Business intelligence tomorrow… and what it means for today
Business intelligence (BI) is quite possibly the single point of constancy in the ever-evolving data ecosystem. Regardless of which technologies emerge or architectural shifts occur, the foundation of data’s overall utility for most organizations will almost always be the ability to serve business users.
Developments in this domain tend to resonate throughout the data landscape as a whole—much more so than other aspects of data management do. The movement toward self-service was born with BI. BI typifies the advantages of cognitive computing via natural language queries, search-based analytics, and statistical AI recommendations.
Still, BI’s crowning accomplishment is more than just exemplifying the democratization of AI, the ubiquity of hybrid and multi-cloud developments, or even the accessibility and reliability that provides governed self-service across business units.
BI’s greatest achievement is in extend-ing the self-service movement beyond individual visualizations, dashboards, and reports to comprehensive solutions in which “BI vendors look at the opportunities and challenges presented by serving many different customers with their singular platform,” acknowledged Charles Schaefer, Tableau senior manager of competitive intelligence.
By merging each component of the self-service experience into a comprehensive platform, contemporary BI centralizes:
- Metadata management: By providing a holistic means of managing all metadata relevant to business users and their tools for working with data at scale, modern self-service BI platforms streamline one of the critical prerequisites for consistently being able to reuse data for timely analytics.
- Semantics: Self-service repositories are relatively worthless without a well-defined, clear understanding of the data’s meaning, which is best facilitated in terms business users comprehend. According to Vijay Anand, MicroStrategy VP of product marketing, effective BI platforms are underpinned by a semantic layer that is not just a siloed way of consuming information but consolidates all of the sources into a singular semantic graph.
- Data cataloging: Enterprise data catalog tools are required for any uniform means of managing data. They’re influential for classifying information and strengthening the governance protocols necessary to prevent self-service from burgeoning into data chaos. Enterprise data catalogs are “really designed to provide a governed layer on top of, for example, a data lake, so you can on-board data, you can manage it, and make it available for consumption in basically a marketplace paradigm,” Mike Potter, CTO of Qlik, revealed.
- Data models: All-encompassing, self-service BI platforms also account for the plethora of data modeling concerns in standardized ways, aligning data of different formats, structures, and schema. “Data sources themselves need to be described and so does the kind of metadata you’re going to use: a name, a region, or some other kind of property,” explained Irene Polikoff, CEO of TopQuadrant. “That’s the definition of the schema that you would be using to enrich and describe those datasets for self-service.”
By unifying all of these aspects of data management for expedient, trustworthy access to data in a single platform, BI empowers much more than analytics. Tomorrow, it’ll serve as the launching point for the most advantageous means of managing data for all users (and use cases), including edge computing, application building, data strategy, hybrid and multi-cloud deployments, and alternative, unstructured data sources.
Natural language search
One of the defining characterizations of post self-service BI is the incorporation of the various dimensions of natural language processing (NLP). “NLP will be really impactful to BI because it allows anyone to perform queries on their data in the same way they might search for information in a search engine,” Schaefer noted. “The technology interprets their question and translates it into an analytical query.” The immediate effect of leveraging BI in natural language is that it puts the power of analytics in the hands of a broader user base that’s not necessarily data-savvy. Although NLP is the overarching term for interacting with IT systems in natural language and accounts for basic semantic understanding, additional NLP capabilities of BI platforms involve:
♦ Natural language generation (NLG): The converse of NLP, NLG evokes responses to queries in natural language. Some platforms utilize NLG so that “any chart or graph within a dashboard can be translated into text-based narratives” Anand said. NLG gives visualizations a verbal component.
♦ Natural language interaction (NLI): NLI is considered a combination of NLP and NLG. It typifies the conversational interfaces that natural language delivers to BI and is implemented via chatbots in some instances in which, “You ask a question in natural language and it gives you an interactive response where it provides analytic responses and visualizations,” Potter said. “It allows you to iterate, to interact, and to disambiguate your question.”
♦ Natural language understanding (NLU): NLU is a sub-category of NLP focused more on intention and understanding of the words used in natural language, as opposed to their semantic meaning.
The entire suite of natural language is influential in what’s termed “augmented analytics”—the use of various dimensions of AI to effectively “augment the analyst by automating and simplifying certain tasks and allowing people to focus on the richer analytics experience, their ability to ask questions, and to continue exploring their data,” Schaefer denoted. Various forms of machine learning abet augmented analytics in a number of ways. According to Anand, a recurring use case for this technology with holistic BI platforms is to “create usage-based recommendations and personalized insights” predicated on previous data discovery or analytics results. The collection and analysis of telemetry and usage behavior “complement the semantic module,” Anand said, and also inform machine learning results in this use case. Additionally, machine learning is influential in helping business users connect to data sources and “know which databases to access, which tables need to be combined to answer their question, and how to join those tables together,” Schaefer posited.