Data is a very valuable asset in today’s business world. Businesses that understand, collect and access better data, are set to become more competitive and profitable. This is the reason why data is characterized as “the new oil”. Nowadays the development of data-driven products and services is propelled by advances in digital technologies such as big data and artificial intelligence (AI). Nevertheless, organizations that jump on the bandwagon of big data and AI, face significant data management challenges as well. Specifically, they must deal with the quantity, quality, and integrity of the data that they use to deploy projects, products, and services. In this direction, modern data-driven enterprises have a lot to gain if their data comprise proper metadata and exhibit FAIR (Findable, Accessible, Interoperable and Reusable) properties.
The FAIR data principles are a set of guidelines that are aimed to support the reusability of digital assets. FAIR data are widely used in research and scientific contexts. However, there are many cases where they can provide benefits for businesses as well. At the very least, they encourage data standardization, while ensuring that datasets and software code published by organizations are made available under a ‘free’ open source/copyleft-type license. Furthermore, the implementation of FAIR data principles can help enterprises understand how their complex data are interconnected and integrated. They also encourage the adoption and use of open and transparent data management systems rather than proprietary and costly platforms.
FAIR Properties and Business Benefits
Each of one of the FAIR properties can deliver business benefits to modern enterprises with data-intensive products and services:
- Findable: Findable Data are datasets that can be discovered through querying catalogues. They are stored in standards-based formats and have an accompanying metadata record that describes their purpose, content and usage restrictions. The findable property helps organizations to ensure that their data are properly organized so they can be easily found and used when needed. Findable data are properly organized to enable data consumers to discovery and access them whenever needed. The metadata of a findable dataset make it easily discoverable for both humans and computers. In the case of computer programs, machine-readable metadata are essential for the automatic discovery of datasets. Findable data facilitate the first step of any data processing pipeline, which is to discover the data needed. In this direction, a good data management practice is to use a search engine in order to accelerate the querying of large volumes of data with rich metadata.
- Accessible: Accessible data facilitate users to gain access to the datasets they need based on proper authentication and authorization functions. Data access must be based on formats that make sense for end-users. As a prominent example, access to geospatial data must be accompanied with proper attribution and information about how the datasets have been created. As another example, if a user wants to access patient medical records from a third-party provider, they must know how much information can be retrieved from this external source as well as whether patient consent has been obtained.
- Interoperable: This property eases the integration of data with other pieces and sources of data. Specifically, the data needs to interoperate with applications or workflows for analysis, storage, and processing. Interoperability has different levels and dimensions. The simplest level of interoperability concerns the adherence to a common standards-based format to facilitate data processing and data engineering across data sources. Nevertheless, there are more complex levels of interoperability such as semantic interoperability. The latter adds semantic metadata to the various datasets to ensure their uniform and unambiguous interpretation across applications. Overall, data interoperability requires the use of well-known and open formats and software, as well as the use of relevant standards for metadata. In this direction, community agreed schemas, controlled vocabularies, thesauri, keywords, and ontologies can be used.
- Reusable: Reusable data comprise well described metadata, which facilitate the replication and integration of datasets in different settings. Metadata are descriptions of data that describe what kind of information it contains and how it should be used. They help users understand what they’re looking at when accessing or using the data set. Metadata can be properly designed to be reusable to enable different consumers to access the same data set without having to go through the low-level work of metadata generation.
The implementation of FAIR principles can also drive data quality improvements. Data quality has different dimensions such as: (i) Accuracy i.e., whether the information provided within a dataset is correct or not. Accuracy involves making sure that any errors made during input into the system, are corrected before they becoming added to final record; and (ii) Completeness i.e., whether all information required for an accurate description has been recorded in the database. Completeness ensures that no relevant piece of information has been left out or ignored when creating a record for an entity in the database. FAIR data management pays emphasis on the creation of complete data assets with rich metadata, which boosts their quality.
On the downside, the development of FAIR datasets can be a challenging and costly task for modern enterprises. For instance, it requires complex data curation, data engineering and metadata specification activities, which increase the overall data management costs. This asks for use cases that justify the data “FAIRification” process and yield a positive Return on Investment (ROI) for FAIR data owners.
Nowadays, data science is a critical and an integral part of the modern business framework. It helps organizations make sense of their ever-growing volumes of data, enhance their existing business processes, improve investment and risk analysis, develop new products and services, as well as to bring in new and profitable customers. Hence, organizations have begun to realize that the quality of the data they use to make their core decisions may impact their bottom lines. That is exactly what has given birth to this movement for good practices for managing datasets such as the FAIR principles for data management. FAIR data management is gradually becoming an essential part of maintaining effective data management practices. It’s easy to get bogged down with policies and procedures, yet it is also incredibly important to gather feedback and to make sure that your customers are satisfied. FAIR data management gives you a real opportunity to do that, while also improving public perception about your organization. By applying FAIR data management strategies, organizations can improve their ability to cultivate a successful competitive edge over their competition. Based on the tangible business benefits that come from revenue generation and data sharing, organizations are likely to see these efforts as worthwhile. It might cost a little time and some money to make your data more FAIR, but there are already many use cases that make it well worth the effort.
