This post is an excerpt from the O’Reilly ebook The Rise of the Knowledge Graph co-authored by Sean Martin, Dean Allemang, and myself. If you’d like to learn more about knowledge graph and how it stitches together the concepts in this post, please check out the ebook as well as the other posts in this blog series.
A data fabric is more than just a design concept for enterprise data architecture; instituting a data fabric in an enterprise constitutes a familiar change in attitude about how data is perceived and treated in the enterprise. Why is it familiar? Because this change is in line with prevailing attitudes for what is required for digital transformation as a whole. Digital transformation changes how the business operates. The adoption of a data fabric is the manifestation of that change in enterprise data management.
Early in The Rise of the Knowledge Graph, we considered the plight of a data strategist faced with a business leader who expects an integrated data experience like they are accustomed to seeing on the web. Given what we know now about the data fabric and knowledge graphs, what advice can we give them? How should they proceed? What will drive the enterprise to make the required cultural shift, and how can an appropriate data strategy make the path smoother?
Motivation for a Data Fabric
The executives in our enterprise want to have an integrated experience when consulting business data. But what drives this desire? What is lacking in their current data experience of using data warehouses, data lakes, and other piecemeal enterprise architecture approaches?
The speed of business in a modern enterprise is accelerating. As products move to just-in-time everything, new products, services, and even business models are being developed at an increasing pace. Companies acquire one another, consolidating business plans and moving into new markets. The faster a company can move and adapt, the more successful it can be.
As the enterprise enters these new areas, how can it manage all the new information it needs to carry out business efficiently and effectively? To be blunt, how can the enterprise know what it is doing, when what it is doing changes so quickly?
Conventional techniques for managing enterprise data are not keeping up. No longer is it sufficient to build a new application for each new business model or product line. Data has become the key driver of the business, and is more durable than applications. Each new product, service, business area, or market brings with it new demands on data.
The profit centers of the business don’t want to be slowed down by application development. New products? New regulations? New business models? The business needs to keep up with these things. The business demands a digital transformation that provides flexibility, extensibility, and agility in its data architecture along with business alignment. For a growing business today, a data fabric is no longer a luxury, but a necessity. This is the situation our data strategist finds themself in.
Timing of the Data Fabric
Since the executive is telling our data strategist they want Google-like access to their data and the board is asking questions about digital transformation of the CEO, we know that the business stakeholders are ready to make some big changes. But is the enterprise ready to begin weaving its data fabric?
To be clear, there is no need to delay due to availability of the necessary technologies. The best technology for assembling a data fabric is a knowledge graph, and knowledge graphs are already successfully used by enterprises in many industries. But the transformational switch in an enterprise to a data fabric isn’t a simple matter of creating a new application, even a knowledge-based one; it requires a cultural change in how the enterprise relates to its data. This suggests an iterative approach to building a data fabric.
To take your first step toward building a data fabric in your enterprise, I highly suggest reading The Rise of the Knowledge Graph if you haven’t already. You’ll be introduced to the technology and concepts that support a knowledge graph approach. Have you experimented with some of that technology yourself? You can download some of the wide array of free software for managing knowledge graphs, or inventoried some of the resources in your field.
The biggest obstacle you will face, moving from an awareness that your enterprise needs a change to actually building a data fabric, will be resistance from parts of the organization. Your application developers see the problem as just another application that needs to be built, and not a sea change in how the enterprise sees data as a resource. Data scientists will focus on the questions they have in front of them, and will not want to concern themselves with the big picture. It is, of course, impractical to build a comprehensive enterprise data fabric before reaping any of the benefits of having one. So how can you proceed? We have a few suggestions.
The Rise of the Knowledge Graph can provide you with one tool for informing your colleagues about data fabrics and knowledge graphs. These are not far-fetched ideas bubbling in an academic lab (though there is a strong academic foundation behind the technologies that support the knowledge graph), but real technologies that are widely in use today in enterprises. We have included examples from many industries, covering a range of data capabilities. We hope you can find some parallels in your own organization, as it is always more persuasive to use local examples. We have also provided a bit of guidance for what a data fabric looks like: it is distributed, it is connected, and it includes reusable metadata. These are the principles that make a data fabric work at scale.
Remember that nothing succeeds like success; you will have to deliver something that provides real business value to bring these ideas home. One good approach is to start small, but pick a problem to solve that delivers true business value, so that when it succeeds, it is noticed within the organization and you can build on that success. But how do you make your first application part of a data fabric before the fabric exists? Or, to take the metaphor perhaps a bit too seriously, how many stitches do you have to knit before your yarn becomes a fabric? See how Bosch, Merck KGaA, and Deloitte got started.
Use the principles in The Rise of the Knowledge Graph as a guide to building your first data fabric application. When you organize the application’s data, don’t organize it just for that application; organize it for the enterprise. Represent it as a graph, and give the entities in the graph global names (using URIs). Find data assets that already exist in your enterprise, and reuse them. Build your application around the data, not vice versa. Think of each data set you use as a potential product that someone else might reuse. Represent your metadata explicitly, using the same principles. This is the beauty of the knowledge graph: while it is the basis of a large, scalable enterprise architecture like a data fabric, it can also be used for a single application. In short, build your first application as a fledgling knowledge graph in its own right. By succeeding with something that is small but that successfully delivers meaningful business value, you attract positive notice to the overall approach. Most importantly, you also earn the right to extend the application or develop a peer application with adjacent data sources using the same methodology, and thereby can continue to prove its efficacy. As you have learned, the knowledge graph is ideal for this kind of expansion.
Just as a journey begins with a single step and a fabric starts with a single stitch, a data fabric begins with a solution to a single business problem—a solution that provides business benefits in its own right, but that also forms the first stitch in something much larger. You’ve got what you need to get started, and your business is waiting. The next application you build will be the first stitch in your new data fabric.
This completes An Integrated Data Enterprise blog series. Find all earlier posts here:
- The differences between Data Fabric, Data Mesh, Data-centric revolution, and FAIR data
- Why a data fabric wins out over traditional architecture
- How knowledge graph technology satisfies the requirements of a data fabric