Rightsizing data science: How to architect analytics around the business need
The following is a guest article from Wes Chaar, chief data and analytics officer at Catalina.
Large, successful companies are increasingly embracing the value of the data science function and its potential to deliver powerful insights, better outcomes and personalized customer experiences.
Growth in the application of data science is evident in new hiring trends, increased spending in areas like technology and data science, as well as an overall cultural shift and recognition of the need for data-inspired decision making at the highest levels of leadership for critical strategic direction.
But the glorification of data science in today's enterprise can be a two-edged sword.
The growing wave of big data applications and activities, along with plenty of media hype and publicity, can sometimes supercharge the expectations and demands of business leaders.
Executives can push for what's shiny and new — the biggest and most complex — data science models to address their perceived business needs.
Perhaps there's also a desire to gain bragging rights among executives as the leader who implemented the most advanced data science techniques in their industry.
While the growing stature of data science and an improved data culture represent positive trends for the field, all the excitement can drive data science teams to fall into a trap I call "Data Science Overdone."
It's an unhealthy condition that leads data science teams to over-engineer solutions and lose focus of a few key guiding principles as they architect their data science processes, methods and solutions.
In lieu of practical, effective approaches, enthusiasm among business leaders, combined with the curiosity and ambition of highly trained and well-qualified data science teams, can ironically sometimes push data scientists toward over-hyped and unnecessarily complex approaches — often to the detriment of the business.
Instead, data scientists should remember the principal of Occam's razor and right size their projects by embracing the simplicity of practical and effective designs.
What follows are three suggestions to help get the data science function on track in rightsizing their data science solutions:
1. Data science team
Experience is the best teacher and experienced data science teams have seen solutions fall short of implementation over time.
Strong, experienced and confident data science leaders understand the value of rightsizing data science rather than creating solutions that excite and astound because of the advanced techniques applied.
2. Senior leaders and cultural understanding
Executives and others outside of the data science function commonly represent the ultimate consumers of analytics solutions. Their exposure to new terms, methods and trends can result in buzzword-fueled requests for the latest and greatest in data science work.
While it's exciting to see the thrill and enthusiasm, analytics clients can lose sight of the importance of focusing on their business needs rather than the latest trends.
It's incumbent on data science leaders to help their clients understand the difference between cutting edge and the surest, most cost-effective, route to solving business challenges.
3. Framing the problems or opportunities
Data science teams need to take the time to work with the business to frame the business problem and, especially, to understand the intended application of the results.
Experience has shown that this is an often-neglected, yet vitally important step in the analytics project lifecycle.
Proper framing will increase the likelihood of implementation of results, whether that includes learnings that lead to business improvements, or integration of a data science solution (in the form of an algorithm, process, method, etc.) into a scaled system for real-time decision-making or data-driven automated processes.