Balancing Old and New
Earlier this month I attended two different conferences in which I got a glimpse of the future. The first was the Minitab Insights Conference, held in Leesburg, VA (Oct 10-11). Many presenters covered topics that would be familiar to all Green Belts and Black Belts – Design of Experiments, Measurement System Analysis, and Sampling. Other speakers covered topics new to me – Regression Trees, MARS, and Random Forests. For the record, all of these are algorithms used in machine learning (AKA artificial intelligence).
TMAC’s Smart Manufacturing Conference was the second symposium which I attended. It was held in Dallas (Oct 22). For anyone not familiar with Smart Manufacturing (also known as Industry 4.0), it combines real-time integration of the operating system (equipment and people) of a factory with the information system (network, software, internet, etc.) to connect the supply chain from suppliers to customers. It can incorporate additive manufacturing (3D printing), collaborative robots, artificial intelligence, cyber security, and virtual reality. By the way, most of the concepts shared in this conference are equally applicable to service companies.
Both conferences featured knowledgeable speakers discussing technical topics who explained key concepts with a mix of theory and real-world examples. The most futuristic tools and methods shared reminded me of science fiction writer Arthur C. Clarke’s Third Law that …
“Any sufficiently advanced technology is indistinguishable from magic.”
Here are some general takeaways from these two conferences:
- Computing Power and Advanced Methods Allow Practitioners to Solve New Types of Problems – Many of the advanced methods under the broad umbrella of Artificial Intelligence have been around for decades. However, they have NOT been previously available to most practitioners because the computing power required was cost prohibitive. Now, with Moore’s Law still holding true, the computing power available in a laptop computer makes these methods feasible. This allows the application of billions of calculations in a fraction of the time required with older computers. Another key enabler that is seldom mentioned is the increasing availability of inexpensive sensors. These sensors can be used to count items, measure temperature, check thickness, assess color, etc. on a real-time, ongoing basis to a degree not seen in the past. For that matter we are seeing such sensors (or RFID or bar codes) adapted in an increasing number of service environments like hospitals. As for the type of problems that AI can solve, think of challenges with thousands of data points and dozens of inputs. One example shared: Using machine learning to solve the challenge of determining the most efficient routing in a job-shop to minimize process lead time.
- Traditional Statistical Analysis isn’t Going Anywhere – Most business processes don’t generate enough data to warrant using machine learning or other advanced methods. This is especially true for service companies. And even for many manufacturers the amount of data generated is still relatively low. Going forward, I feel strongly that the majority of business problems can still be solved with a basic tools – process maps, fishbone diagrams, scatter plots, histograms, Pareto analysis, control charts, and process capability. Some problems do require more advanced methods – hypothesis testing, ANOVA, regression, perhaps even DOE. These tools are standard topics in the LSS curriculum. Bottom line: The tools and methods learned by Lean Six Sigma practitioners over the past 20 years are still effective for solving the vast majority of business problems.
- Change Management Remains Vital – Regardless of whether advanced methods or more traditional tools are used, the management of change must be done carefully – or the change will fail. This was a point brought home to me by a consultant from another firm I’ve known for over 20 years. He’s used traditional lean and six sigma methods since the 1990s. His company is now expanding into artificial intelligence. Determining the ‘right’ solution to a business problem is one thing. Convincing everyone affected by the change to agree to it is another thing altogether. Or to use the terminology we share with all our students at TMAC: R = Q x A (Results = Quality of the Solution x Acceptance Level of the Solution).
Finally – there is much still to be learned about these new tools and methods. A fundamental question is how to know when traditional methods can be used effectively versus when more advanced methods (artificial intelligence) are warranted. That is a topic for a future blog.
