Big-Time Data Terminology

The term "big data" is so broad that when commenting about data management and the industry, it's better to consider topics such as data quality, data consistency or deriving value from data – or at least discuss matters in those terms.

A presentation given this past week by Pierre Feligioni, head of real-time data strategy at S&P Capital IQ, defined "big data" as "actionable data," and sought to portray big data concerns as really being about four issues: integration, technology, content and scalability.

Integration, particularly the centralization of reference data, is the biggest challenge for managing big data, as Feligioni sees it. While structured data is already quite "normalized," unstructured data, which can include messaging, emails, blogs and Twitter feeds, needs to be normalized.

Unstructured data is fueling rapid exponential growth in data volumes, justifying the name "big data." Data volumes are counted in terabytes (1,000 gigabytes), or even petabytes (1,000 terabytes). When it comes to unstructured data at those levels, central repositories that can collect and normalize data – and coordinate it with structured data – are a must, Feligioni contends.

Technology and scalability the building blocks necessary to make such central repositories functional, as he describes it. Natural language processing and semantic data approaches are also being applied. "The biggest challenge is understanding documents and creating analytics on top of this content, for the capability to make a decision to buy or sell," says Feligioni.

Scalability makes it possible to process more and more information, and is achieved through new resources, such as cloud computing, which carry their own issues and require additional decisions [as described in my column two weeks ago, "Cloud Choices"].

Everything that Feligioni calls part of "big data" actually revolves around getting higher quality data by incorporating more sources and checking them against each other to keep that data consistent. It's also about creating new value from data that can be acted upon by trading and investment operations professionals.

So, whatever buzzwords one uses, whether "big data" or sub-categories under that umbrella, what they are really talking about is quality, consistency and value. Other terms just describe the means.