Regis Deshayes, Head of Data Quality, and Stefan Hilbert, Data Scientist at Zeiss Group, speak with Derek Strauss, Chairman Gavroshe, and Editorial Board Member, CDO Magazine, about improving data quality through data mesh architecture, the importance of data sharing, the end-product approach to data quality, effective use cases, and understanding the business impact of data quality.
Zeiss is a 177-year-old German manufacturing company that specializes in industrial metrology, medical device manufacturing, vision care, and semiconductor manufacturing technology.
When asked how data mesh architecture facilitates the improvement of data quality, Hilbert states that it supports the federated model of data management as compared to the siloed approach.
This enables data sharing and after sharing, as a data producer, one has an incentive to look at the data quality, he adds. From a consumer standpoint as well, one must ensure an understanding of data quality and getting high-quality data, says Hilbert.
Therefore, it is critical to improve the data quality and make people aware of the data quality. Adding to that, Deshayes states from the governance perspective, data mesh architecture provides a centralized landscape of data domains where people who understand the business usage of data are given responsibility and ownership.
Such people have the relevant knowledge to define what is fit for purpose, and who has the highest interest in having improved data quality. Thus, this targeted ownership is the key driver to improve data quality, he asserts.
In continuation, Deshayes mentions the data quality methodology perspective. He maintains that data mesh architecture helps organizations with a methodology approach that is also referred to as the end product approach.
Since data quality cannot be measured everywhere, the end product approach enables measuring data quality at data products which are created at the data domains consumption layer at the end of the data journey.
Further, as the program evolves, data quality can be measured closer to the source. Zeiss does it with a data domain ingestion layer which is closer to the source, says Deshayes. Here, data mesh architecture offers a clearer starting point for a data-oriented program at scale.
When asked about use cases, Deshayes mentions creating a web portal for customers to directly access information, which did not turn out well due to poor information quality. This worked as an eye-opener for many in the company, which led to the creation of an MVP in the form of a dashboard to provide data quality measurements.
It also assessed the validity of key data fields displayed on the portal and transparency around data fields between the portal backend and respective data sources.
Thereafter, the regions willing to clean data could go ahead with the dashboard. Moving forward, Hilbert discusses a major enterprise-wide project called “data readiness” which took place while changing the CRM system.
The project addressed two major questions:
How to ensure that only high-quality and fit-for-purpose data has migrated to the new CRM system?
How to ensure that the data stays of high quality after the migration?
Furthermore, Hilbert advises using the rule of 10 to understand the business impact of data quality and customer data. He elaborates saying that if it costs US$ 1 to check a record for duplicates, it might cost US$ 10 to clean up later, or even US$ 100 over time.
Therefore, through this project, the organization provides tools to find and remove duplicates to find data quality issues and help monitor data quality. Also, to ensure that data quality is maintained, the organization must establish responsibilities of data owners and stewards, and train them on data quality.
In conclusion, Hilbert mentions establishing a community of data owners and stewards where they can connect and share knowledge.
CDO Magazine appreciates Regis Deshayes and Stefan Hilbert for sharing their insights and data success stories with our global community.
