An industrial data pipeline architecture described supports annotation of raw data streams into scalable and durable cloud-level storage. A dynamic data mapping process collects industrial data from data tags that have been annotated with metadata indicating that the tags are to be made available for visualization, and connects data from these tags to a data aggregator object of a node-level or edge-level visualization tool. The data aggregator object dynamically adjusts its data collection footprint to include the annotated data tags, which are dynamically associated with the physical layout of the industrial equipment in the visualization. Visualization objects for a specified plant configuration act as data context containers that can be added to a plant visualization panel, and allows the annotated tags to be selected to build visualizations of selected data items.