Jeff Jonas, an IBM Fellow and the chief scientist of Big Blue’s Entity Analytics Group, dropped by theCUBE at Information on Demand 2013 to share his unique perspective on Big Data variety and the importance of errors in delivering better insights.

Kicking off the discussion, Jonas recalls working with a European bank that sought to find customers on Twitter in order to map brand perception and improve complaint handling. The task was proven impossible after an internal survey found that identification accuracy would be far too low to justify the expenditure, an obstacle that Jonas says could have been avoided altogether by simply engaging clients directly.

Achieving high data quality – and pursuing optimal paths – is critical to gaining visibility into the bottom line, but inaccurate information can also prove valuable when managed appropriately.

“Bad data starts to become your friend,” Jonas tells theCUBE hosts John Furrier and Dave Vellante. “The spelling errors, the transposition errors, it turns out you want to remember that – you want to remember that natural variability. The best example I have of this is when you research Google and it says ‘did you mean this,’ it’s not looking in a dictionary; it’s remembering people’s errors. If it didn’t remember the errors, it would not be so smart.”

Organizations seeking to drive meaningful business insights must have a broad observation space that encompasses multiple data sources throughout the supply chain, Jonas continues. Exemplifying with another customer story, he reflects on a government client that turned to IBM for help in tracking technology trends, but only had Twitter as a source of information.

Click the video below for the full interview, including more on asking the right questions and uncovering value in Big Data.