The three big trends I keep talking about in data science and machine learning are these: the explosive adoption of open source technologies, the expanded hype and adoption around non-traditional data science (citizen data science, augmented analytics, automated machine learning), and operationalization. My colleagues Erick Brethenoux, Shubhangi Vashisth and Jim Hare recently published a timely and vital report on “How to Operationalize Machine Learning and Data Science Projects.”
Data science operationalization is most simply defined as the application and maintenance of predictive and prescriptive models. Both clients and vendors are placing an emphasis on the importance of moving data science out of a prototype environment and into a state of production and continuous improvement. This figure shows the concurrent cycles of model development and model operationalization that occur within a working data science team:
The figure can be overwhelming at first glance, so I highly recommend giving the full report a thorough read.
I’ll be the first to admit that operationalization is not the perfect word. It has eight syllables. You don’t have to be a speech pathologist to know that making a hard pivot from an “L” sound to a “Z” sound is tricky. There is a running joke within Gartner’s data science team that every analyst warms up for a day of client inquiries by saying operationalization five times in the mirror. For months, I have been trying to come up with a better term that might stick in the market (in the proud analyst tradition of “name it and claim it”). These efforts have thus far been highly unsuccessful.
Despite its linguistic clunkiness, operationalization seems to be the best option to encompass the people, processes, and technologies that go into moving data science out of the ivory tower and into the real world. The cumbersome pronunciation of the word is what drew my mind back to an episode of The West Wing named after another famously difficult-to-pronounce word: “Shibboleth.”
In the episode, President Josiah Bartlett is faced with the difficult decision of whether to grant asylum to a group Christian Chinese refugees (the parallels to modern events are obvious, but would be best discussed in a different forum). President Bartlett, struggling to determine the authenticity of the group’s faith and story of persecution, references a story from the Book of Judges. The Israelites used the ability to pronounce the word “shibboleth” correctly to distinguish true Israelites from impostors sent across the River Jordan by their enemies. Think of it as a password or a secret handshake to prove membership and fluency.
At the end of the episode, President Bartlett realizes he doesn’t need a shibboleth to believe that the refugees need asylum. Their willingness to risk their lives in pursuit of a better life is proof enough, but the president is reassured anyway when the group’s priest looks him in the eye and without prompting says “shibboleth.”
You can watch the complete scene here.
Author’s note: the second season of The West Wing is a near-perfect stretch of television and this episode contains perhaps the show’s most heartfelt moment (fans will know this as “Charlie and the carving knife”). Highly worthy of your next binge.
Like President Bartlett at the end of the episode, I don’t necessarily need to hear a client or vendor say operationalization to know they are serious about data science. But I do need to hear them talking about the things that go into making data science a working reality: business value, transparency, scalability, release and activation. The focus of data science needs to go beyond the mere creation of analytical assets. Even if operationalization efforts are only in their infancy, a data science and machine learning initiative should be justified by the end goal of sustained positive influence on business actions.
So, to sum up my title and thesis, operationalization is a difficult to pronounce term that proves membership and conversance in the world of data science. In other words: shibboleth.
The Gartner Blog Network provides an opportunity for Gartner analysts to test ideas and move research forward. Because the content posted by Gartner analysts on this site does not undergo our standard editorial review, all comments or opinions expressed hereunder are those of the individual contributors and do not represent the views of Gartner, Inc. or its management.