Research data services are becoming increasingly attuned to current and envisaged user requirements. RRD content and geographic coverage are important, since this fundamentally shapes the nature and range of hypotheses that can drive social investigations. A successful data service should develop and engage with a community of practice around existing and prospective data assets, with likely benefits to all stakeholders. Numerically, most benefits will accrue to the many RRD users, but data services can also avail specialist users of the tacit knowledge necessary to hone raw (or less pre-processed) smart data to more specialized research tasks. Additionally, whilst data owners may not have panoptic awareness of the content and coverage of their own data (particularly with respect to ‘less often heard’ groups), knowledge exchange embedded in the data service can signpost potential shortcomings in metadata. McGrath-Lone et al. (2022) set out the requirements for effective provision of RRD from administrative sources. Their core considerations extend from creation through documentation, maintenance and service delivery of RRD under FAIR data principles, as elaborated upon in Figure 3-9. There are four components to service provision.
First, creation of smart RRD requires vision and breadth. The early pilot studies discussed in Section 2 may originate in commonalities of interest in problem-solving that are very restricted in space, time and scope. Such pilots are useful to establish proof of concept but are unlikely to meet the needs of the broader constituency of research users. Community-wide multilateral licenses should seek the widest possible (e.g. national) coverage and include content relevant to the greatest number of users. A strategic view of smart RRD creation and maintenance also provides that acquisitions are amenable to linkage with minimal or no geographic aggregation – for example to an integrative smart data spine that readily enables cross-classification and new or updated RRD following new data acquisitions. Whilst it can be more challenging to secure multiple data licensing agreements for the full extent of datasets, ambition in acquisition has the greatest potential to drive new RRD creation.
Figure 3-9 The anatomy of an RRD service (after Longley et al., 2024)
Second, smart RRD require adequate documentation. Crucially, many owners of smart data are drawn from the private sector and operate in markets in which they have no monopoly of provision. As a result, data are likely to be biased in content and may be geographically restricted in coverage. This scenario requires documentation in quantitative and qualitative terms – the former with respect to available framework data, such as censuses and property gazetteers, and the latter to confirm how and why the data were generated in terms of the mission statements of the data providers. User experiences can also be added as testimonies of suitability for specific research purposes, and to flag potential RRD extensions or refinements.
This leads to the third consideration of curation of RRD, which extends documentation through dynamic updating of the underpinning data infrastructure of RRD, including but not limited to the addition of new data. It includes procedures of internal validation through longitudinal profiling of underpinning infrastructure, and external validation with respect to framework data and other ancillary sources. These procedures enable improved harmonization of the multiple smart data sources used to create smart RRD and suggests potential new RRD products that the research community might find useful.
Fourth, innovation-led data services must render accessible the pipeline of smart RRD to a large and diverse research user community. Proliferation of data sources creates a requirement for gateways to data catalogs that are accessible to users and that provide clear and streamlined procedures governing access to smart RRD. Ideally data should be available on open access licenses, but issues of ownership, service funding and licensing will typically necessitate additional access protocols – for example providing for data owner clearance in circumstances in which there are potential commercial conflicts of interest.
The final consideration addresses highly skilled requirements for RRD enhancement. Smart data are at the same time data rich and plagued with detail that most researchers will deem extraneous. Innovation-led data services should selectively discard unnecessary detail to improve database manageability, without unduly restricting the range of research questions that the raw data might be used to address. A successful data center is thus both technocratic and research focused, balancing the opportunity costs and benefits of prioritizing different RRD products.