How to Solve the Data Scientist Skills Shortage

Posted by Matthew Littlefield on Thu, Jan 28, 2016 @ 10:00 AM

Find me on:

Big_Data_Analytics_framework-2.jpgWe at LNS Research have spent the past several years researching, writing, advising, and consulting on Big Data Analytics in the industrial sector. As part of our Metrics that Matter research we have shown how manufacturers believe areas like business model transformation and asset value.

Click here to speak with Matthew

We have also discussed how Big Data Analytics need to be deployed as part of an Industrial Internet of Things (IIoT) platform as it is tightly coupled with Cloud, depends on connectivity for a successful deployment, and users expect it to be included as part of next-generation applications.

Despite this, the large majority of industrial companies are still not successfully using Big Data Analytics technology. For Big Data Analytics to become a reality in the industrial sector, we as an industry need to make many additional changes, including:  

  • Improve connectivity
  • Address data model shortcomings 
  • Transform system architecture
  • Develop a simple framework to understand and deploy traditional and next-generation Big Data Analytics

The Data Scientist Shortage

However, even if all of this happens, it will not address a fundamental limiting factor: the number and expense of data scientists. There will never be enough data scientists to go around and not every industrial company is going to be in the position of companies like IBM, SAP, GE most notably, and Siemens more recently to hire 100’s or 1,000s of data scientist from areas like Silicon Valley or Cambridge, MA. Furthermore, not enough will be available to build shared services Big Data and Digitalization services for internal and external users.

Moreover, even if every manufacturing company could afford to take this approach, would it be the best approach? Should we all be trying to teach data scientists the industrial sector to support Industrie 4.0 and Smart Manufacturing initiatives? Or instead should we be teaching industrial leaders and subject matter expertise Big Data and Digital technologies?

Luckily, we have a relatively recent example that can lend some insight into what might be the right approach. It wasn’t that many years ago that Lean and Six Sigma were the trends, and nearly 25% of industrial companies were adopting programs in support of Six Sigma. Although over time there were many different Six Sigma consulting and technology companies launched, for many of the early years it was an entirely homegrown approach were industrial companies themselves developed the tools, methodologies, training, and certification programs to support these efforts.

Building on What was Learned from Six Sigma Initiatives

It is important to note that as the momentum for Six Sigma grew, and companies like Motorola, GE, and others launched programs and initiatives. These companies didn’t go out and hire hundreds or thousands of statistical or financial geniuses. Instead, these companies took manufacturing and engineering subject matter experts and taught enough targeted statistics, finance, and project management skills to launch, lead, and complete successful Six Sigma projects. The point was to teach manufacturers, suppliers, and customers enough statistics to reduce variability in processes and understand the benefits of these reductions, not require a mastery of multi-variate statistics.

Then, as momentum grew even further and the benefits became undeniable, an entirely organic certification program emerged around Six Sigma training, with brown, black, and master black belts all claiming different levels of proficiency and documented project savings. As the years passed, these certifications even became important hiring criteria and resume builders, cementing the importance and value of Six Sigma in our industry.

Big Data Analytics as a Program not Tool

Industrial companies need to take a similar approach to Big Data Analytics today. We as an industry need to organically develop an industry specific Big Data Analytics tool set so that we can teach today’s subject matter experts enough data science to enable digital transformation, not vice versus. For this change to take shape, we will need to see some early adopters take this approach and prove the value to others; which we are already beginning to see. Cisco now has 80 data scientists in its Supply Chain organization, all entirely homegrown through a 2 year internal training program and Cisco is not alone. Others, like Merck, are taking a similar approach with a home grown data science team that is starting to provide strong ROI.

Just as Six Sigma was not sold to the industry by technology and consulting companies, but instead developed and promoted by leading industrial companies; the same has to happen for Big Data Analytics and for this to be successful the following needs to be established by these leading industrial companies:

  • A common language and branding for Big Data Analytics jargon, including: tools, methodologies, training, and certification.
  • An industry specific framework for mapping use cases to proven Big Data Analytics tools.
  • A common method for identifying use cases, estimating potential financial benefits, and tracking actual financial benefits to justify future work.
  • A common training and certification program to help develop and communicate the creation of new human capital within industrial sector.

LNS Research is hoping a few leading industrial companies are up for the challenge. As these companies move down this path we hope to continue to be an advocate for and supportive of this change and be a strong communication channel for this organic development within the industry to help extend and accelerate this trend.

NEW Research Spotlight on strategies and recommendations for minimizing risk through a migration away from monolithic, single-plant MOM architectures through exploration of Cloud and IIoT technologies that are advancing in manufacturing today.Manufacturing Operations Management

Tags: Manufacturing Metrics, Big Data, IIoT