Various portals have been developed to provide an easy way to discover and access public research data sets from various organizations. Data sets are made available with descriptive metadata based on common (e.g., OGC, CUAHSI, FGDC, INSPIRE, ISO, Dublin Core) or proprietary standards to facilitate better understanding and use of the data sets. Provenance descriptions may be included as part of the metadata and are specified from a data provider's perspective. These can include, for example, different entities and activities involved in a data creation flow, such as sensing platforms, personnel, and data calculation and transformation processes. Moving beyond the provider-centric descriptions, data provenance may be complemented with forward provenance records supplied by data consumers. The records may be gathered via a user-driven feedback approach. The feedback information from data consumers gives valuable insights into application and assessment of published data sets. This might include descriptions about a scientific analysis in which the data sets were used, the corrected version of an actual data set or any discovered issues and suggestions concerning the quality of the published data sets. Data providers might then use this information to handle erroneous data and improve existing metadata, their data collection and processing methods. Contributors can use the feedback channel to share their scientific analyses. Data consumers can learn more about data sets based on other people's experiences, and potentially save time by avoiding the need for interpreting or cleaning data sets. The goals of the study are to capture feedback from data users on published research data sets, link this to actual data sets, and finally support search and discovery of research data using feedback information. This paper reports preliminary results addressing the goals. We provide a summary of current practices on gathering feedback from end-users on research data portals, and discuss their relevance and limitations. Examples from the Earth Science domain on how commentaries from data users might be useful in practice are also included. Then, we present a data model representing key aspects of user feedback. We propose a system architecture to gather and manage feedback from end-users. We describe how the core PROV model may be used to represent the provenance of user feedback information. Technical solutions for linking feedback to existing data portals are also specified.