Data Mess to Mesh: Strategic Lessons from the Field

Abstract

As the significance of data and artificial intelligence escalates, firms endeavor to adopt a more data-driven approach. Nonetheless, existing data infrastructures are not inherently structured to accommodate the magnitude and breadth of data and analytics applications. Indeed, current designs often do not provide the anticipated value associated with them. Data mesh is a socio-technical, decentralized, and distributed framework for business data management. The notion of data mesh remains fresh and is devoid of empirical inputs from the field. An grasp of the driving elements for implementing data mesh, the related problems, tactics for execution, its commercial implications, and possible archetypes is lacking. To rectify this deficiency, we do 15 semi-structured interviews with industry specialists. Our findings indicate that organizations encounter challenges in transitioning to federated data governance linked to the data mesh concept, the redistribution of responsibility for the development, provision, and maintenance of data products, and the understanding of the overarching concept. We propose many implementation tactics for enterprises, including the establishment of a cross-domain steering unit, monitoring data product use, achieving early fast wins, and favoring small, specialized teams that prioritize data products. While we recognize that firms must tailor implementation techniques to their own requirements, we also identify two archetypes that provide more detailed recommendations. Our results consolidate views from industry experts and provide academics and professionals with first instructions for the effective implementation of data mesh.