“Without alignment, data science projects fail to drive outcomes.” So cautions Kjell Carlsson, Ph.D., author of Best Practices: Scaling Data Science Across The Enterprise, a report from industry analysts at Forrester Research. Kjell identifies a company’s IT team as stakeholders whose support is critical to the success of data science projects, and we concur with Kjell’s research.
Comprising three interlocking components of people, business processes and systems, the operating model provides insight into why IT involvement is critical to project success. The goal of data science is to create new artificial intelligence as a source of either incremental or disruptive innovation. The former is achieved when machine learning models are embedded in business processes to improve the performance of tasks that previously were completed by human workers. Disruptive innovation aims to create entirely new processes, products, or services that were previously not possible but are now enabled by AI. To successfully change a process or create a new one requires consideration of the people operating the process and the information systems supporting it. Much of the success of interplay between people and systems comes down to the diligent work of IT professionals.
The first step in the machine learning life cycle is to define project objectives. Data management specialists from IT play an enabling role. They ensure that the project team understands what data is currently available and, just as importantly, helps them understand where data deemed necessary for success is not currently available. Active participation of IT professionals during this first step of the life cycle can provide the project team with opportunities to investigate the viability of acquiring data from external sources. This may lead to a project being put on hold or canceled before money and time are wasted.
Once the project team establishes that the necessary data is available, they can move to step two where they acquire and explore all of the relevant data for use in machine learning. Here, the data engineering team, typically located within the IT department, plays a critical role in preparing the data to make it suitable for analysis.
While machine learning models are physically implemented in the final step of the life cycle, planning for their implementation is best initiated in step one when the project objectives are established. Here, IT professionals can raise such questions as, "Where and how will models be embedded in existing systems?", "Does the changed business process require that user interfaces of applications currently in use be changed?", "Will running new models on existing hardware have any impact on service levels already established with user communities?" and "Do we have sufficient hardware and software or a plan and budget to procure these?". Including such questions in the initial planning allows time to be allocated throughout the project timeline to address each challenge and refine potential solutions during the project life cycle. For information on how DataRobot’s automated machine learning platform reduces the effort and timelines usually required for effective model deployment from weeks or months down to hours, see https://www.datarobot.com/wiki/machine-learning-model-deployment/.
About the Author:
Bob Laurent is a Sr. Director of Product Marketing at DataRobot. Prior to DataRobot, he ran product marketing at Alteryx, where he was responsible for driving awareness and growing a loyal customer base of empowered data analysts. He has more than 20 years of marketing, media relations, and telecom network engineering experience with Fujitsu and NYNEX (now Verizon). Bob resides in Dallas with his wife and two boys, and holds a Bachelor of Science degree from Clarkson University, plus an MBA from New York University's Stern School of Business.