This paper addresses the challenge of assuring data quality in high-frequency Internet-of-Things (IoT) streams while migrating to a hybrid edge–cloud architecture. We demonstrate that moving a subset of data-quality procedures—trust-metric calculation, outlier detection, and data-contract validation—from the cloud to edge devices markedly lowers end-to-end latency and reduces cloud load. After surveying existing cloud-centric and edge-centric quality-control solutions, we reveal their limitations: static placement of analytic modules and lack of support for dynamic workload drift. We introduce the concept of edge-oriented data-quality control, in which validation tasks are continuously re-assigned according to real-time network bandwidth and CPU utilisation. A prototype based on Apache Flink implements the proposed scheduler. Experiments with an industrial testbed (300 000 messages/s) show a 37 % reduction in alert latency and a 46 % decrease in cloud CPU consumption compared with a fully cloud-based pipeline. The paper discusses strengths, weaknesses, applicability boundaries, and security threats, and outlines future work on adaptive model selection at the edge, multimodal stream support, and formalised data-quality contracts.

Keywords: data quality control, streaming processing, Internet of Things, cloud computing, Apache Flink, Apache Kafka, anomaly detection, dynamic offloading