What do the most popular consumer web properties such as LinkedIn, Facebook and Google have in common? Apart from being among the highest-flying tech stocks, they all use proprietary versions of graph technology to help them continuously deliver relevant information through easy-to-use interfaces, while continuing to astound and amaze with new features and functionality at a rate unmatched by traditional enterprise-class applications.
What are graphs good for and how do they work? Graphs belong to the family of NoSQL databases, as opposed to “traditional” relational databases such as Oracle and IBM DB2. NoSQL databases are often described as schema-on-read, as opposed to schema-on-write relational stores that require strict definition of a data model, which must be first updated before new data attributes can be stored.
Graph databases therefore allow for faster, more effortless loading of data, and offer agility when new types of data need to be captured. They are also better at determining how different data points are related or how similar they are, allowing "real-world" relationships between people, products and their activities to be quickly uncovered and captured.
Much has been written about the growing popularity of graph databases for the enterprise and how they are being used. The technology powering LinkedIn’s Economic Graph, Facebook’s Social Graph and Google’s Knowledge Graph, were of course built from the ground up to give them the flexibility to extend and scale beyond the original limitations off-the-shelf graph databases, such as scalability issues, and complex data access methods.
Indeed the popularity of graph databases with Hadoop, Spark and the continuous stream of new technologies designed to handle ever increasing volume, variety and velocity, have changed the way data can and should be managed. With hardware and storage costs more affordable than ever, and elastic pay-as-you-go options via the cloud, the temptation might be to just dump everything into so-called big data lakes, and let data scientists comb through the information. However, stitching together all of the pieces required for a complete end-to-end offering can still be a complex undertaking.
In an ever changing and evolving market, business and regulatory conditions continue to dictate that enterprises must continue to adapt. So while off-the-shelf graph databases have improved significantly over the years, competitive advantage still boils down to how quickly contextual applications can make use of the stored data. The ubiquity and popularity of LinkedIn and Facebook have also raised the stakes for applications in general. Coined by the industry as the “Consumerization of IT,” there is a growing expectation that enterprise applications must be simple to use, without significant training, and be accessible anytime, anywhere from any device. This leads business teams to question why they are still saddled with enterprise applications that required labor-intensive manual data entry, jumping between apps to get the complete view they need, and having to sift through complex patterns of information.
There is also the yin of governance, security and reliability of the data, the responsibility of IT teams, which leads them to multi-million dollar master data management projects. And the yang of deriving relevant insights in a timely manner, the goal of business teams, which fuels the popularity of business intelligence and visualization tools.
Such BI and analytics tools are powerful and sophisticated. They enable business analysts and data scientists to independently gain fast access to information. Convergence of data at scale from multiple sources is now possible, however the data is not guaranteed to be clean and accurate prior to analysis. Having business analyzing unreliable data may lead to wrong conclusions, which is of major concern to IT teams.
Although many of these tools claim to be for non-data scientists, they are still beyond the skill of frontline field teams as they go about their day-to-day operations. Any insight is obtained using a standalone analytics tool is also separate from the operational execution that ends up taking place in separate siloed applications downstream. This disconnect means enterprises still can’t accurately measure their ROI, and rely on surveys and conjecture. What’s missing is the ability to tie actual actions taken back to insights to form a closed-loop system that proves and justifies outcomes.
The path to a true enterprise Commercial Graph is more than just a graph database itself. Just like LinkedIn and Facebook it’s about the complete package. Relevant insights and recommended actions have to be delivered through enterprise-class data-driven applications that are both analytical and operational.
With LinkedIn, you get visualizations and insights, in context to your profile and your goals. LinkedIn not only suggests jobs you may like based on your experience and title, but also connections. As an application, it also allows you to take action right there and then by clicking to apply for that job or, sending that connection request.
In an equivalent data-driven application for the enterprise, a sales person may receive recommendations on actions they should take to meet this quarter’s sales targets that they can execute on, without having to switch to another application.
Data-driven applications must be backed by the new Big Data tools and databases available to bring together and make data reliable across external and internal sources, while auditing and clarifying how data is used from sourcing all the way to delivered result.
The added bonus of a closed-loop is that recommended actions can continuously improve through deep learning, making the next set of predictions and prescriptive actions even more likely to yield the desired outcomes.
Companies must get the right data in the hands of the right users and at the right time. Successfully deploying a new breed of enterprise data-driven applications requires a modern data management foundation that includes an enterprise Commercial graph. Rather than spending time managing data lakes, they should be delivering immediately consumable, purified insights direct to their business teams. Giving them the ease-of-use and agility needed to be competitive in today’s market.