Traditionally, the data discovery process was a critical prerequisite to, yet a distinct aspect of, formal analytics. This fact was particularly true for big data analytics, which involved extremely diverse sets of data types, structures, and sources.
However, a number of crucial developments have recently occurred within the data management landscape that resulted in increasingly blurred lines between the analytics and data discovery processes. The prominence of semantic graph technologies, combined with the burgeoning self-service movement and increased capabilities of visualization and dashboard tools, has resulted in a new conception of analytics in which users can dynamically explore their data while simultaneously gleaning analytic insight.
Such analytic exploration denotes several things: decreased time to insight and action, a democratization of big data and analytics fit for the users who need these technologies most, and an increased reliance on data for the pervasiveness of data-centric culture.
According to Ben Szekely, Vice President of Solutions and Pre-sales at Cambridge Semantics, it also means much more–a new understanding of the potential of analytics, which necessitates that users adopt:
âA willingness to explore their data and be a little bit daring. It is sort of a mind-bending thing to say, âlet me just follow any relationship through my data as Iâm just asking questions and doing analyticsâ. Most of our users, as they get in to it, theyâre expanding their horizons a little bit in terms of realizing what this capability really is in front of them.â
Expanding Data Discovery to Include Analytics
In many ways, the data discovery process was widely viewed as part of the data preparation required to perform analytics. Data discovery was used to discern which data were relevant to a particular query and for solving a specific business problem. Discovery tools provided this information, which was then cleansed, transformed, and loaded into business intelligence or analytics options to deliver insight in a process that was typically facilitated by IT departments and exceedingly time consuming.
However, as the self-service movement has continued to gain credence throughout the data sphere these tools evolved to become more dynamic and celeritous. Today, any number of vendors is servicing tools that regularly publish the results of analytics in interactive dashboards and visualizations. These platforms enable users to manipulate those results, display them in ways that are the most meaningful for their objectives, and actually utilize those results to answer additional questions. As Szekely observed, oftentimes users are simply: âApproaching a web browser asking questions, or even using a BI or analytics tool theyâre already familiar with.â
The Impact of Semantic Graphs for Exploration
The true potential for analytic exploration is realized when combining data discovery tools and visualizations with the relationship-based, semantic graph technologies that are highly effective on widespread sets of big data. By placing these data discovery platforms atop stacks predicated on an RDF graph, users are able to initiate analytics with the tools that they previously used to merely refine the results of analytics.
Szekely mentioned that: âItâs the responsibility of the toolset to make that exploration as easy as possible. It will allow them to navigate the ontology without them really knowing theyâre using RDF or OWL at all…The system is just presenting it to them in a very natural and intuitive way. Thatâs the responsibility of the software; itâs not the responsibility of the user to try to come down to the level of RDF or OWL in any way.â
The underlying semantic components of RDF, OWL, and vocabularies and taxonomies that can link disparate sets of big data are able to contextualize that data to give them relevance for specific questions. Additionally, semantic graphs and semantic models are responsible for the upfront data integration that occurs prior to analyzing different data sets, structures and sources. By combining data discovery tools with semantic graph technologies, users are able to achieve a degree of profundity in their analytics that would have previously either taken too long to achieve or not have been possible.
The Nature of Analytic Exploration
On the one hand, that degree of analytic profundity is best described as the ability of the layman business end user to ask much more questions of his or her data in quicker time frames than he or she is used to doing so. On the other hand, the true utility of analytic exploration is realized in the types of questions that user can ask. These questions are frequently ad-hoc, include time-sensitive and real-time data, and are often based on the results of previous questions and conclusions that one can draw from them.
As Szekely previously stated, the sheer freedom and depth of analytic exploration lends itself to so many possibilities on different sorts of data that it may require a period of adjustment to conceptualize and fully exploit. The possibilities enabled by analytic exploration are largely based on the visual nature of semantic graphs, particularly when combined with competitive visualization mechanisms that capitalize on the relationships they illustrate for users. According to Craig Norvell, Franz Vice President of Global Sales and Marketing, such visualizations are an integral âpart of the exploration process that facilitates the meaning of the researchâ for which an end user might be conducting analytics.
Emphasizing the End User
Overall, analytic exploration is reliant upon the relationship-savvy, encompassing nature of semantic technologies. Additionally, it depends upon contemporary visualizations to fuse data discovery and analytics. Its trump card, however, lies in its self-service nature which is tailored for end users to gain more comfort and familiarity with the analytics process. Ultimately, that familiarity can contribute to a significantly expanded usage of analytics, which in turn results in more meaningful data driven processes from which greater amounts of value are derived.