Beauty In — Beauty Out
(OR: GIGO meet BIBO)
Anyone who has spent some time in the world of data knows the acronym GIGO, short for Garbage In, Garbage Out. In other words, the value of any information that comes out of a data collection process is only as good as what went into it.
Input error, missing data, few inputs can all skew results, sometimes in ways that accumulate and compound false outcome predictions. Recent election results are a powerful demonstration of how error can creep into analysis and yield bad information.
Across industries, more and more recognition is being given to the value of getting the first steps right. Manufacturing, in particular, has become obsessed with the accuracy and quality of each step in the work to convert raw materials into a finished product. Business schools, such as the University of Pennsylvania’s Wharton School, have developed entire departments to focus on process, information and quality. In fact, it was a business school professor of mine in an Operations and Information Management class that drove home for me the direct relationship between quality at the end of a process and the quality of every step leading up to the finished product.
To make his case, he showed the class an old advertisement from a well-known German automaker. In the ad, a stern-looking man in a lab coat stared into the camera. In his hands were a small hammer and a fine file. He had a jeweler’s loupe over one eye. The text stated that the last step of the manufacturing process ended with a team of engineers such as this guy. His job was to fix any mistakes.
My professor described this as as an admission of “failure.”
A successful quality process is one that does not rely on a team at the end of the line to discover and correct errors that should never have been made in the first place. Attention to detail along the “value chain”, the series of steps that convert raw materials into a product, ensures the best results in the end.
The process of converting data collected into useful information also follows a value chain. The final product, information, is only as good as the analysis of the data. Most importantly, that analysis and the information it produces, relies on the quality of the initial data collected: Garbage In, Garbage Out.
Tremendous effort is being invested by data intensive industries to sort and analyze an ever-growing set of inputs. Big data platforms such as IBM’s Watson and analytical tools such as Palantir are, rightfully, getting a lot of attention. All too often, though, these systems are the technological equivalent of the engineer at the end of the automobile production line.
Massive amounts of computing power and analysis could be saved if the initial data was collected in an error-free, structured and digital format initially. Garbage In, Garbage Out meet Beauty In, Beauty Out.
By making the initial data collection process simple, error-free and beautiful, all subsequent steps in the data value chain are simplified and improved. While my company makes a fantastic product that helps governments with this first step, every data process owner can benefit from considering the advice of my professor and pay attention to quality every step of the way, starting at the beginning.
