Along with the many topics at Predictive Analytics World SF this week representing the leading edge of analytic sophistication, there were also presenters speaking to the blocking and tackling necessary to nurture analytic thinking and scale adoption within their organization.
Since many organizations are facing these same analytic adoption and scale problems, I’d like to highlight some useful recommendations from a sampling of these excellent talks.
Data Science is a Team Sport
There is a shortage of data scientists, there are challenges with helping business teams understand how predictive analytics can help them, and data science teams need to understand how they can improve the business outcome with an actionable analytic result. All the while keeping IT in the loop so deployment is fast and feasible.
Tina Owenmark of Cisco presented how they have a central data science team with a diverse skill set, that leverages a standard approach of decision modeling and CRISP-DM to ensure business understanding, effective analytic model development and successful deployment.
The importance of a diverse team was underscored by Professor J. Bryan Bennett of Northwestern University who recommends a “DataScienceStein” approach, building a data science team from people currently on staff with a variety of skills: Data Analyst, Business Analyst, Predictive Modeler, Programmer, Visualization Specialist, and Subject Matter Expert.
You Can’t Hire Your Way Out of The Problem
Just hiring more data scientists isn’t going to solve the blocking and tackling challenges of predictive analytics adoption and scale. Data Scientists are in short supply, and processes and approaches have to scale, too.
Professor Bennett’s presentation provided some hard numbers wrt the peak demand for Data Scientists resulting in the current supply and demand problem. Training current staff and creating diverse teams is a successful complementary approach.
Paul Speaker of The Dow Chemical Company, in his presentation, “An Industrial Revolution for Analytics” echoed the problem, saying once a data science team is successful, demand is 20X supply. Just hiring more data scientists is not going to get you there. In order to scale, processes have to change. Paul emphasized:
- Project-Based Analytics does not scale. Data science teams have to move from an artisanal approach of building a specific analytic for a specific ask, to an industrial approach.
- Getting away from the typical reactive analytics project process is key, by modeling the business at a fundamental scale. This enables a component approach to analytics models for analytics assembly.
Shared Business Understanding
There were many conversations among attendees about the need to engage with business partners and deliver an actionable result from analytic projects, with measurable business value. Yanai Golany of Verizon offered tips and tricks in his keynote. Understanding the business problem involves breaking down the problem, with understanding the symptoms of the problem often leading to a better result. Sharing the business metrics with the analytics team and creating a climate for change will result in faster acceptance and adoption of analytics.
Chemere Davis at Wells Fargo shared how the data science team developed a sophisticated communications strategy to build a bridge from analytics to the business.
Overall I took away that a diverse team – especially when supported by a standard approach like decision modeling that breaks the analytic problem down, defines business metrics and deployment context upfront, and drives a shared understanding and actionable outcomes – eases the challenges of the data scientist shortage and positions organizations for analytic success.
Making better data-driven decisions is the fundamental driver of the rush of investments in data infrastructure and analytical tools. My thanks to Dr. Eric Siegel of Prediction Impact and the team at Rising Media for assembling a wide range of analytic experts covering technical topics and the blocking and tackling required to improve business decisions with data science.
To learn more about decision modeling:
- The Analytics Value Chain: Operationalizing Analytics with Decision Modeling
- Bringing Clarity to Data Science Projects with Decision Modeling: A Case Study
- Analytics Teams: 6 Questions To Ask Your Business Partner Before You Model