Eight insights get to the heart of big data value and point to a future in which we synthesize and make sense of vast data stores.
Now and then I find myself thinking about the big principles of big data; that is, not about Hadoop vs. relational databases or Mahout vs. Weka, but rather about fundamental wisdom that frames our vision of "the new currency" of data. But maybe the new oil better describes data. Or perhaps we need a new metaphor to explain data's value.
Metaphors aren't factual or provable, but they do illuminate certain truths about topics of interest. They make complex concepts understandable, much like the following set of quotations I've collected that you could say explain basic big-data principles. I'll offer eight truths about big data -- you've surely already bought into at least a few -- ordered roughly chronologically. Then I'll take a look ahead at a "future truth."
1. "Correlation is not causation."
We hear this over and over (or at least I do). I learned one version of the underlying fallacy, when I was in college studying philosophy, as post hoc ergo propter hoc, or "after the thing, therefore because of the thing."
You can read a smart take in the O'Reilly Radar blog, where in "The vanishing cost of guessing," Alistair Croll observes: "Overwhelming correlation is what big data does best... Parallel computing, advances in algorithms and the inexorable crawl of Moore's Law have dramatically reduced how much it costs to analyze a data set," creating a "data-driven society [that] is both smarter and dumber." Bottom line? Be smart and respect the difference between correlation and causation. Patterns are not conclusions.
2. "All models are wrong, but some are useful."
Accidental statistician George E.P. Box wrote this in his 1987 textbook, Empirical Model-Building and Response Surfaces. Box developed his thoughts on modeling, which very much apply to big data, over the length of his career. See in particular the article "Science and Statistics," published in the Journal of the American Statistical Association in December 1976.
3. Big data knows (almost) all.
If you don't already, it's time to accept Scott McNealy's 1999 statement, "You have zero privacy anyway... Get over it." McNealy was cofounder and CEO of Sun Microsystems, quoted in Wired magazine. Examples of big data's growing invasiveness are plentiful: Analysts' ability to infer sex and sexual orientation from social postings and pregnancy from buying patterns; the on-going expansion of vast, commercialized consumer-information stores held by Acxiom and the like; the rise of Palantir and Riot-ous information synthesis; the NSA Prism vacuum cleaner.
4. "80% of business-relevant information originates in unstructured form, primarily text, (but also video, images, and audio)."
I wrote this in a 2008 article, although as I said then, this bit of pseudo-data factoid dates back to at least the early 1990s. It's a factoid because it is far too broadly drawn to be precise; as far as I know, it's not derived from any form of systematic measurement ever performed. Still, per statistician Box, "80% unstructured" is a useful notion, even if not precisely correct. Whatever number works for you, text and content analytics belong in your toolkit.
5. "It's not information overload. It's filter failure."
Clay Shirky made this observation at the September 2008 Web 2.0 Expo in New York. Corollaries of Shirky's filter observation are truisms such as, "More data does not imply better insights," which happens to be one I made up. But don't overdo it; avoid what Eli Pariser terms "the filter bubble," an inability to see beyond what automation makes immediate.
6 Tools to Protect Big DataMost IT teams have their conventional databases covered in terms of security and business continuity. But as we enter the era of big data, Hadoop, and NoSQL, protection schemes need to evolve. In fact, big data could drive the next big security strategy shift.
Big Data Brings Big Security ProblemsWhy should big data be more difficult to secure? In a word, variety. But the business won’t wait to use it to predict customer behavior, find correlations across disparate data sources, predict fraud or financial risk, and more.