We are excited to share a new, outstanding achievement accomplished by TH-OWL (OWL University of Applied Sciences and Arts), within the context of TwinAir project!
The article “A Survey on Platform-Level Data Quality: From Visibility to Controllability” has been published at “IEEE Xplore” Journal.
The authors of the article are: Tony Rosset; Lukasz Wisniewski; Stefano Scanzio and its respective DOI is the following: 10.1109/ACCESS.2026.3676924
The abstract and the keywords of the publication can be found below:
Abstract
Modern data platforms integrate heterogeneous data sources—ranging from IoT sensors and APIs to databases and event streams—and deliver information to diverse analytical and operational applications. Ensuring data quality in such environments is increasingly complex, as platforms often possess only partial visibility into both data origins and usage contexts. Most traditional data quality frameworks implicitly assume full knowledge of the source schema or consumer requirements— assumptions that rarely hold in large-scale, multi-tenant ecosystems. This survey consolidates three decades of research on data quality and reinterprets it through the constraint of partial visibility. Over seventy academic and industrial sources are organized into ten thematic clusters encompassing foundations, governance, streaming, metadata, ML/AI, profiling, integration, storage, economics, and trust. By analyzing these clusters through the lens of platform capabilities, the paper exposes a systematic gap between what platforms can observe and what they can control. The surveyed literature is synthesized into a visibility–controllability taxonomy that distinguishes four regimes of platform-level quality management: inferential, interface, adaptive, and operational. This taxonomy explains why many existing approaches fail to scale to platform settings and identifies the conditions under which quality guarantees remain feasible despite incomplete information. The study highlights open challenges, including quality estimation without ground truth, integration of observability and governance, adaptive thresholds, and distributed accountability. The results establish a foundation for inference-driven, context-aware data quality management—enabling platforms to ensure trustworthy Quality of Service (QoS) even when full transparency of sources and consumers is unattainable.
Keywords:
Data quality, data platforms, observability, partial visibility, governance, quality of service.
You can reach the full article following this link: https://www.mdpi.com/2813-4168/3/4/29