Another consideration that leads to Data Products is the use of data outside of governance. At this point, a data product can represent a tactical approach, as Yaad Oren, Global Head of Research and Innovation at SAP, suggests: “If data sets are used across teams without strict governance, clearly defined processes or clear responsibilities, companies are recommended to develop a data product. Data products that are anchored in a unified database eliminate silos, create a common understanding of the data and establish secure, standardized access to it.”
A third way to use data products strategically is to develop them in a reusable form for defined customers in order to achieve efficiency gains. If such a data product requires combining multiple data sources, the vision statement and qualified business value are particularly important. Christopher Zangrilli, vice president of technology strategy at compliance service provider Vertex, explains: “Leaders should ask themselves whether the data optimizes cycle times, improves decision accuracy or mitigates compliance risks in order to assess business impact. When governance, change management, quality and measurement are integrated from the start, data products transform from experimental tools to strategic resources.”
2. Standardize data products
Products in the supermarket come with packaging that contains a detailed list of ingredients, an expiration date and a price. Those responsible for data governance should take a similar approach – and standardize how data products are defined, cataloged and managed. How and why, explains Abhi Sharma, co-founder and CEO of AI provider Relyance AI: “Every modern data product should clearly answer four questions: Where does the data come from, how is it transformed across systems, who or what uses it, and what governance obligations are involved? Without this consistent context, teams develop functions based on data they don’t fully understand.”
Although food manufacturers publish their ingredients and label them with an eye on risks such as allergic reactions, only a few document the origin of their raw materials and the route they take from the producer to the retailer. However, when it comes to developing data products in strictly regulated industries, it may be necessary to do exactly that – and capture the data lineage. This is particularly important when it comes to standardizing data sources for AI applications.
Carter Page, Executive Vice President of Research and Development at dev specialist Astronomer, knows what will happen otherwise: “Without data lineage, teams are operating blindly and governance degenerates into reactive troubleshooting. However, if teams can understand where the data comes from, how it was transformed and which systems rely on it, updates can be predicted, the right pipelines tested, affected stakeholders notified and fundamental changes documented. Before incidents arise as a result.”
3. Manage data products sustainably
For APIs, applications, or AI models, lifecycle management requires setting a release schedule for optimizations, bug fixes, and other necessary updates. When it comes to data products, however, several related disciplines come together, as Ulf Viney, EVP of Engineering at AI data specialist Precisely, explains: “In order to manage the life cycle of data products, you need versioning, testing, structured deployments and stakeholder communication.”
