and comparable KPIs for several Power BI reports. The challenge was to model numerous fact tables with varying temporal and content depth consistently. The solution had to scale both for aggregated KPIs and for fine-grained individual documents.

For this purpose, a star schema data model was developed which, by separating fact and dimension tables, provides a stable and extensible architecture. Specially adapted DAX measures were used to handle the differences in granularity. They allow flexible calculations across different time levels and analysis contexts – with consistent report logic. The implementation was carried out in close cooperation with the business departments and resulted in a robust, reusable model for all sales-related analyses.

Based on this model, the introduction of role-based access control (row-level security, RLS) was also prepared. This means content can be tailored to specific target groups and restricted to individual regions or sales staff in compliance with data protection requirements. Access is controlled via single sign-on (SSO) in combination with Active Directory (AD), Microsoft Azure and Snowflake.

The newly built KPI system now reaches more than 100 users, and the number is growing. New operational data is being integrated, and the functionality is being extended to further business areas such as product management, production and logistics. The potential of the data foundation for data science and machine learning is being explored for new use cases.